The study on the reproducibility of behavioral experiments with insects has now been published, providing evidence that some results cannot be fully reproduced. This “reproducibility crisis” affects different disciplines, including biomedical research and behavioral studies on mammals. However, there have been no comparable systematic studies on insects – until now.

A team of researchers from the Universities of Münster, Bielefeld, and Jena (Germany) conducted a multi-laboratory approach to test the reproducibility of ecological insect studies. They performed three different behavioral experiments using different insect species: the turnip sawfly, meadow grasshopper, and red flour beetle.

Each experiment was carried out in laboratories in Münster, Bielefeld, and Jena, and the results were compared. The studies examined the effects of starvation on behavior in larvae of the turnip sawfly, the relationship between body color and preferred substrate color in grasshoppers, and the choice of habitat in red flour beetles.

To the research team’s knowledge, this study is the first to systematically demonstrate that behavioral studies on insects can also be affected by poor reproducibility. This was surprising, as insect studies generally use large sample sizes and could provide more robust results. However, reproducibility was higher compared to other systematic replication studies not carried out on insects.

The results are of particular interest to scientists in behavioral biology and ecology but also for all disciplines where behavioral experiments are conducted with animals. The research team concludes that deliberately introducing systematic variations could improve reproducibility in studies with living organisms.

The amazing diversity of life on our planet is a testament to the multitude of biological solutions that have evolved to secure and maintain energy. However, despite its central role in biology, measuring energy consumption remains a challenging task. One major drain for many animals is movement, making it an ideal lens through which to estimate energy usage. While methods exist for measuring animal movement, they are often limited by the physical size of the equipment used.

In a groundbreaking study published in the Journal of Experimental Biology, researchers from the Marine Biophysics Unit at the Okinawa Institute of Science and Technology (OIST) have developed an innovative method for measuring energy usage during movement using video and 3D-tracking via deep learning. This new approach opens up the possibility of studying energy consumption in animals that were previously inaccessible due to the reliance on wearable equipment.

The current state-of-the-art method, Dynamic Body Acceleration (DBA), involves measuring oxygen consumption while an animal performs a specific behavior in a lab setting. However, this method has limitations when applied in the wild, where reliably measuring oxygen consumption is impossible. To overcome these challenges, researchers have used physical accelerometers that weigh at least ten times less than the animal, but this still rules out the study of many small species.

The OIST researchers’ solution to this problem is elegantly simple: they use two cameras to capture video footage of an animal’s behavior from multiple angles, reconstructing its movement in 3D space. A deep learning neural network is then trained on a few frames of the videos to track the position of body features such as eyes, allowing researchers to subsequently measure the movement-related acceleration.

This new video-based DBA method has opened up possibilities for studying energy consumption in animals that were previously inaccessible, potentially enabling many new research avenues into the breadth of life on our planet. For example, researchers can now investigate the energy expenditure during schooling of small fish, which has long remained mysterious. By accurately measuring energy usage during free-ranging animal behavior, scientists can gain a deeper understanding of the ecology and evolution of various species.

The use of wood in construction has been a staple for millennia, from the majestic Japanese cypress to the sturdy ponderosa pine. Despite its environmental benefits, wood’s susceptibility to damage from sunlight and moisture often pushed it aside in favor of steel and concrete. However, with the growing interest in sustainable building practices, wood is making a comeback.

To overcome the challenges associated with wooden structures, researchers at Kyoto University have developed a groundbreaking method for detecting early signs of coating deterioration. This simple yet effective approach combines mid-infrared spectroscopy with machine learning to predict the extent of degradation before it becomes visible.

The team’s innovative technique uses partial least square regression and genetic algorithms to identify subtle chemical changes in wood coatings. These slight alterations, often too small to detect visually, can be accurately captured by infrared spectroscopy and predicted by the model. This enables researchers to diagnose early coating deterioration with high accuracy, reducing the need for costly visual inspections and preventing further decay.

By integrating chemistry and data-driven modeling techniques, this research demonstrates how traditional craftsmanship and modern data science can work together to support smarter maintenance of sustainable buildings. As Teramoto notes, “We hope this technology will help bridge the gap between traditional craftsmanship and modern data science.”

The researchers are now conducting tests on real wooden buildings, with plans to improve their model for application in new paint and coating product development. Beyond wood, this method may also be applied to materials like concrete or metal, unlocking new possibilities for diagnosing early material failure and improving sustainability across various industries.