The extent of increased brain activity’s contribution is uncertain, but functional reorganization in brain during early development should allow some kinds of compensation. Since functional reorganization only appears in some patients but not the others, patterns of memory impairments may be inconsistent despite similar brain lesion. Therefore, it is difficult to determine whether a brain region is necessary for a particular cognitive function. Besides brain plasticity, successful adoption of compensatory strategies could sometimes lead to different presentations of overt impairments. This was shown in patient AV, who suffered from DA but with normal verbal IQ (Brizzolara et al., 2003). Although AV was impaired in both episodic and semantic memory tasks, it was found that AV advantaged relatively more from adoption of semantic strategy than controls. This means that strategies could compensate memory impairments to a certain extent. While most of the studies do not investigate patients’ strategy use, it is uncertain whether their preserved semantic memory is due to normal or compensatory cognitive functioning. As a result, it leads to discrepancies between patients’ impairments and makes it difficult to map cognitive process neurologically. Given the plasticity of brain and adoption of compensatory strategies, the predictability of memory impairments based on brain lesion is low. When neither of the models could accurately predict patterns of impairment, it is difficult to tell whether one is superior to the other. Although it is appealing to justify either Squire or A’s model based on superficial evidence, this essay has discussed their limited generalizability, testability and predictability. The dilemma between single dissociation in human cases and double dissociation in animal studies makes it difficult to determine whether an associative or dissociative model better describes amnesia. Moreover, testability limited by nature of memory leads to problems in verification of models. Cognitive process of memory is not overt, and organic amnesia provides a chance for visualization of it in a biological way. Nevertheless, brain plasticity and compensatory strategies obstruct localization of neural substrates underlying declarative memory. Currently, there is no promising treatment for amnesia, but establishing a good model could facilitate our understanding of it. This essay has discussed the difficulty in judging whether Squire or A&B’s model is better, but by no means they should be disregarded. Despite limited testability, models of amnesia provide framework and direction for future studies. With sufficient caution during evaluation, the strengths and weaknesses of evidence could still enrich our understanding of amnesia and memory in general.