Rosenholtz’s reframing of visual attention as a process in which an observer chooses and performs a task raises important questions about the goals of the visual system – of the observer – and reveals how a more complete understanding of these goals can help us develop more coherent and ecologically valid theories of attention. Considering possible goals and the choices that might accompany them, a significant oversight of broader attention theory is the assumption that the perceptual tasks we choose are in service of perceptual goals. Instead, for embodied human observers, we argue that the primary role of perceptual selection is to support action. Beyond the laboratory, visual perception is often only useful to the extent that we can act on what we perceive: It is not enough to see the berries and notice the approaching tiger – we want to pick the berries and avoid the tiger. Thus, rather than directly adopting Rosenholtz’s approach, we propose an orthogonal paradigm shift in how we conceptualize visual attention. We suggest framing it as part of an integrated sensorimotor system that links vision and action. This challenges the traditional “see → decide → act” framework and advocates for a broader view in which attention extends beyond visual perception alone. In real-world scenarios, successful interactions require not just locating target objects but also reaching for and manipulating them to achieve goals. This complex behavior depends on the seamless coordination of multiple brain systems responsible for processing sensory signals, selecting tasks consistent with our goals (i.e., allocating attention), and controlling visually guided actions. Therefore, we argue that visual attention is not passive but actively shaped by the demands of action (Nakayama, Moher, & Song, Reference Nakayama, Moher and Song2023; Song, Reference Song2019).

Our recent work (Moher, Anderson, & Song, Reference Moher, Anderson and Song2015) highlights the complex interactions between these systems. Traditional models suggest that physically salient objects, such as flashing hazard signs, capture attention regardless of the observer’s goals (Anderson, Laurent, & Yantis, Reference Anderson, Laurent and Yantis2011; Theeuwes, Reference Theeuwes1994; Theeuwes & Godljn, Reference Theeuwes and Godljn2002). However, we demonstrated that these same cues can trigger a suppression mechanism during action selection, but not during perceptual tasks. Specifically, high-contrast stimuli were more distracting in perceptual tasks, in line with traditional models, but less disruptive when rapid motor actions were required. Notably, these findings also generalized to stimuli that were salient due to their association with reward. This suggests a salience-triggered suppression mechanism during goal-directed action, where highly salient distractors are quickly suppressed to minimize interference. Conversely, in perception-based visual search tasks, strong distractors caused more interference. This might seem counterintuitive as one would expect stimuli affecting conscious vision to similarly influence visually guided actions. We propose that the discrepancy arises because the cost of distraction is greater when maintaining an action course under time pressure. This leads the brain to adapt by suppressing stronger distractors more effectively during goal-directed actions. These findings underscore the need to integrate visually guided action paradigms with traditional psychophysical approaches to fully understand the coordination of perception and action. This integration would offer new insights into visual attention’s role in achieving behavioral goals and resolving competing processes in complex visual environments (Song, Reference Song2017).

Research on perception and action also reveals a strong link between visual attention, hand movements, and tool use. Neurons in the monkey parietal cortex, particularly in the intraparietal sulcus, show dual receptive fields – somatic and visual. This provides a neural basis for enhanced visual processing near the active hand (Iriki, Tanaka, & Iwamura, Reference Iriki, Tanaka and Iwamura1996). This extends to humans, where reaction times are faster for stimuli near the hands, supporting the notion of heightened visual attention in peripersonal space (Reed, Grubb, & Steele, Reference Reed, Grubb and Steele2006). Crucially, this effect depends on action: visual advantages arise only when tools are actively used, not merely held (Reed et al., Reference Reed, Betz, Garza and Roberts2010). Further research by Thomas (Thomas, Reference Thomas2015, Reference Thomas2017) showed that different hand actions modulate visual attention through distinct pathways. Precision grips, requiring fine spatial resolution, enhanced parvocellular processing, while power grips, involving dynamic processing, engaged magnocellular pathways. Our own work (Guo & Song, Reference Guo and Song2019) demonstrated that a concurrent grasping task enhanced orientation discrimination. Furthermore, prior precision-grasping training also improved sensitivity to orientation discrimination. These findings suggest that action preparation or enhancement boosts perceptual discrimination, emphasizing the bidirectional relationship between motor and visual performance. Rather than merely relying on perception, action actively shapes it.

These findings, among many others, underscore the need to move beyond constrained models of “attention for perception” and to explicitly consider action’s role in shaping the goals and selection mechanisms of the visual system. Critically, this approach does not necessitate more complex models of attention. Instead, acknowledging the pervasive interaction between perception and action could yield more parsimonious models by more accurately describing what selection is for, by capturing unanticipated constraints on task complexity, and by grounding the study of visual attention in more ecologically valid contexts.

This integration would also facilitate the generation of more comprehensive psychophysical datasets. Rather than reducing complex perceptual experiences to discrete keypresses, incorporating continuous action-based tasks in vision research could yield deeper insights into distractor interference, response competition, and decision confidence, revealing dynamics that traditional methods may overlook (Hayhoe, Reference Hayhoe2017; Song & Nakayama, Reference Song and Nakayama2009; Wöstmann et al., Reference Wöstmann, Störmer, Obleser, Addleman, Andersen, Gaspelin, Geng, Luck, Noonan and Slagter2022). These insights could help reconcile previously conflicting findings and provide a clearer understanding of anomalies by framing them within a perspective that more accurately reflects the (embodied) goals of both the visual system and the observer.

In conclusion, the evolving understanding of visual attention underscores its dynamic relationship with sensorimotor integration. While Rosenholtz’s critique of current attention models prompts valuable reconsideration, a broader framework that incorporates action processes provides a more comprehensive perspective. Rather than functioning solely as a perceptual filter, visual attention is dynamically shaped by the demands of action. We hope this perspective inspires refinements to attention theory and encourages a reassessment of prior findings in visual attention research.