Attention is selective: change blindness and inattentional blindness

Every interface assumes that a mark placed in front of a user will be seen if it matters. Change blindness and inattentional blindness show why that assumption is unsafe, and that the failure is predictable. The strongest counter is that the measure is partly a reporting criterion, so a miss counts what a person says rather than only what was seen (Nartker et al., 2024). Change blindness is the failure to notice a large change marked by a brief disruption; inattentional blindness is the failure to notice an unexpected object while attention is committed elsewhere. Both are widely replicated outside the laboratory, and both move with the similarity of the unexpected item to what is tracked, with the difficulty of the task, and with how the change is signalled. The honest design rule is that detection depends on which channel the primary task leaves free. Unit 2 asked how the eye builds an object. This one asks why a built object is still missed.

Vision does not keep a copy of the scene

The founding measurement is Rensink, O'Regan and Clark (1997). Observers saw a photograph alternating with a modified version, separated by a brief blank field. Identifying the changes became extremely difficult even for large, repeated changes. A verbal cue made identification much faster, so poor visibility was not the cause. Changes to important objects were found sooner. The conclusion was that observers never form a complete representation of a scene.

The mechanism is the disruption rather than the occlusion. O'Regan, Rensink and Clark (1999) placed six small high-contrast shapes over a photograph, arranged so that they did not cover the change, and large changes were still missed. A brief transient normally marks a change, so anything that removes or masks it removes the change with it. Motion is therefore double-edged: an animation can carry an update into awareness, or hide one.

Work a failure in an interface. A settings panel redraws when a save completes. One option flips from off to on with no animation and no focus change, while the user is reading the confirmation below. The state that matters has changed in full view, and it is not registered. The mechanism is the missing transient. A change that carries no signal therefore waits on the user noticing by chance. The case shows why a silent update is fragile. It does not say how many users will miss it.

Change blindness is not an artefact of screens. Simons and Levin (1998) had an experimenter stop pedestrians to ask directions, then be replaced mid-conversation by a different person. Only seven of the fifteen noticed, and detection depended on whether the pedestrian shared the experimenters' social group. Feedback that asks the user to compare one state with another is therefore fragile, and a larger change does not repair it.

An unexpected object goes missing in full view

The parallel effect for objects appeared two decades before the gorilla. Neisser and Becklen (1975) asked observers to follow one of two optically superimposed films, and found that monitoring both at once was difficult and that unexpected events in the ignored film were widely missed. Selection, they concluded, follows an event rather than filtering the ignored stream.

Mack and Rock (1998) made the effect a laboratory standard. Observers judged a cross at fixation while an unexpected object appeared in the same place, and many failed to report it, which the authors read as perception without attention; Nartker and colleagues (2024) record that some of those rates exceeded 80 per cent. The best known demonstration is the gorilla. Simons and Chabris (1999) had observers count passes in a filmed game, and found that noticing the unexpected figure depended on its similarity to the other objects and on how hard the monitoring task was, and not on how close it came to the attended locations.

The miss is conditional, and the conditions are design variables

Most and colleagues (2001) varied that relation directly: the more similar an unexpected object is to the attended items, the more likely it is to be noticed. Nearly 30 per cent of observers missed a bright red cross with a unique colour, shape and motion path, on screen for five seconds. Being different is not enough, because an item is noticed when it belongs to the class that the task has primed.

Chabris and colleagues (2011) simulated a police chase and staged a fight along the route: 35 per cent of runners noticed it at night and 56 per cent in daylight, and raising the task load reduced noticing further. Wallisch and colleagues (2023) ran three high-powered experiments with 4,493 participants and varied the speed of the unexpected object. Fast objects were noticed readily and slow ones often missed, and because the fastest was also the briefest on screen, exposure did not explain the result. The authors also report that the original study's rates of 8 and 67 per cent rose sharply when its video was reused.

A miss is a process, not a trait of the person who misses

Kreitz and colleagues (2015) tested whether a miss is a fact about the person. They compared static and dynamic, central and peripheral tasks against measures including working-memory capacity and attention breadth. None predicted noticing except a weak working-memory association in the central static task, and noticing in a static task only weakly predicted noticing in a dynamic one.

Simons, Hults and Ding (2024) pooled that literature. Their meta-analysis coded 38 records, 31 of them yielding effect sizes across 74 samples, and tested 14 cognitive and 19 personality measures against noticing. It found little evidence that ability or personality predicts a miss, so the claim now rests on the field's synthesis rather than on one study.

Work the diagnostic question on one alert. A monitoring dashboard raises an alert while an operator is completing a form. On the evidence here, whether it is noticed depends on the alert's relation to the form and to the ignored material, and not on a stable trait of the operator. The case moves the remedy from retraining the person to the alert's channel and timing, since an operator who works through a screen reader meets the alert only as an announcement the interface chooses to make.

Contested question: unseen, unreported, or unheld?

Two disputes run through the unit, and neither is about whether the effects exist. The first concerns change blindness. Is a change missed because the scene was never represented in enough detail, or because it was represented and the before and after were not compared? Rensink, O'Regan and Clark (1997) argued for the first. Simons and Rensink (2005) survey a decade of work and dismiss some of the more extreme claims drawn from the effect. They hold that a core of inferences about visual processing survives, which places the dispute in the mechanism rather than the phenomenon. The representational account rests on alternation studies, and the comparison account on the same data read differently, so neither is decisively ahead.

The second concerns inattentional blindness and whether the unnoticed item is genuinely unseen. Mack and Rock (1998), and the usual reading of the gorilla, treat a failure to report as a failure of awareness. Nartker and colleagues (2024), in the largest set of inattentional-blindness studies to date, found that participants who denied noticing could still report the location, colour and shape of the stimulus above chance. Their absent trials also revealed a bias towards reporting that nothing had been noticed. On that reading the yes or no question measures a criterion rather than perception. Wood and Simons (2019) replicated two classic processing-without-noticing tests. They found support for grouping under inattention but none for word-stem priming. The report-as-awareness reading is the older and broader; the criterion reading rests on the largest dataset and on absent trials, so it has the firmer footing.

Settling either dispute is a matter of measurement. For inattentional blindness, forced-choice probes of features the participant denies seeing, run with absent trials that estimate the reporting criterion, separate sensitivity from bias. For change blindness, the test dissociates representation from comparison. It uses recognition memory for a changed feature and for an unchanged one, with confidence ratings that show which process failed.

While the dispute is open, a designer can act on what both readings allow. Put anything that must not be missed on a channel the primary task does not occupy, and mark a change with a visible transient. Treat a non-report as weak evidence and probe it. Do not train the operator harder, because noticing behaves as a process rather than a trait.

Where the argument is heading

Work on the topic is moving from demonstrating the effect to measuring it. Nartker and colleagues (2024) turn the noticing question into a signal-detection problem with a criterion to correct. Wallisch and colleagues (2023) vary the eliciting event parametrically, and report that speed, not exposure or salience, drives detection. Kreitz and colleagues (2015), and the pooled meta-analysis of Simons, Hults and Ding (2024), show that noticing behaves like a process rather than a personal trait.

The sources this unit rests on

• Chabris, C. F., Weinberger, A., Fontaine, M. and Simons, D. J. (2011). You do not talk about Fight Club if you do not notice Fight Club: inattentional blindness for a simulated real-world assault. i-Perception, 2(2), 150-153. doi:10.1068/i0436 Staged a fight along a simulated chase route: 35 per cent of runners noticed it at night and 56 per cent in daylight, with noticing falling as task load rose.
• Kreitz, C., Furley, P., Memmert, D. and Simons, D. J. (2015). Inattentional blindness and individual differences in cognitive abilities. PLoS ONE, 10(8), e0134675. doi:10.1371/journal.pone.0134675 Tested static and dynamic, spatial and central tasks against cognitive measures and found no measure predicting noticing except a weak working-memory association in the central static task.
• Mack, A. and Rock, I. (1998). Inattentional Blindness. MIT Press. doi:10.7551/mitpress/3707.001.0001 Reports trials in which observers judging a cross at fixation failed to see an unexpected object in the same place, and argues for perception without attention. It established the effect and gives the baseline that later critiques are set against.
• Most, S. B., Simons, D. J., Scholl, B. J., Jimenez, R., Clifford, E. and Chabris, C. F. (2001). How not to be seen: the contribution of similarity and selective ignoring to sustained inattentional blindness. Psychological Science, 12(1), 9-17. doi:10.1111/1467-9280.00303 Varied the unexpected object's similarity to the attended and ignored sets; in a further experiment nearly 30 per cent missed a bright red cross visible for five seconds with a unique colour, shape and motion path.
• Nartker, M. S., Firestone, C., Egeth, H. E. and Phillips, I. B. (2024). Sensitivity to visual features in inattentional blindness. eLife, 13, e100337. doi:10.7554/elife.100337 In the largest set of inattentional-blindness studies to date, participants who denied noticing a stimulus still reported its location, colour and shape above chance, and absent trials showed a bias towards reporting that nothing was noticed.
• Neisser, U. and Becklen, R. (1975). Selective looking: attending to visually specified events. Cognitive Psychology, 7(4), 480-494. doi:10.1016/0010-0285(75)90019-5 Had observers follow one of two optically superimposed films and found that monitoring both at once was hard and that unexpected events in the ignored film were widely missed. It predates the gorilla study.
• O'Regan, J. K., Rensink, R. A. and Clark, J. J. (1999). Change-blindness as a result of 'mudsplashes'. Nature, 398(6722), 34. doi:10.1038/17953 Showed that six small high-contrast shapes scattered over a photograph, deliberately not covering the change, were enough to make large changes go unnoticed. The disruption, not the occlusion, causes the miss.
• Rensink, R. A., O'Regan, J. K. and Clark, J. J. (1997). To see or not to see: the need for attention to perceive changes in scenes. Psychological Science, 8(5), 368-373. doi:10.1111/j.1467-9280.1997.tb00427.x Placed brief blank fields between alternating versions of a photograph: identifying large, repeated changes became extremely difficult, and a verbal cue helped, so poor visibility was not the cause. It is the founding measurement of change blindness.
• Simons, D. J. and Chabris, C. F. (1999). Gorillas in our midst: sustained inattentional blindness for dynamic events. Perception, 28(9), 1059-1074. doi:10.1068/p281059 In the gorilla study, detection of the unexpected figure depended on its similarity to the other objects in the display and on the difficulty of the monitoring task, and not on its proximity to the attended locations.
• Simons, D. J. and Levin, D. T. (1998). Failure to detect changes to people during a real-world interaction. Psychonomic Bulletin and Review, 5(4), 644-649. doi:10.3758/bf03208840 In a street experiment one experimenter was replaced mid-conversation by another, and seven of the fifteen pedestrians noticed; detection depended on whether the pedestrian shared the experimenters' social group.
• Simons, D. J. and Rensink, R. A. (2005). Change blindness: past, present, and future. Trends in Cognitive Sciences, 9(1), 16-20. doi:10.1016/j.tics.2004.11.006 Reviews the first decade of change-blindness research and argues that a core of inferences about visual processing survives, which locates the live dispute in the mechanism rather than the existence of the effect.
• Simons, D. J., Hults, C. M. and Ding, Y. (2024). Individual differences in inattentional blindness. Psychonomic Bulletin and Review, 31(4), 1471-1502. doi:10.3758/s13423-023-02431-x Meta-analysed 38 inattentional-blindness records, 31 with usable effect sizes across 74 samples, and found little evidence that ability or personality predicts noticing. It replaces a single study with the field's synthesis, so the unit's claim that a miss is not a trait now rests on the pooled evidence.
• Wallisch, P., Mackey, W. E., Karlovich, M. W. and Heeger, D. J. (2023). The visible gorilla: unexpected fast, not physically salient, objects are noticeable. Proceedings of the National Academy of Sciences, 120(22), e2214930120. doi:10.1073/pnas.2214930120 Ran three high-powered experiments with 4,493 participants, varying the speed of the unexpected object: fast objects were noticed readily, slow ones often missed, and because the fastest was briefest on screen, exposure did not explain the pattern. The authors also record that the original study's rates of 8 and 67 per cent rose sharply on video reuse.
• Wood, K. and Simons, D. J. (2019). Processing without noticing in inattentional blindness: a replication of Moore and Egeth (1997) and Mack and Rock (1998). Attention, Perception and Psychophysics, 81(1), 1-11. doi:10.3758/s13414-018-1629-1 Replicated two classic tests of processing without noticing with large online samples of 260 and 448 participants, finding support for grouping under inattention but not for word-stem priming.