Long-term potentiation (LTP) resembles memory in that it is initially unstable and then, over about 30 min, becomes increasingly resistant to disruption. Here we present an hypothesis to account for this initial consolidation effect and consider implications that follow from it. Anatomical studies indicate that LTP is accompanied by changes in spine morphology and therefore likely involves cytoskeletal changes. Accordingly, theta bursts initiate calpain-mediated proteolysis of the actin cross-linking protein spectrin and trigger actin polymerization in spine heads, two effects indicative of cytoskeletal reorganization. Polymerization occurs within 2 min, has the same threshold as LTP, is dependent on integrins, and becomes resistant to disruption over 30 min. We propose that the stabilization of the new cytoskeletal organization, and thus of a new spine morphology, underlies the initial phase of LTP consolidation. This hypothesis helps explain the diverse array of proteins and signaling cascades implicated in LTP, as well as the often-contradictory results about contributions of particular molecules. It also provides a novel explanation for why LTP is potently modulated by factors likely to be released during theta trains (e.g., BDNF). Finally, building on evidence that normal patterns of activity reverse LTP, we suggest that consolidation provides a delay that allows brain networks to sculpt newly formed memories.