Neuroscience review panels have quietly reorganized around a single model organism. Mouse atlases, two-photon rigs, and optogenetic toolkits now set the standard for what counts as a strong proposal, and human cortex work pays for it. This piece traces how that happened and what it costs.

Mouse Maps Crowd Out Human Cortex

A reviewer opening a grant panel folder sees a familiar pattern. The mouse connectome proposal has a clean pipeline, a public dataset, and a track record of published circuits. The human cortex proposal has a tissue-sharing agreement, a scanner queue, and a sample size that will not survive a power calculation. The first one scores higher almost by default.

That default is not a conspiracy. It is what happens when a field standardizes its infrastructure around one species and then asks every other species to compete on the same terms. Human cortex work needs fresh postmortem tissue, living surgical samples, or high-field MRI time, and each of those has a queue, a consent process, and a cost line that rodent work does not carry. A single high-field MRI session for a human cortex study can run into thousands of dollars, while a mouse imaging session costs a fraction of that. The tissue itself is scarce: a brain bank might release only a few dozen human cortical samples per year, each with limited viability for certain assays. These constraints mean that a human cortex proposal often requires a budget an order of magnitude larger than a comparable rodent study, with no guarantee that the tissue will yield publishable results.

Panels reward proposals they can evaluate with confidence. A mouse study fits the template. A human study often does not, and reviewers who have never run one may not know how to weigh its risks. The result is a funding gradient that favors the model organism with the best tooling, not the one closest to the clinical question.

How Rodent Tools Became the Standard

The Allen Institute's mouse connectivity atlas gave the field a shared reference brain. Once that existed, every new circuit paper could be placed in a common coordinate space, and every reviewer could compare it to a known map. Human cortex has no equivalent at the same resolution, largely because the tissue is harder to get and harder to keep viable. The atlas also provided a common set of tools: standardized image registration, cell-type classification, and connectivity matrices that any lab could download and use. That infrastructure lowered the barrier to entry for rodent work and created a virtuous cycle: more labs used the atlas, more data were contributed, and the atlas became even more useful.

Two-photon calcium imaging matured in mice. The skull is thin, the cortex is accessible, and genetic lines express indicators reliably. Optogenetics was validated in rodents too, which means the causal tools that let a lab say a circuit does something were built and benchmarked in the same animal. Human tissue rarely supports the same methods, so the human work cannot borrow the toolkit. Even when human tissue is available, it is often from older donors or patients with neurological conditions, which complicates the interpretation of circuit-level findings. The result is that human cortex research often relies on correlational methods like functional MRI, which measure blood flow rather than direct neural activity, and cannot establish causality in the way that optogenetics can in mice.

This is a case of method diffusion running in one direction. A technique perfected in a model organism becomes the field's definition of a good experiment, and researchers who cannot use it are judged against a standard they cannot meet. A related piece on this site describes how diffusion tensor imaging moved from brain tracts into porous media flow, which is the same pattern in a different field: a method travels, and the destination field reorganizes around it.

The Economics of Model Organism Choice

Mouse lines are cheap to maintain and breed quickly. A colony can support several parallel projects, and a graduate student can run a full circuit study within a thesis timeline. Primate work costs far more per animal and carries longer approval and observation windows. Human cortex work sits at the far end of that spectrum, needing tissue, imaging time, and often a clinical partner. For example, a single nonhuman primate can cost tens of thousands of dollars, and the approval process for primate research can take years. Human cortex work is even more expensive: a postmortem tissue sample might cost several thousand dollars, and high-field MRI time can run into hundreds of dollars per hour. These costs are not just financial; they also involve ethical and logistical hurdles that rodent work avoids.

Publication speed reinforces the gradient. Rodent studies publish faster, and faster papers feed tenure clocks and renewal applications. A panel that funds a mouse proposal can expect outputs within a grant cycle. A human cortex project may produce its first major paper after the cycle ends, which makes it look like a worse bet even when the science is stronger. This temporal mismatch is exacerbated by the fact that many human cortex studies require long-term follow-up or postmortem analysis, which can take years to complete. In contrast, a mouse study can be designed, executed, and published within two to three years, aligning neatly with the typical grant reporting period.

Panels default to low-risk, high-throughput proposals because that is what their scoring rubrics measure. The rubric rarely asks whether the model organism can answer the question. It asks whether the methods are established, the sample is adequate, and the timeline is realistic. Mouse work wins on all three, and the human relevance of the question is treated as a separate, softer criterion.

What Gets Lost in Translation

Human cortex has cell types and laminar organization that rodents do not share. Some interneuron subtypes, some projection patterns, and some gene expression profiles are human-specific or primate-specific. A circuit claim established in mouse cortex may simply have no counterpart in the human tissue, and the paper will not say so because the experiment was never run there. For instance, the human cortex contains a higher proportion of certain inhibitory neurons, such as double bouquet cells, that are rare or absent in rodents. These differences matter for understanding conditions like schizophrenia and autism, where the balance of excitation and inhibition is thought to be disrupted. If the mouse model lacks the relevant cell types, then any circuit-level finding may not translate.

Rodent findings often fail to replicate in humans. Disease models miss human-specific gene expression, which is a particular problem for psychiatric and neurodegenerative conditions where the mouse model cannot reproduce the human genetic architecture. Circuit claims rarely cross the species boundary cleanly, and the field has accumulated a large literature of findings whose human relevance is assumed rather than tested. A 2019 study in a leading journal found that only a small fraction of rodent findings in certain areas of neuroscience were replicated in human studies, though the exact figure is debated. The replication crisis in neuroscience is not limited to rodents, but the reliance on a single model organism amplifies the problem because it narrows the range of hypotheses tested.

The cost is not only scientific. It is also infrastructural. Every dollar spent on mouse-specific tooling is a dollar not spent on human tissue banks, high-field imaging access, or the surgical partnerships that make human cortex work possible. The imbalance compounds, because the field with better infrastructure produces more papers, and more papers justify more infrastructure.

Signs of a Correction

The BRAIN Initiative has pushed human tissue pilots and technology development aimed at human-relevant measurement. Some panels now weight human relevance more heavily, and a few funding calls explicitly ask for cross-species validation. Consortia have begun sharing human cortex data, which lowers the entry cost for labs that cannot collect their own tissue. For example, the NIH BRAIN Initiative has funded projects to create a human brain atlas at cellular resolution, though it is far from complete. These efforts are promising but still underfunded compared to rodent initiatives. The imbalance persists because the infrastructure for rodent work is mature and self-sustaining, while human work requires ongoing investment to build comparable resources.

Rodent work still dominates by volume. The correction is real but partial, and it runs through a system that rewards the model organism with the best tools. A related piece on this site notes that climate models absorb most of their cost after publication, which is a reminder that infrastructure choices shape a field long after the grant that paid for them has closed.

Reviewers who want to shift the balance have limited levers. They can ask whether a rodent model answers a human question, and they can fund the human-relevant infrastructure that makes the comparison possible. Neither is a quick fix, and both cost money that panels have to take from somewhere.

Practical Steps for Researchers and Reviewers

Ask whether the rodent model actually answers the human question before scoring a proposal on its methods alone. A clean mouse circuit paper is not evidence about human cortex unless someone has checked the cross-species claim.

Budget for human tissue or imaging access in the first draft of a grant, not as an afterthought. A line item for tissue procurement or scanner time signals that the project takes the human side seriously.

Cite cross-species validation gaps explicitly in the background section. Naming what is unknown about the human counterpart makes the proposal harder to dismiss as incremental.

Push panels to fund human-relevant infrastructure, including tissue banks and shared imaging time, rather than treating it as a separate initiative. The infrastructure is what makes the human work competitive.

Track funding ratios for animal versus human work at your institution or in your program area. The numbers are often more persuasive to a review committee than an argument about model validity.