Survival training in a safe space

A group of meerkats. These African mammals use controlled learning to prepare their young for the dangers of everyday life.
Photo Credit: Dušan Veverkolog

Scientific Frontline: Extended "At a Glance" Summary: Protected Learning Environments in Animal Development

The Core Concept: Protected learning is a biological mechanism in which adult animals create staged, risk-mitigated developmental spaces, allowing offspring to safely acquire essential survival skills without facing immediate lethal consequences.

Key Distinction/Mechanism: Unlike unassisted trial-and-error learning in the wild—which poses a significant threat to inexperienced juveniles hunting dangerous prey—this process relies on graduated risk exposure (e.g., adult meerkats offering dead, then disarmed, then fully intact venomous scorpions to their young). A critical finding is that if the developmental environment is too safe and diverges significantly from reality (analogous to "helicopter parenting"), maladaptation occurs, leaving the animal unprepared to cope with genuine risks in adulthood.

Major Frameworks/Components:

• Two-Phase Learning Framework: A developmental model simulating the transition from a protected juvenile stage to an unprotected, hazardous adult environment.
• Dynamic Programming: A mathematical optimization method used to calculate the theoretically ideal behavioral strategy under varying environmental conditions.
• Reinforcement Learning: A computational approach employed to simulate the trial-and-error processes through which individuals acquire survival strategies over time.

Branch of Science: Computational Biology, Theoretical Biology, Ethology (Animal Behavior), and Evolutionary Biology.

Future Application: The models provide a baseline for future research into complex social learning (learning through observing others). They offer frameworks for understanding early-life education, species-specific variations in parental care, and potential cross-disciplinary applications in human developmental psychology.

Why It Matters: This research provides rigorous theoretical and mathematical proof for parental care strategies long observed in nature. It illustrates precisely how early-life experiences shape adult performance and how biological organisms successfully balance developmental risks against evolutionary rewards to ensure survival.

Visualisation of the mathematical model used in the study. Panel A (top) illustrates decision-making during hunting and the possible outcomes and consequences. Panel B (bottom) depicts the two-phase learning model.
Image Credit: Dharanish Rajendra / Universität Würzburg

To improve their chances of survival, animals must learn – and that can be dangerous. A new study from the University of Würzburg shows how gradual learning under parental supervision can reduce these risks. 

Adaptation is essential for survival. Across species, it occurs over many generations through evolution and natural selection. Individual animals, however, can also adapt within their own lifetimes – through learning. For many species, humans included, this process is vital. 

The challenge arises when adaptation must take place in a hazardous environment, such as when hunting dangerous prey. “In such cases, learning through exploration can pose a serious threat to a juvenile’s survival,” explains Dharanish Rajendra. He is a doctoral researcher working with Professor Chaitanya Gokhale, Chair of Computational and Theoretical Biology at Julius-Maximilians-Universität Würzburg (JMU). 

One solution is parental care: adults create protected learning environments in which offspring can gradually approach the challenges of real life. 

In a new publication, the two researchers use mathematical and computational models to investigate how such learning environments can be structured to promote successful adaptation. 

Step by step towards competence 

Many predators specialize in prey that represent a substantial risk for inexperienced hunters. Wolves take on elk and bison many times their size; pythons must subdue porcupines, and meerkats frequently prey on venomous scorpions. 

These small mammals, native to southern Africa, live in colonies of up to 45 individuals and are well known for their complex social behavior. To prepare their young for handling dangerous prey, adults proceed in stages: first offering dead scorpions, then scorpions with the sting removed, and only later presenting fully intact, dangerous prey once the juveniles have developed sufficient skill. 

When protection goes too far 

In their study, the researchers modelled this developmental transition using a two-phase learning framework, simulating the shift from a protected juvenile stage to an unprotected adult environment. 

Their findings highlight a risk reminiscent of so-called “helicopter parenting”: if the learning environment is too safe, or differs too greatly from real conditions, maladaptation can occur. Individuals may reach adulthood insufficiently prepared and struggle to cope with genuine risks. 

“What matters most for successful protected learning is that the environment remains sufficiently similar to reality,” says Chaitanya Gokhale. “A gradual increase in risk bridges the gap between a safe developmental space and the demands of the real world.” 

The mathematics behind behavior 

To analyze these behavioral strategies, the researchers combined two complementary modelling approaches. 

Dynamic programming was used to calculate the theoretically optimal strategy under different environmental conditions. Reinforcement learning was then employed to simulate how individuals acquire such strategies through trial and error. 

Together, these methods show how an animal learns to weigh risks against rewards and how early-life experience shapes later performance. 

“Our research provides a theoretical foundation for parental care strategies that have long been observed in nature, but whose underlying mathematical logic has remained unclear,” says Dharanish Rajendra. The findings contribute more broadly to our understanding of early-life learning in both animals and humans, as well as species-specific differences in parental care. 

Because many species that provide extended parental care are also highly social, the researchers see promising directions for future work. In particular, they aim to investigate how protected developmental environments interact with social learning – learning through observing and drawing on the experiences of others. 

Published in journal: Proceedings of the Royal Society B

Title: Optimising play for learning risky behaviour

Authors: Dharanish Rajendra, and Chaitanya S. Gokhale

Source/Credit: Julius-Maximilians-Universität Würzburg | Lutz Ziegler

Reference Number: bs030226_01

Privacy Policy | Terms of Service | Contact Us