The knowledge-centric approach advocated by Phillips et al. represents a welcome advancement in theory of mind research, and I am in complete agreement with this proposed shift in focus. My concern, however, is that Phillips et al. have overlooked an important source of evidence available to this emerging project – the neuroscience of knowledge attribution. Capable of providing insights even when undetectable in behavioral measures, as well as independent lines of converging evidence, hemodynamic (e.g., functional magnetic resonance imaging [fMRI] and functional near-infrared spectroscopy [fNIRS]) and neurophysiological techniques (e.g., electroencephalography [EEG] and magnetoencephalography [MEG]) serve as powerful tools in theory of mind research. Crucially, neuroimaging has already begun to provide direct support for Phillips et al.'s central claim that knowledge attribution is more basic than belief attribution – belief attribution seems to demand neural resources that knowledge attribution does not (Bricker, Reference Bricker2020). All this give us compelling reason to think that the neuroscience of knowledge attribution has a vital role to play in the nascent knowledge-centric theory of mind research program.

It is not without good reason that neuroimaging techniques have been widely employed in the effort to understand our theory of mind systems (for overviews, see Carrington & Bailey, Reference Carrington and Bailey2009; Heleven & Van Overwalle, Reference Heleven and Van Overwalle2018; Mahy, Moses, & Pfeifer, Reference Mahy, Moses and Pfeifer2014; Schurz, Radua, Aichhorn, Richlan, & Perner, Reference Schurz, Radua, Aichhorn, Richlan and Perner2014). A considerable amount of evidence indicates that the cognitive processes supporting human theory of mind capacities are both associated with identifiable neural correlates (see Heleven & Van Overwalle, Reference Heleven and Van Overwalle2018) and distinct from more generalized executive function in the brain (see e.g., Bradford, Brunsdon, & Ferguson, Reference Bradford, Brunsdon and Ferguson2020; Hartwright, Apperly, & Hansen, Reference Hartwright, Apperly and Hansen2015; Pacella et al., Reference Pacella, Scandola, Beccherle, Bulgarell, Avesani and Moro2020; Samson, Houthuys, & Humphreys, Reference Samson, Houthuys and Humphreys2015). If the knowledge-centric theory of mind program is to achieve success comparable to that of its belief-centric counterpart, this observation is key. Mental-state attributions are best understood not as cognitive, but rather as neurocognitive processes.

The identifiable neural correlates of theory of mind processing enable neuroimaging techniques to provide additional lines of evidence that can converge with the findings of other methods. For example, fMRI studies indicating that the perspective taking and self-perspective inhibition components of theory of mind are largely supported by distinct regions in the brain (e.g., Hartwright et al., Reference Hartwright, Apperly and Hansen2015; Özdem, Brass, Schippers, Van Der Cruyssen, & Van Overwalle, Reference Özdem, Brass, Schippers, Van Der Cruyssen and Van Overwalle2019; Schuwerk et al., Reference Schuwerk, Döhnel, Sodian, Keck, Rupprecht and Sommer2014; van Der Meer, Groenewold, Nolen, Pijnenborg, & Aleman, Reference van Der Meer, Groenewold, Nolen, Pijnenborg and Aleman2011) have offered considerable support to the claim that these are indeed separate neurocognitive processes, which was initially suggested by Samson et al. primarily on the basis of lesion studies (Reference Samson, Apperly, Kathirgamanathan and Humphreys2005).

Moreover, neuroimaging methods are especially valuable in their capacity to provide insights into theory of mind processing even when those insights aren't salient on behavioral measures. To take an example from fMRI, Hartwright et al. found differences in hemodynamic activity indicating that self-perspective inhibition during mental-state attribution is distinct from inhibition during non-mental tasks, a finding that was not detectable in their behavioral data (Reference Hartwright, Apperly and Hansen2015). Taking a similar example from EEG, an N400 paradigm employed by Bradford et al. revealed initial egocentric processing during the attribution of false beliefs – even when those attributions were ultimately computed successfully (i.e., altercentrically) – providing key evidence that “egocentric processing is the default perspective for information integration” in such cases (Reference Bradford, Brunsdon and Ferguson2020, p. 276). Again, this evidence was not salient in their behavioral data.

All this provides a general sense of the value of neuroimaging in theory of mind research. However, the neurocognitive findings most directly pertinent to the research program imagined by Phillips et al. come from my own EEG study (Bricker, Reference Bricker2020). The first step in a broader research project dedicated to understanding the neurocognitive mechanisms of knowledge attribution, the design of this study was simple, with participants varyingly judging whether a cartoon character sitting at a table knew/believed that there were two cylinders on the table. This study provided two key results: (1) There were no significant difference in response time between belief attribution and knowledge attribution. (2) Differences in P3b amplitude indicated that the belief attribution tasks demanded a level of neural resources significantly greater than that of the knowledge attribution tasks, which is most likely explained by a greater demand for self-perspective inhibition during belief versus knowledge attribution.

These findings are relevant to the account presented in the target article for at least two distinct reasons. First, these results provide additional evidence for the target article's central claim that knowledge attribution is at least as basic as belief attribution. As with the response time evidence discussed by Phillips et al. (sect. 5.1), the idea that knowledge attribution relies on something like a belief attribution stage is inconsistent with the observation of comparable response times for belief and knowledge attribution tasks. We see something similar to the neurophysiological results, which indicate that belief attribution can entail processing demands that exceed those of knowledge attribution. This again suggests that knowledge attribution is the more basic of the two processes.

However, beyond simply providing further evidence that belief attribution does not come before knowledge attribution, the results of this study also illustrate why knowledge-centric theory of mind research works best when understood as a neurocognitive endeavor, highlighting both the advantages of neurocognitive techniques outlined above. Not only did the neurophysiological results of the study provide an additional line of evidence for the conclusion suggested by behavioral measures, but these findings also offered a further insight not salient on behavioral measures – Belief attribution appears to be a more resource-intensive process than knowledge attribution, likely because of differential demands for self-perspective inhibition.

Although it is too early to speculate whether this knowledge-focused theory of mind research will ultimately attract the same attention as its belief-centric counterpart, it is clear that neurocognitive techniques have a good deal to offer this emerging project. Through the integration of behavioral and neuroimaging methods with the characterization of knowledge states offered by epistemologists (see especially sect. 2 of the target article; Bricker Reference Bricker2020, sect. 1.1), we stand to make significant strides toward understanding the mechanisms underlying our judgments about knowledge, which are at present still largely unknown.