Rosenholtz claims attention research was in a crisis, as defined in Kuhn’s (Reference Kuhn1962) paradigm theory. To solve the crisis, Rosenholtz calls for a paradigm shift, replacing the concept of attention with other mechanisms for stimulus-driven and task-driven vision. However, from a philosophical and psychological viewpoint, attention research fails to meet the criteria of a scientific crisis. Calling for a paradigm shift seems unwarranted and would impede progress in attention science.

According to Kuhn (Reference Kuhn1962, Ch. 7), a scientific crisis occurs if the scientific community loses faith in the ability of the leading paradigm (the field’s fundamental theories and methodologies) to resolve persistent conflicts (“anomalies”) between theory and empirical evidence. Shifting to a new paradigm resolves the conflicts. The paradigm shift is revolutionary: Old and new paradigms are incommensurable, and there is no conceptual continuity between them. The new paradigm cannot be derived by introducing new assumptions in the old paradigm. Thus, progress becomes an illusion because with each paradigm shift, new theoretical foundations and methods are developed from scratch, cutting links to knowledge from the past.

Contrary to Rosenholtz’ claims, attention research violates both the criteria of scientific crisis and those of revolution that would rule out progress in the field. Attention refers to a broad class of mechanisms accommodating the brain’s limited processing capacity for perception by setting processing priorities (e.g., Bundesen, Reference Bundesen1990; Desimone & Duncan, Reference Desimone and Duncan1995). Rosenholtz claims to have found “anomalies” revealing the failure of the “attention paradigm” to provide useful theories. Uselessness is inferred from an increasing complexity of the theories to accommodate new findings (pp. 3–4). For instance, that the gist of a visual scene is processed faster than objects within the scene has been explained by assuming additional mechanisms for scene processing that complement attention (Wolfe et al., Reference Wolfe, Võ, Evans and Greene2011). Critically, an increase in complexity does not indicate a crisis (cf. Kuhn, Reference Kuhn1962). To the contrary, the possibility to accommodate new findings by introducing new assumptions demonstrates a continuous progression in theory development without abandoning core posits entirely (cf. Lakatos, Reference Lakatos, Worrall and Currie1978). In scene perception, for example, newly assumed mechanisms (Wolfe et al., Reference Wolfe, Võ, Evans and Greene2011) can be perfectly reconciled by amending previous attention theories (Wolfe, Reference Wolfe1994). This goes against a crisis of attention research.

For a crisis, there need to be findings challenging the fundamental assumptions of a research field. For attention, as a class of priority-setting mechanisms (see above), this would be the case if findings challenged the assumption that there was any prioritization in perception and cognition. Critically, this challenge cannot be found by focusing on specific subareas of attention research, such as peripheral vision (p. 5–8), gist/scene perception (p. 3–4), or dual-tasking (p. 5–8), even if there were findings conflicting with prioritization in processing (e.g., losing information within the computation of a summary statistic as opposed to losing it because it had been filtered-out in a selection process, p. 8). This is because such a focus would necessarily neglect other areas that clearly demonstrate prioritization. For instance, that prioritization must exist in perception and cognition is evident from the fact that humans can bring their limited effectors (say, two hands) to operate on one out of several (say more than three) objects at a time and can do so, flexibly, to perform a specific (instruction-based) task (cf. Neumann, Reference Neumann, Heuer and Sanders1987). Adopting the view of Lakatos (Reference Lakatos, Worrall and Currie1978) rather than Kuhn’s (Reference Kuhn1962) paradigm theory, one might say that the assumption of priority was the “hard core” of attention research. Attentional mechanisms (e.g., those that select stimuli for perception [covert attention] and/or as targets for the next saccadic eye movement, bringing them onto the high-resolution fovea [overt attention], cf. Deubel & Schneider, Reference Deubel and Schneider1996, or those that dynamically allocate neurons’ receptive fields to processing a specific stimulus, Desimone & Duncan, Reference Desimone and Duncan1995) would constitute hypotheses auxiliary to the hard core. Auxiliary hypotheses can be tested, modified, and potentially refuted or replaced without letting the core theory be challenged directly, forming a “protective belt” around the hard core (Lakatos, Reference Lakatos, Worrall and Currie1978). This seems in line with attention research, where auxiliary hypotheses of specific mechanisms performing prioritization (or selection) are tested, amended, or refuted, without questioning the principle of accommodating processing capacity limitations by prioritizing specific processes.

Now, what happens if we call for a paradigm shift without crisis? Paradigm shifts cut access to previous theories, methodologies, and findings. So, banning the concept of attention, as Rosenholtz suggests (p. 10), would disconnect new research from a plethora of theories and findings connected by the assumption of processing priorities (i.e., the broad concept of attention). This strategy risks a new replication crisis (Open Science Collaboration, 2015) because the (conceptual) replication of findings, which is the safeguard against erroneous findings (due to statistical variation), rests on these connections that allow to find, synthesize, and meta-analyze previous findings in concert with current ones. Rosenholtz suggested that “attention” was too broad and unspecific a concept to be useful for scientific progress (p. 4). However, it could be precisely the breadth of the concept that ensured stable progress. Rather than calling for a paradigm shift, the breadth of the concept could serve the important function of organizing the variety of prioritization mechanisms (auxiliary hypotheses) that compete as explanations of visual phenomena under a common superordinate principle (prioritization as the “hard core”). Adopting such an organization allows for a sustainable science of attention: Under a superordinate principle, previous findings, methodologies, and theories can be found, accessed, and understood. This ensures that they can be integrated into current revisions of attention theory, which would place the theory on a richer evidence base as compared with the situation of focusing on a specific sub-area. Thus, surprisingly, we need the broad and unspecific concept of attention (cf. Rosenholtz, p. 4) to achieve an incremental and sustainable science of attention (cf. Lakatos, Reference Lakatos, Worrall and Currie1978; Platt, Reference Platt1964) rather than ruling out progress by waxing and waning Kuhn (Reference Kuhn1962) paradigms.