The ancient and cross-culturally prevalent pattern of caregiving suggests that long-term caregiving is species characteristic for humans. If so, then an evolutionary account of the adaptation(s) that underwrite this caregiving is necessary, particularly for the one-sided and long-term nature of Alzheimer's caregiving. Four standard evolutionary explanations are evaluated: kin selection theory, the grandmother hypothesis, direct reciprocity, and indirect reciprocity. Each is found inadequate to explain caregiving because of the lack of reproductive benefits. These evolutionary accounts also assume that relationships are only valuable to the degree that they provide benefits and that relationship partners are predominantly motivated by self-interest. Attachment provides another explanation, which evolved initially to ensure infant protection and nurturance, but was exapted for important adult relationships. Attachment relationships naturally include caregiving and engender long-term relational commitment. Yet attachment theory is ambiguous about whether relationships are maintained for the sake of security benefits or because they have inherent value. This ambiguity undermines the explanatory value of attachment theory for Alzheimer's caregiving. Therefore, a shared identity theory is offered that highlights the inherent value of the relationship and the loved one, transcending the predominant focus on beneficial individual outcomes. The theory emphasizes the frequent human motivation to benefit others because of their mutual commitment, shared identity, and shared goals. The conclusion is that fully understanding and supporting the arduous efforts of caregiving for loved ones with Alzheimer's requires psychologists to fully appreciate and support the deep and meaningful motivations that often inspire the humanity seen in caregiving.