A decade of studies on long-term habituation (LTH) in the crab Chasmagnathus is reviewed. Upon sudden presentation of a passing object overhead, the crab reacts with an escape response that habituates promptly and for at least five days. LTH proved to be an instance of associative memory and showed context, stimulus frequency and circadian phase specificity. A strong training protocol (STP) (≥15 trials, intertrial interval (ITI) of 171 s) invariably yielded LTH, while a weak training protocol (WTP) (≤10 trials, ITI = 171 s) invariably failed. STP was used with a presumably amnestic agent and WTP with a presumably hypermnestic agent. Remarkably, systemic administration of low doses was effective, which is likely to be due to the lack of an endothelial blood-brain barrier. LTH was blocked by inhibitors of protein and RNA synthesis, enhanced by protein kinase A (PKA) activators and reduced by PKA inhibitors, facilitated by angiotensin II and IV and disrupted by saralasin. The presence of angiotensins and related compounds in the crab brain was demonstrated. Diverse results suggest that LTH includes two components: an initial memory produced by spaced training and mainly expressed at an initial phase of testing, and a retraining memory produced by massed training and expressed at a later phase of testing (retraining). The initial memory would be associative, context specific and sensitive to cycloheximide, while the retraining memory would be nonassociative, context independent and insensitive to cycloheximide. Key wordsSimple and basic models to study memory Simple-system approaches using invertebrate species have led to considerable progress in our understanding of the mechanisms underlying learning and memory (e.g., the mollusks Aplysia (1,2) and Hermissenda (3,4), the fruit fly Drosophila (5), and the honey bee Apis (6)). A virtue usually attributed to invertebrates is that they are simple models though the use of this term has been highly controversial. Indeed, the CNS of invertebrates cannot be considered simple from the perspective of a neural network researcher since the amount of neurons involved in a current mnemonic process is assumed to be far larger than that required to obtain a satis-