Is attention a redundant concept?

A major argument against attention as a theoretical construct is its purported redundancy. Critics claim that attentional effects can be explained through alternative mechanisms such as Bayesian decision processes (Anderson, Reference Anderson2011) or low-level constraints such as peripheral vision proposed by Rosenholtz. These perspectives suggest that empirical anomalies in attentional research result from misattributed causal mechanisms rather than failures in theoretical formulation.

However, this reasoning conflates explanatory adequacy with the overuse or misuse of the terminology. As Wu (Reference Wu2024) argues, attention should not be seen as an explanation in itself but rather as the effect that needs to be explained. Replacing attention with constructs such as “coarse-grained processing in peripheral vision” does not necessarily lead to greater precision in understanding cognition and behavior; rather, it shifts the label without resolving the underlying mechanisms. The assertion that attention is redundant assumes that alternative explanations are sufficiently comprehensive, yet they have not demonstrated the predictive power necessary to fully replace attentional theories. In science history, Kepler’s law generates good predictions that can be generalized from the solar system to other systems. But what about the “peripheral vision”? It is only a description of the phenomenon instead of any kind of cognitive theory that can generate testable and generalizable predictions.

The role of peripheral vision in attentional tasks: A confound, not a replacement

Rosenholtz proposes that many attentional effects can be better explained as limitations imposed by the structural and intrinsic features of peripheral vision (e.g., crowding and extraction of summary statistics) rather than an attentional bottleneck. While it is true that uncontrolled eye movements and peripheral processing influence task performance, these factors do not negate the existence of attentional mechanisms. The key question is not whether peripheral vision plays a role but rather how much variance in attentional tasks can be accounted for by peripheral factors alone.

Empirical evidence suggests that attentional processes cannot be fully reduced to sensory limitations. First, pupil size—a physiological marker of attentional effort—is associated with enhanced detection performance in peripheral vision and is modulated by cognitive demand (Eberhardt et al., Reference Eberhardt, Strauch, Hartmann and Huckauf2022; Vilotijević & Mathôt, Reference Vilotijević and Mathôt2024). Second, studies have shown that attention can enhance neural gain even when sensory input remains unchanged (Hillyard & Anllo-Vento, Reference Hillyard and Anllo-Vento1998), reinforcing its role beyond mere sensory encoding. Third, well-documented phenomena such as temporal attention (e.g., attentional blink, Luck, Vogel, & Shapiro, Reference Luck, Vogel and Shapiro1996; Raymond, Shapiro, & Arnell, Reference Raymond, Shapiro and Arnell1992) and feature-based attention (Zhang & Luck, Reference Zhang and Luck2009) suggest that attention functions extend beyond mere sensory constraints. Specifically, attentional blink reflects transient competition for attentional resources rather than simple perceptual masking (Chun & Potter, Reference Chun and Potter1995). Four, the contrast between resource and sensory limitations in distractor processing further supports the need for an “attentional theory” (Lavie & De Fockert, Reference Lavie and De Fockert2003). Lastly, attention as a theoretical construct can and should have strong utility, for instance as an account for cognitive process such as effort (Xie & Zhang, Reference Xie and Zhang2023).

The approach in the target article conceptually aligns with the notion of data-limited processing, which is distinct from attentional effects (i.e., resource-limited processing), as proposed by Norman & Bobrow (Reference Norman and Bobrow1975). Specifically, cognitive performance is constrained by the quality of available perceptual information (e.g., weak or ambiguous sensory input), such that additional cognitive processes, including attention, cannot enhance performance. Consistent with this perspective, the target article argues that many phenomena traditionally attributed to attention can instead be explained by the inherently low quality of peripheral visual information, rather than by top-down attentional modulation. In this sense, visual cognition operates under data-limited constraints, with peripheral vision representing one specific form of data limitation. Norman and Bobrow (Reference Norman and Bobrow1975) clearly demonstrated that even after accounting for data-limited constraints, cognitive resources such as attention are still necessary to modulate and optimize processing beyond these inherent constraints.

The pragmatic case for attention as a cognitive construct

Attention as a multifaceted construct (Luck & Vecera, Reference Luck and Vecera2002) involves multiple sub-processes and mechanisms, such as sensory enhancement versus noise reduction (Hillyard & Anllo-Vento, Reference Hillyard and Anllo-Vento1998). It is clearly inconsistent with the unified account as proposed in the target article. Even if attention lacks a single, universally agreed-upon definition, it remains a useful heuristic for organizing empirical findings and guiding theoretical development. The history of science demonstrates that concepts often persist despite definitional ambiguities—consider “representation” in cognitive neuroscience (Palmer, Reference Palmer1978) or “capacity” in working memory research (Kane et al., Reference Kane, Bleckley, Conway and Engle2001; Oberauer et al., Reference Oberauer, Farrell, Jarrold and Lewandowsky2016; Zhang & Luck, Reference Zhang and Luck2008). These terms continue to provide valuable insight even when their underlying mechanisms are debated.

Replacing attention with alternative paradigms does not necessarily yield theoretical elegance. Unlike Kepler’s laws that shifted the field in astronomy by replacing geocentric (Earth-centered) models with a heliocentric (Sun-centered) framework, the proposed alternatives to attention remain fragmented and context-dependent. Given that attentional research has yielded significant progress in fields ranging from neuroscience (Petersen & Posner, Reference Petersen and Posner2012; Posner & Petersen, Reference Posner and Petersen1990) to translation research (e.g., ADHD, Shalev, Tsal, & Mevorach, Reference Shalev, Tsal and Mevorach2007) and artificial intelligence (Soydaner, Reference Soydaner2022), rejecting attention prematurely would introduce conceptual voids without guaranteeing superior explanatory models.