Haueis and Colaço’s target article shows that brain metabolism is a critical and often overlooked factor in psychology, neuroscience, philosophy, and artificial intelligence. Among its many strengths, their perspective encourages researchers to get specific about how to integrate brain metabolism – as a source of empirical evidence – into traditional cognitive accounts, either evaluatively constraining them or generatively expanding them. Obviously, this is a long-term project, but even in the short term, it has an important effect: It encourages researchers to interrogate placeholder terms, like “neural activity,” which embed assumptions into research paradigms. As a term, “neural activity” is so ubiquitous as to seem innocuous; but it both lumps neural, hemodynamic, and metabolic events into a single construct (Singh, Reference Singh2012), and embeds a traditional stimulus–response logic at the foundation of neuroscience (Hutchinson & Barrett, Reference Hutchinson and Barrett2019; Keller & Mrsic-Flogel, Reference Keller and Mrsic-Flogel2018). The proposed metabolic perspective calls this term into question and points to an alternative.

Haueis and Colaço are likely aware that “neural activity” has these problems, as their paper deftly avoids it. What they say instead is worth highlighting: They generally refer to “neural activities.” The difference is subtle but important. “Neural activity” implies a distinction between active and inactive brain states (Raichle, Reference Raichle2010, Reference Raichle2015). By contrast, “neural activities” imply that neural populations occupy many states, and that empirical measures – e.g., single unit recordings, local field potentials (LFPs), cerebral blood flow, and oxygen/glucose metabolism – are independent of theoretical interest, as opposed to all being better or worse approximations of an underlying, latent, and singular “neural activity.”

One might object, saying that no one actually thinks these measures are interchangeable indices of latent “neural activity”, but, in practice, they are often treated this way. This is most apparent in the blood-oxygen level-dependent (BOLD) response, the mainstay measure in fMRI, elicited by events like sensory stimulation or cognitive work. The event-elicited BOLD response is influenced by the ratio of oxygenated to deoxygenated hemoglobin, and BOLD increases are caused by localized increases in blood flow, drawing more oxygenated hemoglobin to the “active” region (Kwong et al., Reference Kwong, Belliveau, Chesler, Goldberg, Weisskoff, Poncelet, Kennedy, Hoppel, Cohen and Turner1992; Ogawa et al., Reference Ogawa, Lee, Kay and Tank1990, Reference Ogawa, Tank, Menon, Ellermann, Kim, Merkle and Ugurbil1992). In 2001, Logothetis reported a key correlation between event-elicited BOLD increases and LFPs (Logothetis et al., Reference Logothetis, Pauls, Augath, Trinath and Oeltermann2001). This work fostered an interpretation of BOLD as an approximate measure of local synaptic transmission, licensing researchers to substitute references to BOLD signal with terms like “neural activity” or “neural processing” – despite cautions by Logothetis that “none of [these] techniques [e.g., BOLD fMRI, single unit recordings, LFPs] is a substitute for the others” (Logothetis, Reference Logothetis2008, p. 877). This interpretation even carried over to resting state fMRI, despite evidence that resting state BOLD fluctuations are more weakly correlated with LFP (for review, see Drew, Reference Drew2019), are caused by a variety of non-neural factors (see Fig. 9 in Haueis, Reference Haueis2017), and support a variety of non-neural functions, heterogeneously distributed throughout the brain (Drew, Reference Drew2022). The point is that the BOLD signal follows from particular neural, metabolic, and hemodynamic activities, but these important details are lost when it is simplified into abstract “neural activity.” This abstraction slots BOLD fMRI into a simple theoretical paradigm (specifically, a stimulus–response paradigm), but halts lines on inquiry that could trigger a paradigm shift.

Refocusing on “neural activities” has at least two advantages. First, it helps address a “cognitive myopia,” where neural activities not directly related to cognition are dismissed as noise (Haueis, Reference Haueis2018). The brain is not just for thinking (Theriault et al., Reference Theriault, Katsumi, Reimann, Zhang, Deming, Dickerson, Quigley and Barrett2025): it also “maintains the body and its own structural integrity … [with] neural mechanisms that maintain and repair [its parts]” (Haueis, Reference Haueis2017, p. 252). For example, even the event-elicited BOLD response may support cognitive and noncognitive functions. Specifically, event-elicited BOLD increases cooccur with an atypical metabolic shift, a shift that metabolizes glucose nonoxidatively and threatens to acidify “active” brain regions (for review, see Dienel, Reference Dienel2019; Theriault et al., Reference Theriault, Shaffer, Dienel, Sander, Hooker, Dickerson, Barrett and Quigley2023). Beyond delivering oxygen, BOLD hemodynamics may flush lactic acid produced by this nonoxidative metabolic shift, regulating local pH levels within ranges for normal neural firing (DiNuzzo et al., Reference DiNuzzo, Dienel, Behar, Petroff, Benveniste, Hyder, Giove, Michaeli, Mangia, Herculano-Houzel and Rothman2024). The event-elicited BOLD increase, then, may not be a pure sign of neuro-cognitive processing; rather, it may, in part, stem from neuroprotective feedback mechanisms that buffering against the consequences of atypical and acidifying metabolic and neural activities.

Second, refocusing on “neural activities” may help contextualize the massive metabolic expense of the resting brain (∼20% of the whole-body budget; Clarke & Sokoloff, Reference Clarke and Sokoloff1999; Yu et al., Reference Yu, Herman, Rothman, Agarwal and Hyder2018). Event-elicited metabolism, by contrast, is a metabolic blip, increasing energy consumption ∼5% from baseline (Raichle, Reference Raichle2010). Further, the nonoxidative nature of event-elicited neural activities is distinct from the predominantly oxidative brain metabolism at rest (for review, see Dienel, Reference Dienel2019; Theriault et al., Reference Theriault, Shaffer, Dienel, Sander, Hooker, Dickerson, Barrett and Quigley2023). In other words, the brain does expensive work all the time. To understand why it pays this expense, it may be necessary to scrutinize the dominant stimulus–response paradigm (see Cisek, Reference Cisek1999; Dewey, Reference Dewey1896; Lashley, Reference Lashley1951; Logothetis, Reference Logothetis2008; Powers, Reference Powers1973; Yin, Reference Yin2020) – a paradigm implicit in the project of brain mapping, where stimulus features are mapped to the “neural activity” they elicit (Hutchinson & Barrett, Reference Hutchinson and Barrett2019). Alternatives posit that stimuli are context sensitive: their features mean nothing to an organism on their own, because their meaning is codetermined by how the organism is prepared to receive them (Shannon and Weaver (Reference Shannon and Weaver1964[Reference Shannon and Weaver1949]) – i.e., information encoding requires a preexisting neural context. Thus, the massive expense of the resting brain may go toward neural activities that maintain that context, whereas event-elicited neural activities may help propagate bottom-up perturbations of it (see Theriault et al., Reference Theriault, Shaffer, Dienel, Sander, Hooker, Dickerson, Barrett and Quigley2023).

The “neural activity” concept has encouraged stimulus–response paradigms that focus on a fraction of the many neural activities occurring in the brain. Haueis and Colaço’s proposal may force neuroscientists to grapple with and resolve the contradictions that appear when “neural activities” are taken at face value. This will be a difficult project, but we believe it will ultimately pave the way for an integrative systems neuroscience, unified through a common metabolic thread.