A long-standing theory for Alzheimer's disease (AD) has been that deterioration of synapses and depressed neuronal activity is a major contributing factor. We review the increasing evidence, in humans and in mouse models, that show that there is often neuronal hyperactivity at early stages rather than decreased activity. We discuss studies in mouse models showing that hyperexcitability can occur long before plaque deposition and memory impairment. In mouse models, a generator of the hyperactivity appears to be the dentate gyrus. We present evidence, based on mouse models, that inhibition of muscarinic cholinergic receptors or medial septal cholinergic neurons can prevent hyperactivity. Therefore, we hypothesize the novel idea that cholinergic neurons are overly active early in the disease, not depressed. In particular we suggest the medial septal cholinergic neurons are overly active and contribute to hyperexcitability. We further hypothesize that the high activity of cholinergic neurons at early ages ultimately leads to their decline in function later in the disease. We review the effects of a prenatal diet that increases choline, the precursor to acetylcholine and modulator of many other functions. In mouse models of AD, maternal choline supplementation (MCS) reduces medial septal cholinergic pathology, amyloid accumulation and hyperexcitability, especially in the dentate gyrus, and improves cognition.