Planarians are comparatively simple invertebrates, evolutionary close to the predecessors of vertebrates, featuring a rope-ladder nervous system containing only tens of thousands of neurons. We studied Girardia tigrina, which has about 40,000 neurons in total, roughly a third of them in the cerebral ganglia, homologous to the vertebrate brain. We aimed to determine whether this animal can make an early decision when faced with two equally attractive targets.

Sixty-five animals were placed in a Petri-dish with two millimeters of water at a marked starting point. Two black cubes, each measuring 20x20x20 mm, were positioned 25 mm away from the start. The angle between the straight lines from the starting point to the middle of the cubes was approximately 90°. Each animal performed five runs. Of 325 runs in total, 324 ended at a cube. When averaging all paths leading to the left cube and all paths leading to the right cube, these two average paths formed a distinct V-shape.

We simulated a gradient descent based on a one-dimensional circular optical field. The overlay of hundreds of simulated paths produced a Y-shape with a high central node, nearly a T-shape. These findings indicate that the planarians were able to segregate the visual input into two separate targets and decide, right from the start, which to approach. The mechanisms allowing the planarians to do this are still unclear - they must be more complex than simple gradient descent.