Feedback on target rewards alters attentional bias during redundant-target detection

A flicker of feedback can change the way people search the visual world, even when the rewards and odds attached to different locations are carefully balanced. In a study published in Attention, Perception, & Psychophysics, researchers report that explicit performance feedback transformed participants’ attentional strategy during a rapid target-detection task. Without feedback, participants appeared to process information from two possible locations in parallel. Once feedback was added, however, their behavior indicated a shift toward sequential processing—effectively checking one location before the other. The result suggests that attention is not governed solely by the value of a target or the probability that it will appear in a particular place. Instead, the brain’s search strategy may be recalibrated by the broader context in which decisions are made, including whether people receive clear information about how well they are performing.

The finding addresses a long-standing question in cognitive science: when several factors compete to guide attention, does the mind combine them into a unified priority signal, or does it favor one source of information over another? In everyday vision, locations can differ in both their expected usefulness and their likelihood of containing something important. A driver may watch a frequently changing section of road, while a radiologist may concentrate on a region where abnormalities are more likely. Rewards add another layer: a correct response in one location may be worth more than a correct response elsewhere. Previous research has shown that learned value can capture attention and that spatial probability can bias visual search. But it has remained uncertain how these influences interact when they point in different directions. The new study isolates that interaction by assigning target locations equal expected value while separating reward magnitude from target probability.

The researchers used a redundant-target detection task, a design intended to reveal how multiple sources of information are processed. Participants viewed displays in which targets could appear at the top of the screen, at the bottom, or at both locations simultaneously. They reported whether a target was present. The two locations were matched in expected value, meaning that the combination of reward size and likelihood produced the same overall payoff across locations, even though the ingredients of that payoff differed. One location could be associated with a larger reward but a lower probability of containing a target, while the other carried a smaller reward but appeared more often. This arrangement prevented simple differences in total expected benefit from explaining any attentional preference. The critical comparison was between trials conducted with performance feedback and trials conducted without it.

The study’s central technical tool was Systems Factorial Technology, or SFT, a mathematical framework for distinguishing processing architectures that can look similar in ordinary reaction-time data. Conventional analyses often compare average response times or error rates across conditions. Those measures can indicate that two signals help detection, but they do not necessarily reveal whether the signals are processed simultaneously, one after another, or through some form of interaction. SFT instead examines how experimental factors combine across the full distribution of response times. Its logic makes it possible to test predictions from competing models of information processing at the level of an individual participant. In this study, the method was used to determine whether the top and bottom target channels operated in parallel or whether attention effectively prioritized one channel before turning to the other.

In a parallel architecture, information from both possible target locations can accumulate at the same time. Detection is then determined by whichever channel reaches a decision threshold first, a pattern sometimes described as a “race” between processing channels. In a serial architecture, by contrast, the observer inspects one location and then, if necessary, shifts to the other. These strategies can produce overlapping average performance, particularly when the task is simple, which is why the researchers relied on the more detailed SFT analysis. The framework evaluates relationships among response-time distributions and uses factorial interactions to test whether the evidence is consistent with simultaneous or sequential processing. The approach is especially useful when the underlying cognitive architecture is uncertain, because it does not require researchers to assume in advance that attention must operate in one particular way.

When no feedback was provided, the results indicated parallel allocation across the two target locations. Neither the difference in reward magnitude nor the difference in target-location probability produced a detectable attentional bias under these conditions. This does not mean that reward and probability are irrelevant to vision in general. Rather, in the carefully balanced setting of the experiment, neither factor was sufficient to push participants toward one location or to reorganize the way the two locations were processed. The result is important because it challenges a simple version of the idea that attention automatically follows the most valuable or most probable target. When the expected payoff was equated and participants were not shown explicit information about their performance, their visual system behaved as though it could monitor both locations concurrently.

The pattern changed when feedback was introduced. Under feedback conditions, the SFT results supported a sequential allocation strategy. Participants no longer appeared to process the two possible target locations in parallel; instead, their behavior was consistent with checking locations in an ordered manner. The abstract does not identify a single fixed order that all participants followed, nor does it claim that one reward dimension universally dominated the other. The key result is the change in processing architecture itself. Feedback appears to have encouraged participants to adopt a more selective strategy, potentially because it made the consequences of their decisions more explicit or increased the perceived importance of managing uncertainty. Rather than simply amplifying an existing bias toward high reward or high probability, feedback altered how attentional resources were deployed across the display.

That distinction matters for theories of attention and perceptual decision-making. Many models treat attention as a weighting system: information judged more relevant receives greater processing priority, while less useful information receives less. The new findings suggest that context can influence not only the weights assigned to locations but also the sequence in which those locations are evaluated. Feedback may prompt observers to exploit regularities, reduce uncertainty, or trade the speed of parallel monitoring for the selectivity of serial inspection. Such a trade-off could be adaptive. Parallel processing allows rapid detection when either location may contain a target, but it may be less efficient when the observer can learn a reliable inspection order. Sequential processing can concentrate resources, yet it risks missing information at a location that is checked later. The participants’ shift therefore points to strategic flexibility rather than a permanent attentional preference.

The study also reinforces the idea that perceptual decisions are shaped by information arriving after, as well as before, a response. Feedback is often assumed to improve accuracy by teaching people which choices are correct, but its effects can extend to the organization of the decision process itself. Here, explicit feedback was associated with a move from simultaneous to sequential processing, even though the reward and probability structure of the task remained the same. The researchers link this result to the role of uncertainty: when people receive clearer evidence about performance, they may revise how they sample competing sources of information. That conclusion remains bounded by the experiment’s design. The studies were not preregistered, and the article presents the result as evidence for contextual dependence, not as proof that feedback always produces serial attention. De-identified trial-level data and analysis code are publicly available through the Open Science Framework, allowing the finding and its analytical approach to be examined further.

The implications reach from basic vision science to situations in which people must search efficiently under pressure. Training systems, user interfaces, medical-image interpretation and monitoring tasks often provide feedback while also manipulating the probability and value of events. If feedback changes the architecture of attention, then performance gains or losses may reflect altered search strategy rather than changes in sensory sensitivity alone. The work may also help explain why the same cue can guide attention differently depending on whether observers are receiving reliable information about outcomes. For now, the most striking message is that attention is not merely a spotlight whose intensity follows reward. It is also a strategy for organizing evidence over time. A simple signal that says “correct” or “incorrect” may be enough to make the mind stop watching two places at once and begin inspecting them in sequence.