Memory helps living things make better decisions, but holding onto and using past information costs energy and resources. Researchers at the University of Tokyo’s Institute of Industrial Science, working with RIKEN, built a mathematical framework describing how organisms with limited resources balance that tradeoff, publishing their analysis in Physical Review Letters.

The model lets a simulated organism draw on both its current sensory input and memories of past observations as it tries to make sense of a changing environment. “When resources are scarce, the best strategy is to ignore memory and react only to current information,” said lead author Takehiro Tottori. Once enough resources become available, the model shows, the optimal strategy switches abruptly to a memory-based approach rather than shifting gradually — a pattern the researchers describe as a kind of phase transition in decision-making strategy.

The team also found that memory is most valuable when sensory information is moderately uncertain; it adds little when information is already very clear, and not much more when it is too noisy to be reliable. “Whether memory is useful depends not only on the resources available, but also on environmental uncertainty,” said senior author Tetsuya J. Kobayashi. The authors note their results line up with recent behavioral experiments suggesting humans also adjust how much they lean on memory based on both resource availability and sensory uncertainty, offering a theoretical basis for understanding how memory-intensive biological computation, including the brain, may have evolved.

Endnotes

1. Takehiro Tottori and Tetsuya J. Kobayashi, “Theoretical analysis of resource-induced phase transitions in estimation strategies,” Physical Review Letters, July 28, 2026, DOI: 10.1103/5ynb-7k4v.
2. University of Tokyo Institute of Industrial Science, “The cost of memory: how organisms balance thinking and sensing,” news release via EurekAlert!, July 29, 2026.