Why do some distractions become difficult to ignore while others fade into the background? A preregistered experiment suggests that reward alone may not be enough. What people selectively attend to while learning appears to help determine which signals later gain the power to pull attention away from a task.
The finding matters because modern environments are crowded with cues designed to acquire value: notification badges, promotional colours, loyalty signals and other visual prompts can repeatedly coincide with rewards. The experiment does not test advertising or smartphone use directly, but it offers a controlled look at a more fundamental mechanism: how attention helps bind a feature to a rewarding outcome.
Testing what attention contributes to reward learning
Francisco Garre-Frutos and colleagues recruited 160 participants for an online visual-search experiment. Seven were excluded under preregistered criteria, leaving 153 participants, with 77 in a colour-report group and 76 in a location-report group. The final sample had a mean age of 21.4 years and included 117 participants who self-identified as women.
Participants searched displays for a diamond-shaped target while a uniquely coloured circle could appear as a distractor. One distractor colour predicted a high-value reward and another a low-value reward. Crucially, participants were not told about this colour-reward relationship. Instead, the researchers manipulated what they had to pay attention to: one group occasionally reported the distractor’s colour, while the other reported where it had appeared.
The task contained six blocks of 48 trials. High-value distractors appeared on 20 trials per block, low-value distractors on another 20, and eight trials contained no distractor. The reward system made fast, correct performance valuable, with points multiplied by ten on high-value trials. This meant that being captured by the high-value distractor could actually reduce the points participants earned.
The researchers analysed response times with linear mixed-effects models and separately assessed participants’ awareness of the colour-reward contingency. The study was preregistered, and its materials, data and analysis scripts were made publicly available.
A small delay revealed a clear difference between groups
Across the final sample, responses were slower when a high-value distractor appeared than when a low-value distractor appeared. Mean response time was 734.2 milliseconds in high-value trials, compared with 728.4 milliseconds for low-value trials and 684.2 milliseconds when no distractor was present.
The more revealing result was the interaction with the attention manipulation. Participants required to report the reward-predictive colour showed a value-modulated attentional-capture effect of 10.4 milliseconds, with a 95% confidence interval from 4.0 to 16.8 milliseconds. Those reporting the distractor’s location showed an estimated effect of only 1.2 milliseconds, with a confidence interval spanning zero from -5.3 to 7.6 milliseconds.
The group-by-value interaction was statistically significant (p = 0.025). Accuracy remained high, at roughly 95%, and the researchers found no evidence that the response-time pattern was simply a speed-accuracy trade-off.
Awareness did not explain the result away
A key question was whether the colour-report group had simply become more consciously aware that colour predicted reward. The two groups did not significantly differ on the study’s awareness measures. Although greater contingency awareness was itself associated with stronger reward-driven distraction, controlling for those individual differences did not eliminate the effect of the attention manipulation.
After awareness was included in the model, the colour-report group still showed a significant estimated capture effect of 7.2 milliseconds, while the location-report group’s estimate was -0.9 milliseconds and not significant. The authors therefore argue that selective prioritisation of a reward-predictive feature and explicit awareness can make partly independent contributions to learned attentional bias.
A supplementary meta-analysis placed the result alongside earlier experiments using closely related designs. Reward-driven distraction was significant when participants were encouraged, through instructions or task demands, to attend to the predictive feature: the pooled standardised mean difference was 0.36, with a 95% confidence interval from 0.25 to 0.47. When participants were uninstructed or attended to a reward-irrelevant feature, the pooled estimate was 0.08 and its confidence interval crossed zero.
What the experiment can and cannot tell us
The practical interpretation is not that every reward-associated colour will automatically hijack attention. Rather, the experiment suggests that learning which environmental features matter may depend on selective processing during the learning episode. A signal that repeatedly accompanies reward may become a stronger future distractor when attention has been directed specifically towards that signal.
That distinction is relevant to decision-making environments where people are exposed to many simultaneous cues. Retail interfaces, financial apps, games and digital platforms routinely combine visual features with feedback and rewards. The present experiment cannot establish how large or durable these effects are in those real-world settings, but it helps explain why merely counting exposures to a rewarding cue may miss an important part of the learning process: whether people actually prioritised the predictive feature.
There are also reasons for restraint. The final sample was relatively young, the task was conducted online, and the observed response-time differences were measured in milliseconds in a tightly controlled visual-search paradigm. The split-half reliability of the value-modulated capture measure was also low in this experiment, which the authors note may reflect the dual-task design. Generalising the magnitude of the effect to everyday behaviour therefore requires further work.
Even with those limitations, the experiment sharpens a longstanding idea in learning science: attention does not simply respond to what has already been learned. It can help decide what gets learned in the first place. In environments engineered around incentives and competing signals, what people are prompted to notice may be as important as the reward attached to it.
Source Information
Study Title: The role of selective attention in value-modulated attentional capture
Authors: Francisco Garre-Frutos, Miguel A. Vadillo, Jan Theeuwes, Dirk van Moorselaar, Juan Lupiáñez and colleagues
Journal: Psychonomic Bulletin & Review
Year: 2026
DOI: 10.3758/s13423-026-02964-x
