A study from Kiel and Copenhagen shows for the first time that even simple nervous systems can learn
Jellyfish can learn from experience, similar to humans or other complex creatures - this has now been shown by a team of biologists from Kiel University (CAU) and Copenhagen University. They trained Carribean box jellyfish (Tripedalia cystophora) to learn to spot and dodge obstacles. The study shows that even simple nervous systems are capable of an advanced form of learning. This may indicate that the evolutionary roots of learning and memory are older than previously thought. It is possible that they were one of the most important evolutionary advantages of creatures with nervous systems from the very beginning. The study appears today (22 September) in the journal Current Biology.
It is no bigger than a fingernail, has a very simple structure and has only a few nerve cells. Yet the box jellyfish has a complex visual system with 24 eyes. It uses them to steer through murky waters of Caribbean mangrove swamps, hunt water fleas and avoid underwater tree roots. "Although they are such simple animals, they have an impressive visual capacity that they use to change their behaviour," says Dr. Jan Bielecki from the Institute of Physiology at Kiel University, describing the scientific appeal of these creatures. He is fascinated by how such simple nervous systems are able to learn and what can be transferred from nature to technical areas such as robotics.
Learning ability is older than previously assumed
Bielecki and his colleagues at the University of Copenhagen have now shown for the first time, that box jellyfish can acquire the ability to avoid obstacles through associative learning. This means that an organism changes its behaviour or attitude based on an experience it has had. "This is a higher form of learning than you would expect from such a creature," says Bielecki, who trains the jellyfish in his laboratory. From an evolutionary point of view, jellyfish are among the first animals to have a nervous system. "If already these animals are able to learn, it could be a basic ability of neurons or neural networks. This suggests that it has existed since the beginning of evolution and thus earlier than previously assumed in research."
For their experiments, the research team simulated the jellyfish's natural habitat with a water tank and grey and white stripes on the inside wall. The grey stripes represented the mangrove roots the jellyfish wants to avoid, the white stripes represent the water environment. The box jellyfish uses colour contrasts to perceive spatial distances, so the researchers varied the contrasts during the experiment.
At the beginning of the experiment, the jellyfish often hit the simulated roots on the tank wall. But after just a few minutes, they had already increased their average distance from the wall by about 50 per cent and were only bumping into them half as often. "These results suggest that jellyfish can learn by combining visual and mechanical stimulus experiences," says Anders Garm, Professor of Marine Biology at the University of Copenhagen, Denmark.
Box jellyfish learns surprisingly fast
"We were really surprised by how fast these jellyfish learned," says Bielecki. This is mainly due to the fact that the researchers worked with the animals' natural behaviour. Avoiding obstacles is something the jellyfish know from their everyday life; it is a "meaningful" behaviour for them. "Learning means combining something new with something familiar. This makes learning a very individual process." In biology, it is also called "SSDR", Species-specific Defence Reaction, when it is a very species-specific behaviour.
To better understand the underlying processes of associative learning in the cube jellyfish, Bielecki then isolated the animals' visual sensory centres, the so-called rhopalia. Each of the four centres contains six eyes, but only 1,000 nerve cells. In addition, electrical signals are generated here that control the jellyfish's movements. Bielecki showed the rhopalium moving grey bars to simulate the jellyfish approaching an obstacle. But it was only when he applied weak electrical stimuli to the rhopalium - simulating an impact against the wall - that it reacted and produced signals that made the jellyfish take evasive action. This enabled Bielecki not only to change the jellyfish's behaviour, but also to localise their learning processes in their rhopalia for the first time.
Applying evolutionary insights to technical pattern recognition
"If you want to understand complex structures like the nervous system, it helps to first study structures that are as simple as possible," says Bielecki. "That way it's often easier to understand the connections and then transfer them." This is exactly what he intends to do in the Collaborative Research Centre 1461 "Neuroelectronics", into which the results of the study will flow. The interdisciplinary research association at the Kiel University, of which Bielecki is a member, is investigating how principles from biological information processing can be transferred to technical systems.
"The fact that the cube jellyfish can recognise patterns with so few nerve cells makes it an ideal model organism for our research," says CRC spokesperson Professor Hermann Kohlstedt from Kiel University. The aim of the large-scale research project is to develop hardware such as electronic circuits that can be used for pattern recognition, for example. "Up to now, this has been done using computer software, which consumes a lot of energy. But we know from nature and evolution that there are much more energy-efficient ways of processing information."
Bielecki et al., Associative learning in the box jellyfish Tripedalia cystophora, Current Biology 33 (2023), 22.09.2023, DOI: 10.1016/j.cub.2023.08.056 https://doi.org/10.1016/j.cub.2023.08.056
Since the beginning of 2021, the German Research Foundation (DFG) has been funding the SFB 1461 "Neurotronics: Biologically Inspired Information Processing" at Kiel University with around 11.5 million euros for an initial four years. Here, researchers from the fields of neuroscience, biology, psychology, physics, electrical engineering, material sciences, network sciences and nonlinear dynamics want to develop new hardware for information processing. The aim is to transfer findings about the information pathways in nervous systems to technical information processing in order to improve, for example, pattern and speech recognition or the energy efficiency of existing systems. Partner institutions in addition to CAU as the speaker university are: Ruhr University Bochum, Ilmenau University of Technology, Leibniz Institute for Innovative Microelectronics Frankfurt/Oder, Leibniz Institute for Science and Mathematics Education Kiel, University Medical Centre Hamburg-Eppendorf and Lübeck University of Technology.