Scientists from the Skolkovo Institute of Science and Technology (Skoltech) have developed a mathematical model of memory that suggests the optimal number of senses for learning and retaining information is seven. The research, published in the journal Scientific Reports, offers new insights into the fundamental principles of memory and could have significant implications for the development of artificial intelligence and robotics. The study challenges the notion that our five senses represent a biological limit, proposing a mathematical basis for a more nuanced sensory experience.

The team’s work is built around the concept of memory “engrams,” which describe how memories are physically stored. An engram can be thought of as a small, dedicated group of brain cells, or neurons, that activate together when a specific memory is recalled. In their model, each concept, like a banana, is an object defined by several features in a “conceptual space.” These features correspond to sensory inputs—what a banana looks like, smells like, tastes like, and so on. For humans, this creates a five-dimensional concept. The model simulates how these engrams evolve, becoming sharper with learning or fading through forgetting, based on interactions with the outside world.

Analyzing the model’s dynamics, the researchers discovered that the system of engrams eventually reaches a stable, or “steady,” state. In this mature state, the number of distinct concepts held in memory remains constant. The most surprising finding came when they tested how the dimensions of this conceptual space—the number of senses—affected its total capacity. The team found that the system’s capacity for holding different concepts was at its highest when the space had exactly seven dimensions.

“It appears that when each concept retained in memory is characterized in terms of seven features—as opposed to, say, five or eight—the number of distinct objects held in memory is maximized,” said study co-author Nikolay Brilliantov, a professor at the Skoltech Artificial Intelligence Center, in an institutional press release. While he notes that applying this to human evolution is speculative, the discovery provides a potentially crucial design principle for AI. By programming robots with seven distinct sensory inputs, engineers could maximize their ability to learn about and remember their environment.

References

• Otieno, W., Tyukin, I. Y., & Brilliantov, N. (2025). The critical dimension of memory engrams and an optimal number of senses. Scientific Reports, 15(1), 29972. https://doi.org/10.1038/s41598-025-11244-y

• Posunko, N. & Skolkovo Institute of Science and Technology. (2025, September 8). Mathematical model of memory suggests seven senses are optimal. Tech Xplore; Skolkovo Institute of Science and Technology. https://techxplore.com/news/2025-09-mathematical-memory-optimal.html