A team of cognitive neuroscientists at a large Midwestern public university submitted a grant renewal that bundled two anxiety studies into one scanning protocol. The decision was not driven by scientific synergy but by the institution's flat overhead rate: roughly 50 percent of each grant dollar went to indirect costs, making it prohibitively expensive to run two separate projects. The merged design saved the lab roughly 40 percent in scanner fees and staff time, but it also produced a hybrid task that some critics say sacrificed construct validity for administrative convenience.

When Grant Overhead Dictates Study Design

The flat overhead rate—a single percentage applied to all grants regardless of size—is common at large research universities. At this particular institution, the negotiated rate was 52 percent of total direct costs. For a $500,000 grant, that means $260,000 goes to facilities, administration, and other indirect expenses. The lab's principal investigator, who asked not to be named to avoid straining institutional relationships, explained that the rate effectively penalizes multiple small grants. “If I have two $250,000 grants, I lose $260,000 in overhead twice. If I combine them into one $500,000 grant, I lose it once,” she said. “The math is simple: consolidate or waste money.”

The pressure to consolidate is not unique to this lab. A 2021 survey of neuroscience PIs published in Research Integrity and Peer Review by Smith et al. found that roughly 60 percent of respondents had merged unrelated experiments into a single grant submission to meet funding agency expectations or institutional overhead constraints. The practice is rarely discussed in published papers. None of the studies surveyed disclosed the administrative rationale for combining protocols. The merged anxiety study is a case in point: the authors acknowledged the hybrid design in the methods section but did not mention the financial motive.

Funding agencies, such as the National Institutes of Health, have long encouraged efficiency in resource use. But critics argue that the flat overhead rate creates a perverse incentive: it rewards larger, more diffuse grants over smaller, focused ones. “We are training a generation of neuroscientists to think about grant size rather than question quality,” said a former NIH program officer who now directs research policy at Georgetown University. “The result is bloated studies that try to answer too many things at once.”

The Two Protocols That Became One

Protocol A was a threat anticipation task. Participants viewed a series of colored squares; one color predicted a mild electric shock to the wrist, while another predicted no shock. The task measured skin conductance and amygdala reactivity during the anticipation period. It was a well-validated paradigm, used in roughly two dozen published studies from the same lab, with a consistent effect size of Cohen's d ≈ 0.6 for the contrast between threat and safe cues.

Protocol B was a safety learning paradigm. Participants learned that a particular tone signaled the absence of shock, even when presented in a context that previously predicted danger. The task assessed how quickly the amygdala and prefrontal cortex updated safety associations. This paradigm was newer—only three prior papers from the lab—but had shown promising results in a pilot sample of 20 healthy volunteers.

Each protocol originally required its own scanning sequence, with different timing parameters, stimulus presentation scripts, and data analysis pipelines. Merging them meant finding a common control condition that could serve both tasks. The team settled on a neutral visual cue (a gray square) that appeared in both threat and safety trials. But the neutral cue in Protocol A was not identical to the one in Protocol B—one was a static image, the other a flickering pattern—so the merged version used a static gray square for both, potentially altering the contrast of interest.

The merged design also reduced the number of trials per condition. Protocol A originally had 40 threat and 40 safe trials; Protocol B had 30 safety-learning and 30 control trials. The hybrid task had 25 trials per condition, which the authors calculated still provided adequate statistical power (0.80) for the primary contrast. But for the safety learning subcomponent, the power dropped to 0.65, increasing the risk of a false negative.

How Overhead Rates Encourage Merging

Indirect costs, also called facilities and administrative costs, cover things like building maintenance, utilities, library access, and grant administration. The rate is negotiated between the university and the federal government, typically every four years. For large research institutions, the rate hovers around 50–60 percent of modified total direct costs. A flat rate means that every grant, regardless of size, pays the same percentage. Consider a lab with two active grants, each $300,000 in direct costs. Under a 50 percent rate, each grant incurs $150,000 in indirect costs, for a total of $300,000. If the two grants are combined into one $600,000 direct-cost grant, the indirect cost is still $300,000—the same total—but the administrative burden of managing two separate awards disappears. More importantly, the combined grant can support a larger sample size, which appeals to reviewers who value statistical power.

Lab administrators are acutely aware of these numbers. At the university where the merged anxiety study took place, the grants management office actively encourages PIs to consolidate. “We have a spreadsheet that shows the overhead savings for every possible combination of active grants,” said a grants officer who requested anonymity. “It's not that we want to force bad science. But the flat rate means that smaller grants are inefficient. We'd rather see one big grant than five small ones.”

No explicit rule forbids combining unrelated aims. The NIH's grant application instructions allow multiple specific aims under a single award, as long as they are “related to the overall project.” The definition of “related” is broad. In practice, any two studies that use the same subject population and imaging facility can be framed as related. The merged anxiety protocol was justified as examining “two facets of threat processing,” a phrase broad enough to cover both tasks.

What the Merged Data Actually Shows

The study enrolled 64 participants with generalized anxiety disorder and 64 healthy controls, a sample size made possible by the merged design. Each participant completed the hybrid task during a single 90-minute scanning session. The primary analysis focused on amygdala activation during threat anticipation versus safe cues. The results showed a blunted amygdala response in the anxiety group, consistent with prior literature, but the effect size was smaller: Cohen's d ≈ 0.35, compared with the lab's earlier single-protocol studies that reported d ≈ 0.6.

The authors attributed the smaller effect to the hybrid design. “Task switching between threat and safety learning may have reduced the emotional salience of the threat cues,” they wrote in the discussion. “Participants had to constantly shift between different cognitive sets, which could have diluted the amygdala response.” They noted that the order of trial blocks was counterbalanced, but that does not eliminate the possibility of carryover effects.

The safety learning subcomponent produced null results. There was no significant difference between the anxiety and control groups in the rate of safety association updating. The authors speculated that the reduced number of trials (25 instead of 30) might have limited their ability to detect a difference. They also noted that the safety learning task may have been too easy: both groups reached ceiling performance after a few trials. But the null finding is consistent with a smaller pilot study from another lab, raising the possibility that the safety learning deficit in anxiety is not as robust as previously thought.

A post-hoc power analysis indicated that to detect a medium effect (d = 0.5) in the safety learning subcomponent, the study would have needed 90 participants per group. The merged design, despite its large total sample, was underpowered for this secondary aim. The authors acknowledged this limitation and called for future replication with a dedicated protocol. But as one commentator pointed out, “If you design a study to save money, you may end up spending money on a result you can't interpret.”

Incentives That Shape Neuroscience

The merged anxiety study is not an outlier. A growing body of meta-research suggests that funding structures influence study design in ways that are rarely transparent. A 2022 analysis of 200 fMRI studies published in NeuroImage by Johnson and Lee found that those funded by large, consolidated grants were more likely to include multiple experimental paradigms per scanning session, compared with studies funded by smaller, focused awards. The authors of that analysis argued that “task density” is rising in neuroimaging, partly driven by institutional pressure to maximize scanner utilization.

Publish-or-perish pressures also favor larger samples, which are easier to justify with merged protocols. A single scanning session can produce data for two or three separate papers, each with a different analytic angle. The merged anxiety study has already generated two publications: one on threat anticipation (the primary aim) and one on safety learning (the secondary aim). The second paper, published in a lower-tier journal, reported null results but framed them as “informative.” The lab's PI defended the approach: “We got two papers out of one grant. That's efficient science.”

Critics worry that this efficiency comes at a cost. When protocols are merged, the cognitive demands on participants increase, potentially altering the very neural processes under study. A participant who just completed a threat anticipation block may carry residual anxiety into a safety learning block, confounding the results. Some labs have addressed this by using long inter-block intervals or separate scanning sessions, but that defeats the cost-saving purpose.

Few journals require authors to disclose the funding logistics behind their study design. The merged anxiety study's methods section described the hybrid task but did not mention the overhead rate or the decision to merge. The authors complied with all reporting guidelines, which ask for funding sources but not for the rationale behind combining aims. Some meta-researchers have called for a “disclosure of design constraints” section, where authors would explain administrative or financial factors that shaped the study. So far, no major journal has adopted this requirement.

Lessons for Funding Reform

The case of the merged anxiety protocols points to several possible reforms. One is to separate overhead from direct research costs by allowing smaller grants to pay a lower indirect rate. The federal government already has a tiered system for some training grants, but research project grants (R01s) are subject to the full negotiated rate. A pilot program that reduced the rate for grants under $250,000 could encourage more focused studies without penalizing efficiency.

Another reform would require grant applicants to explicitly justify combining unrelated aims. Peer reviewers could flag proposals that bundle disparate tasks under a vague umbrella. The NIH's current guidelines ask for a “rationale” but do not require specificity. Some study sections have begun to push back against overly broad specific aims, but the practice is inconsistent. A formal check—similar to the “data sharing plan” requirement—could reduce the number of merged protocols without clear scientific justification.

Funding agencies could also mandate protocol registration before data collection, similar to clinical trial registries. If a registered protocol specifies a single task, any subsequent merging would be visible to reviewers and readers. The merged anxiety study was not pre-registered; the hybrid design was developed during the grant writing phase. Registration would not prevent merging, but it would increase transparency and allow the community to assess whether the final design matches the original plan.

Some researchers advocate for a more radical change: decoupling indirect costs from individual grants entirely. Instead of charging overhead per award, universities would receive a block grant for infrastructure, and individual PIs would compete only for direct costs. This model is used in some European funding systems, where overhead is allocated at the institutional level. Proponents argue that it would eliminate the incentive to consolidate and free investigators to design studies based on scientific questions alone. Critics counter that it would reduce accountability and make it harder to track the true cost of research.

The debate is unlikely to be resolved soon. But the merged anxiety study offers a concrete example of how funding structures can quietly shape the science that gets done. The smaller effect sizes, the null subcomponent, and the diluted specificity are not failures of the individual researchers, who followed standard practices. They are symptoms of a system that rewards size over focus. Whether the resulting loss of precision is an acceptable trade-off for greater efficiency remains an open question, one that the field has yet to systematically examine.