More than 4,500 wild animals tracked by GPS across six years changed how they used their habitat the moment people arrived, with over 65 percent of species shifting behavior in response to human presence alone. The finding, drawn from roughly 11.8 million location points, separates the fleeting effect of people passing through an area from the permanent stamp of roads, buildings, and agriculture. Published in Science on May 21, 2026, the study lands as outdoor recreation continues to grow across the United States, forcing a harder look at how even casual human activity reshapes wildlife movement in places that appear relatively undisturbed.
What the GPS data actually show
The research team analyzed location records from more than 4,500 individual birds and mammals, generating approximately 11.8 million GPS fixes to map how animals responded to two distinct pressures: the physical alteration of land and the real-time arrival of people. To measure that second variable, the researchers used counts of mobile devices and vehicles tallied by U.S. census block during 2019 and 2020, a window that captured both normal activity levels and the sharp drop in human movement during pandemic lockdowns.
That quasi-experimental setup, sometimes called the “anthropause,” gave the team a natural before-and-after comparison. When human traffic dropped, some species expanded their ranges or shifted activity patterns; when it returned, those gains reversed. The result was a clear signal: human presence affected more than 65 percent of species in the dataset, and the effects were strongest in areas with less physical development. Roughly 60 percent of the species that responded showed interdependent effects, meaning the behavioral change appeared only when both dynamic human activity and some degree of habitat modification occurred together. According to the University of St Andrews summary, these combined pressures were most evident where animals still had options to move away, such as forest edges, rangelands, and mixed-use recreation corridors.
That interaction is the study’s sharpest contribution. Earlier peer-reviewed work had already established that human presence and the built environment are non-equivalent drivers of wildlife habitat use, with one widely cited analysis showing that even low-intensity recreation can displace sensitive species in ways that differ from permanent infrastructure. The new GPS dataset, spanning multiple taxa and regions, adds temporal resolution: it shows animals changing behavior as human numbers rise and fall over weeks and months, not just across years of development.
Animals in lightly modified settings, places like exurban edges and trail networks near protected areas, proved most sensitive to incremental increases in foot and vehicle traffic. In heavily developed zones, many species had already adjusted or disappeared, leaving fewer detectable responses. That pattern suggests a saturation point: once a landscape is sufficiently altered, additional hikers or cars may not produce measurable new effects because the remaining wildlife community is already composed of generalists and disturbance-tolerant species.
What remains uncertain
The publicly available summaries report aggregated percentages but do not release species-level response rates or the statistical model coefficients behind the interaction terms. Without those details, outside researchers cannot yet determine which individual species drove the headline numbers or how steeply behavioral shifts scaled with rising human activity. The full paper, published in Science (vol. 392, issue 100, pp. 879-884, DOI 10.1126/science.adq3396), may contain those breakdowns, but the journal article itself is not open access at the time of the institutional releases.
The mobile-device proxy also carries limits that the available summaries do not fully address. Census-block-level device counts capture aggregate traffic but cannot distinguish a hiker on a trail from a driver on a highway, or a group of birdwatchers from commuters passing through. Whether the proxy was validated against on-the-ground field counts, camera traps, or acoustic monitoring is not described in the released materials. That gap matters, because misalignments between where people are counted and where animals actually encounter them could blur or exaggerate the apparent effects.
Separate research on mammal responses during the pandemic found that behavioral shifts depended on trophic group and the degree of surrounding habitat modification, with carnivores, herbivores, and omnivores responding differently to reduced traffic and noise. That earlier work aligns with the new paper’s interaction result but also suggests that predator and herbivore responses may diverge in ways the top-line 65 percent figure does not capture. Without species-level tables, it is impossible to say from the current summaries whether, for example, large carnivores were more likely than small mammals to expand their ranges during low-traffic periods, or whether some groups primarily shifted in time rather than space.
No direct interview statements from the lead authors have been published alongside the institutional summaries. The Yale news office provided framing language, but without on-the-record quotes, the researchers’ own interpretation of the interaction effect and its conservation implications is filtered through press-office paraphrase rather than attributed in their own words. That leaves unanswered questions about how they view trade-offs between recreation access and wildlife disturbance, and whether they see their findings as grounds for new restrictions, better spatial planning, or more targeted outreach to visitors.
How to read the evidence
Three tiers of evidence support the headline claim. The strongest is the Science paper itself, a peer-reviewed study in a top-tier journal with a defined sample size, a replicable proxy for human activity, and a specific time window that includes both normal conditions and the pandemic slowdown. The second tier consists of institutional records from the University of St Andrews and Yale that confirm bibliographic details, core statistics, and the study’s framing. The third tier is syndicated press coverage that repeats those numbers without adding independent verification or new data, sometimes simplifying the interaction effect into a single talking point about recreation impacts.
Readers evaluating the 65 percent figure should note what it does and does not say. It describes the share of species in the sample whose spatial behavior changed detectably when human presence increased. It does not describe the magnitude of those changes, whether they harmed individual fitness or population viability, or whether the shifts were permanent. A deer that moves 200 meters farther from a trail and a raptor that abandons a nesting site entirely could both register as “affected” in the dataset, but the conservation consequences differ enormously. Nor does the percentage tell us how many individual animals within each species responded; some may be habituated while others remain highly sensitive.
The interaction finding, that roughly 60 percent of responding species changed behavior only when human presence and habitat modification coincided, is best read as a warning about cumulative impacts. Trails, roads, and exurban development do not just carve up habitat; they also funnel people into the remaining patches, amplifying disturbance where animals are already squeezed. That perspective echoes prior work showing that disturbance from recreation can rival or exceed the effects of infrastructure in some contexts, particularly for species that rely on predictable refuges from human contact.
At the same time, the study does not imply that all human presence is uniformly harmful or that protected areas must be closed to visitors. The anthropause comparison shows that many species can and do adjust their movements when pressure eases, expanding into spaces that were previously avoided. This flexibility offers a lever for management: by concentrating high-use recreation in already disturbed corridors, limiting access during sensitive breeding seasons, or designing buffer zones around key habitats, land managers may be able to preserve much of the benefit people derive from outdoor spaces while reducing the most disruptive encounters.
Until the full dataset and model outputs are available, the new research is best treated as a strong, system-wide signal rather than a species-by-species rulebook. It confirms that animals are watching us, adjusting to us, and in many cases retreating from us, even in landscapes that appear relatively intact. Translating that insight into policy will require finer-grained analyses, local studies that ground-truth device-based proxies, and public conversations about what level of disturbance we are willing to accept in the places we share with wildlife.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
