We theoretically investigate long-term potentiation (LTP) in the hippocampus using a simple model of a neuron stimulated by three different time-structured input signals (regular, Markov, and chaotic). The synaptic efficacy change is described taking into account both N-methyl-D-aspartate (NMDA) and non-NMDA receptors. The experimental results are successfully explained by our neuron model, and the remarkable fact that the chaotic stimuli in the nonstationary regime produce the largest LTP is discussed.