In the target article, Rosenholtz argues that “attention theory” is a scientific paradigm in crisis (Rosenholtz, Reference Rosenholtz2025). This formulation comes from the philosopher of science Thomas Kuhn’s influential idea that a dominant scientific theory can encounter unexplained anomalies in data, which first lead to overcomplications in the theory and eventually give way to its abandonment (Kuhn, Reference Kuhn1962). Rosenholtz reports that she banned the use of the word “attention” in her laboratory for a year as an in-house experiment in theory shedding. But this Kuhnian move raises a question: is attention a theory?

The term “attention” is used widely and variously in psychology and neuroscience, but attention is rarely, if ever, conceptualized as a scientific theory. Rather, like memory and emotion, attention is a psychological concept that captures a broad class of phenomena. The key feature of attention is the prioritization of some items at the expense of others (Carrasco, Reference Carrasco2011; Denison, Reference Denison2024; James, Reference James1890; Reeves, Reference Reeves2020). Decades of careful psychophysical work have shown that prioritization of a spatial location (Carrasco, Reference Carrasco2011; Montagna, Pestilli, & Carrasco, Reference Montagna, Pestilli and Carrasco2009; Pestilli & Carrasco, Reference Pestilli and Carrasco2005), point in time (Denison, Carrasco, & Heeger, Reference Denison, Carrasco and Heeger2021; Denison, Heeger, & Carrasco, Reference Denison, Heeger and Carrasco2017; Jing et al., Reference Jing, Jin, Li, Wu, Chen and Huang2022), or feature such as the color red (Liu, Reference Liu2019; White, Rolfs, & Carrasco, Reference White, Rolfs and Carrasco2015) improves perception of the attended item while impairing perception of other items, relative to a neutral baseline. These experiments have controlled for potential confounds such as eye movements and eccentricity, such that the stimulus remained constant while only instructions about what to prioritize changed from trial to trial. The findings of perceptual trade-offs—benefits for prioritized items accompanied by costs for nonprioritized ones—indicate some form of selectivity, the hallmark of attention. So, although attentional phenomena need theoretical explanations, it is simply a fact that we regularly use our minds to focus on some things and ignore others. Attention is not a theory.

What then, does Rosenholtz mean by “attention theory”? She seems to be referring to a particular class of theoretical accounts of attention based on feature integration theory (Treisman & Gelade, Reference Treisman and Gelade1980). “Attention theory,” she says, proposes a selective mechanism that gates access of visual sensory information to higher-level processing. Her Figure 1 shows a “preattentive processing” stage followed by a “selective attention” stage with a parallel “distributed attention” pathway. Rosenholtz describes the selective attention gate (“in its simplest form”) as all or none, so that only the attended item would pass through to higher, cognitive stages. The distributed attention pathway, not present in the original feature integration theory, accommodates findings such as the ease with which people can report the gist of a scene (Ariely, Reference Ariely2001; Treisman, Reference Treisman2006).

This “attention theory” is a simple box-and-arrow model that makes a strong theoretical commitment to discrete stages of processing in the visual system. Too strong, it turned out. Feature integration theory was developed over decades (Treisman, Reference Treisman2006; Treisman & Gelade, Reference Treisman and Gelade1980), but its notion of discrete preattentive and attentive stages is no longer prominent for three main reasons that emerged from behavioral and neuroscience research. First, attentional selection need not be all or none. Inattention can reduce the precision of visual representations without eliminating them (Anton-Erxleben & Carrasco, Reference Anton-Erxleben and Carrasco2013; Kay, Weiner, & Grill-Spector, Reference Kay, Weiner and Grill-Spector2015). Second, attentional modulation happens concurrently at multiple levels in the visual system, starting with the earliest stages of cortical (Dugué et al., Reference Dugué, Merriam, Heeger and Carrasco2020; Motter, Reference Motter1993) and even subcortical (Briggs, Reference Briggs2020; McAlonan, Cavanaugh, & Wurtz, Reference McAlonan, Cavanaugh and Wurtz2008) visual processing. Indeed, attention affects basic dimensions of vision that had been considered preattentive (Carrasco, Reference Carrasco2011). Third, research has revealed multiple mechanisms for prioritizing visual information. This research has lead the concept of attention to be subdivided based on the object of prioritization (space, time, features, etc.) and the mode of prioritization, including voluntary or involuntary (Cheal & Lyon, Reference Cheal and Lyon1991; Müller & Rabbitt, Reference Müller and Rabbitt1989), presaccadic (Li, Hanning, & Carrasco, Reference Li, Hanning and Carrasco2021), and experience-driven (Kim & Anderson, Reference Kim and Anderson2019) prioritization. So, while Rosenholtz rightly points out problems with two-stage theories of attention, most of the field has long moved beyond these early conceptual models, productive as they once were for driving attention research.

Rosenholtz worries, in the Kuhnian framework, that subdividing attention into different types is a slide into epicycles—those extra orbits tacked on to planetary motion to preserve the theory that the earth was the center of the solar system in the face of anomalous evidence. She argues that the complexity of the current attentional landscape signals a scientific paradigm in crisis. But the field of visual attention research has long embraced pluralism about attentional phenomena and recognized that different mechanisms may explain different kinds (Maunsell, Reference Maunsell2015; Nobre & van Ede, Reference Nobre and van Ede2023)—in the same way that memory research has identified different kinds of memory and associated mechanisms. This pluralism, in which researchers delineate multiple mechanisms of attention with increasing precision, is a sign of the healthy progress of a productive field, not epicycles of crisis. Complexity does not signal a crisis for attention, because attention is not a theory.

Although seeking a single, unified theory of diverse attentional phenomena may be counterproductive, the field has made progress in developing formal (mathematical and computational) theories that explain certain attentional phenomena. For example, the normalization model of attention (Reynolds & Heeger, Reference Reynolds and Heeger2009) predicts psychophysical and neural data from spatial and feature-based visual attention tasks (Herrmann, Heeger, & Carrasco, Reference Herrmann, Heeger and Carrasco2012; Herrmann et al., Reference Herrmann, Montaser-Kouhsari, Carrasco and Heeger2010; Ni & Maunsell, Reference Ni and Maunsell2019) and has been extended to temporal attention (Denison et al., Reference Denison, Carrasco and Heeger2021). By combining simple computations such as gain control and divisive normalization, these models help us understand how attention modulates visual cortical activity. But Rosenholtz is right that existing theories do not capture all the attentional phenomena we would like to understand, and her catalog of phenomena to be explained is a useful guide to theory development. She also helpfully centers the question of what causes perceptual and cognitive capacity limits—which make attention useful in the first place. Hence, although attention is not a theory, the science of attention must continue to develop theoretical accounts of the variety of ways in which humans prioritize and select information.