How Long Does the Brain Take to See One Object? A Clockwise Search Trick Finally Measures It

How long does the human visual system need to process a single object? The question sounds deceptively simple, yet it has frustrated psychologists for decades, because every reaction time measurement is contaminated by processes that have nothing to do with seeing: the brain must register that something appeared, decide what to do about it, select a response, and finally command the fingers to move. A new study published in Attention, Perception, & Psychophysics by Talke Michaelsen, Markus Janczyk, and Heinrich R. Liesefeld of the University of Bremen now offers a clever escape from this impasse, using a visual-search paradigm in which the order of inspection is brought under strict experimental control. The result is what may be the cleanest behavioral estimate yet of how long the mind takes to process each individual stimulus.

The traditional approach, presenting one stimulus and clocking the reaction, fails for several reasons. Motor execution time can be trimmed away by measuring when the movement begins, and backward masking can truncate early perceptual processing, but even the most sophisticated masking studies confound the true processing duration with the time needed to notice that a stimulus appeared at all. The brain’s readiness to process incoming information fluctuates rhythmically, so on most trials some time passes after stimulus onset before the system is even prepared to begin. Worse, a mask interrupts only the earliest perceptual stages; categorization can continue based on partial information extracted before the mask erased the iconic trace. And by presenting just one stimulus at a time, these methods ignore the visual system’s massively parallel capabilities, biasing estimates toward overestimating processing time.

For forty years, researchers hoping to showcase the visual system’s full potential turned to visual search, varying the number of items in a display and reading the processing time per item off the slope of the reaction-time function. The logic, formalized by Bergen and Julesz in 1983, assumes that observers search at random, never inspect the same element twice, and stop the moment the target is found. Under this serial, self-terminating model, target-absent slopes directly reveal the time per nontarget. But modern research has demolished these assumptions. Searchers are guided by preattentive priority maps, skip unpromising stimuli, quit prematurely on target-absent trials, and frequently reinspect items they have already rejected. The number of actually processed stimuli is simply unknown, which makes the classic set-size slope an invalid yardstick.

Set-size manipulations carry a second, more prosaic flaw: they are confounded with eccentricity or density. If the display grows while item density stays constant, targets land farther from fixation in peripheral vision, where stimuli are genuinely harder to process. If eccentricity is held constant instead, crowding increases local feature contrast and target salience, which almost every current model of search identifies as a major driver of search times. Either way, the slope is contaminated by factors that have nothing to do with the intrinsic processing duration of a single item.

The Bremen team’s solution, the systematic-search paradigm introduced by Michaelsen and colleagues in 2024, sidesteps the problem entirely by making the number of processed stimuli an experimental variable. Twelve stimuli are arranged like the hours of a clock, and participants are induced to search in strict clockwise order starting at the 1 o’clock position. A single lure, which matches the target definition but always sits two to four positions after the true target, discourages shortcuts and keeps the search systematic. Because the target can appear only at positions one through seven, the target’s position tells the experimenters exactly how many nontargets were processed before it. Median response times regressed on target position yield a strikingly linear function, with R-squared values above 0.99 for virtually every participant, and the equation is elegantly simple: response time equals an intercept plus the slope multiplied by the number of nontargets already inspected.

The crucial theoretical move is that the two parameters of this function dissociate cleanly. Everything unrelated to nontarget processing, from registering display onset and initiating search to processing the target itself, selecting the response, and executing the button press, is absorbed by the intercept. The slope isolates exactly what the researchers want: the time needed to process each individual nontarget stimulus. The new study validates this interpretation with two preregistered experiments. In Experiment 1, thirty-two participants searched displays containing either simple stimuli, nested squares and circles with one cut-out layer, or complex stimuli with an additional, entirely task-irrelevant shape that increased visual complexity without changing informational content.

The results were as clear as a psychophysicist could wish. For simple stimuli, the mean position slope was 289 milliseconds per stimulus; for complex stimuli it rose to 383 milliseconds, a highly significant difference confirmed by a Bayes factor of 1,856. Crucially, the slope distributions for the two conditions barely overlapped, and almost every individual participant produced a steeper slope with the complex shapes, with only a single participant showing a reversed pattern. Both slopes were far above the 50-millisecond floor, consistent with the 58 to 92 milliseconds thought to be the minimum for shifting attention between items. The linear fits were almost perfect, averaging an R-squared of 0.98 in both conditions, and structured interviews confirmed that serial, one-after-the-other inspection was the strategy participants actually experienced.

Experiment 2 provided the complementary half of the validation: if the slope is a pure measure of stimulus processing, it should be immune to manipulations of the response. The same stimuli now had to be reported either with a simple two-choice button press or with a four-choice mapping in which the luminance of the outer shape selected the hand and the luminance of the inner shape selected the finger. As predicted, this manipulation left the slopes essentially untouched, at 336 versus 339 milliseconds, while inflating the intercepts from a mean of 1,156 to 1,581 milliseconds, with a Bayes factor of 1,697 in favor of the intercept difference. Harder responses cost time, but all of that cost landed exactly where the model says it should.

The authors are careful about what their measure includes and excludes. The position slope bundles in the time to disengage, shift, and reengage attention, and it may depend on how similar nontargets are to the searched-for template, since stimulus processing always has a purpose. Eye movements, with their roughly 200 to 250 millisecond minimum per saccade, do not set a floor for the observed slopes, some of which fall well below that rate, and the team argues that overt shifts of attention are as legitimately part of stimulus processing as covert ones. The estimates, averaging between roughly 290 and 380 milliseconds across conditions, sit intriguingly close to classical measurements of attentional dwell time, although the two constructs are conceptually distinct: dwell time measures how long one item monopolizes resources, while processing time measures how quickly an item is dealt with.

The paradigm does have boundaries. It works only when displays induce clump scanning, the systematic inspection of adjacent groups of stimuli that becomes efficient when targets are of low salience, and Michaelsen’s earlier work showed that when targets are highly salient, participants abandon the intended order and the position functions lose their linearity. But within its domain, the method gives cognitive science something it has never cleanly had: a behavioral dial that moves when seeing gets harder and stays still when acting gets harder. Since processing speed is considered a core facet of intelligence and cognitive aging, and stimulus-processing time indexes the information load that shapes perception, attention, and memory, a valid, preregistered, openly available measure of that fundamental quantity, with all data and code shared on the Open Science Framework, may prove one of those quiet methodological advances that reshapes an entire field.

Subject of Research: Measurement of stimulus-processing time using position slopes in a systematic visual-search paradigm

Article Title: Estimating stimulus-processing time via position slopes in visual search

Article References: Estimating stimulus-processing time via position slopes in visual search. (n.d.). https://doi.org/10.3758/s13414-026-03325-5

Image Credits: AI Generated

DOI: 10.3758/s13414-026-03325-5

Keywords: visual search, stimulus-processing time, attention, psychophysics, position slopes, systematic search, reaction time, perception, clump scanning, cognitive psychology, Bayesian statistics, visual processing