In the previous companion paper, the possibility of learning by Chasmagnathus in field conditions was demonstrated. Here, we study long-term memory inquiring to what extent an internal representation could be maintained in a complex environment. Two 45-min training sessions, each of 15 visual danger stimulus presentations with 3-min intertrials, were given at a 24-h interval. Throughout the first training session and during the first 22.5 min of re-training on day 2, crabs kept the same hiding response level but then, during the second phase of re-training, the re-emerging increased up to the point that 32% of surface crabs ignored the stimulus. Each session was followed by a 22.5-min testing without stimulation. At testing on day 2 after re-training, crabs showed a change in the usual exploring strategy. Results reveal long persistency in responding despite a rest interval of 24 h followed by a gradual decrease in response until it vanishes. The statistical analysis was performed by comparing performances at day 2 (Rescorla in Am Psychol 43:151-160, 1988) and then confirmed through comparisons between day 1 and day 2. However, it is not possible to identify separate and taxonomically well-defined learnings but rather a tangled collection of processes that influence each other blurring some of the diagnostic features of each learning.