Rosenholtz argues that visual attention is in crisis because phenomena typically described as attentional can be explained without appealing to attention-like mechanisms. Because of this, the field of visual cognition might be better served by abandoning, or at least significantly revising, the concept of attention. While we agree that inaccurate, incoherent, or merely inconsistent definitions of attention hold back theories of perception and cognition, as has been argued many times before (Anderson, Reference Anderson2011; Hommel et al., Reference Hommel, Chapman, Cisek, Neyedli, Song and Welsh2019; Zivony & Eimer, 2021), we believe that the concept of attention provides critical explanatory power and should be retained. Rosenholtz forwards two alternative sources of the capacity limits usually attributed to attention: peripheral vision and the drawing of decision boundaries for different tasks. We specifically challenge the effectiveness of replacing attention theory with a framework of limits on decision boundaries and instead defend at least one form of selection as “attention”, in which perceptual experiences, and in some cases even the representational spaces of perception themselves, are reshaped by attentional selection.

Even if a framework focused on tasks and decision boundaries can explain perceptual effects of various capacity limits, it is unclear how it would do so in a way that avoids the very crises that Rosenholtz attributes to attention theory. The decision framework argues that tasks are solved by implementing decision boundaries that best discriminate between the features or objects that are relevant and irrelevant for a given task. Under this framework, decisions are constrained by the complexity of the decision boundaries and how well these boundaries divide the space between different items. While this appears like a promising idea that could highlight how decision-making is limited in its ability to draw complex boundaries given the high dimensionality of visual representations, we do not believe this theory can fully replace attention.

For one, focusing on task-related capacity limits alone will actually require more complex mechanisms that would fail to coherently predict results across paradigms compared to clearly defined manipulations of attention. While attention includes a multitude of principles, they all share a common, well-specified function: the selection of information that is relevant for action (Wu, 2023). Replacing this clear function of attention with a set of mechanisms that draw decision boundaries through an abstract task space, or of changing tasks, could result in an unconstrained theory capable of arbitrarily complicated predictions (but little ability to make falsifiable predictions). It also remains unclear what decisions in the high-dimensional task space achieve if not the selection of the most task-relevant information – which Rosenholtz questions the usefulness in arguing against attention as a simple gating mechanism.

What, then, is the rightful domain of “attention?” Rosenholtz provides a strong argument against the idea that attention resolves a bottleneck in visual processing and related views like the feature integration theory of attention. However, more recent theories conceptualize attention in other ways that are not contested by Rosenholtz and other critics. For example, many of the known capacity limits in visual processing are clearly predicted by the organization of the visual representations themselves, thus arising from perceptual stages, not decision-making stages (Alvarez & Cavanagh, Reference Alvarez and Cavanagh2004; Störmer, Alvarez, & Cavanagh, Reference Störmer, Alvarez and Cavanagh2014; Störmer & Alvarez, Reference Störmer and Alvarez2014). Other patterns of behavior are similarly well predicted by attention but not decision boundaries or limits on peripheral vision. For example, auditory cues have cross-modal effects on visual phenomenology (Störmer, McDonald, & Hillyard, Reference Störmer, McDonald and Hillyard2009; Störmer, Reference Störmer2019), and attentional manipulations affect neural signatures of and confidence about perceptual experiences, and not merely accuracy or response time (Störmer & Alvarez, Reference Störmer and Alvarez2016; Störmer et al., Reference Störmer, McDonald and Hillyard2009). Furthermore, attentional manipulations can reshape entire representational spaces (Chapman, Chunharas, & Störmer, Reference Chapman, Chunharas and Störmer2023), demonstrating that it is not just decision boundaries discriminating between static visual representations. Finally, selection effects also occur in the absence of any explicit task instructions: incidental learning about frequently task-relevant information influences visual search efficiency (Addleman & Lee, 2021; Golan & Lamy, Reference Golan and Lamy2024), and neural markers of early stages of selection like the N2pc (Duncan, Van Moorselaar, & Theeuwes, Reference Duncan, Theeuwes and Van Moorselaar2023; Ortego, Addleman, & Störmer, 2024). These effects can occur implicitly (Jiang, Reference Jiang2018), in that participants selectively process frequently task-relevant information even when they can’t identify what that task-relevant information is; it is ambiguous how such implicit and incidental selection biases would be reflected in a theory of decision boundaries.

To explain these and other effects, we recently proposed a theoretical framework in which attention is not wholly about the selection of discrete targets but reflects a shift in the way visual information is represented in the brain to support behavior (Chapman & Störmer, Reference Chapman and Störmer2024). For example, one set of findings regards the effects of attention on the appearance of visual stimuli. Several studies have shown that attended items are perceived as higher in contrast than unattended items (Carrasco, Ling & Read, Reference Carrasco, Ling and Read2004; Gobell & Carrasco, Reference Gobell and Carrasco2005; Störmer et al., Reference Störmer, McDonald and Hillyard2009), which results in attended items standing out more among competing items. Likewise, searching for a singleton color target can alter its perceived color, such that targets appear more distinct from the distractors (Chapman & Störmer, Reference Chapman, Chunharas and Störmer2023; see also, Navalpakkam & Itti, Reference Navalpakkam and Itti2007). This suggests that attention has effects on our visual experience that cannot be easily attributed to changes in decision boundaries. This also maps onto a large literature showing effects of attention at early cortical processing stages, where attention increases the strength of sensory signals (e.g., Hillyard & Anllo-Vento, Reference Hillyard and Anllo-Vento1998; Martinez-Trujillo & Treue, Reference Martinez-Trujillo and Treue2002; Reynolds & Chellazi, Reference Reynolds and Chelazzi2004).

In conclusion, as vision scientists who study attention, we advocate that attention provides significant explanatory power in understanding perceptual processing limits and phenomenology. The “attention” label may frequently be applied inappropriately to other selective processes, but any alternative theories do a poor job of explaining many phenomena in which perceptual processing is reshaped or otherwise altered by a person’s “attentional” state. Instead of trying to determine which tasks or behaviors require attention, we think a more fruitful approach is to explore the role of attention in how the brain orchestrates the flow of information in support of current behavioral goals.