Many insects are able to use skylight e-vector patterns to deduce their heading direction. Crickets have been well known to orient themselves to certain e-vector orientations to keep their walking direction. However, it is still unknown if crickets are able to utilize polarized light information for spatial recognition. Using an experimental paradigm similar to the Morris water maze for rodents, here we examine the possibility that the cricket Gryllus bimaculatus can utilize polarized light information to find the target place. Crickets were placed in a round arena with a heated floor, a portion of which was cooled, and a cross-shaped e-vector pattern was presented from the top of the arena so that the cricket could find the cool spot by walking along the e-vector direction. When the arrangement of the e-vector pattern and the cool spot were fixed throughout the experiments, the time and the walking distance to find the cool spot were significantly decreased with increasing trials, but not when the e-vector pattern was rotated between each trial. Moreover, a model selection indicated that the visual stimulus contributed to the decrease in time and distance. To investigate the cricket's exploration patterns in the arena, a test trial in which the whole floor was uniformly heated was performed before and after the training trials. In the test trial, the crickets trained with the positionally fixed e-vector pattern showed wall-following behavior for a significantly longer time than those untrained and those trained with random e-vector patterns.