(1) The standard treatment for mild to moderately severe Alzheimer's disease is donepezil, an anticholinesterase with some beneficial effects (progression is slowed in about 10% of patients, by six months on average) and mainly gastrointestinal adverse effects. (2) Memantine, a drug first developed several decades ago, belongs to the family of NMDA glutamate receptor inhibitors. Marketing authorisation was recently granted for memantine in moderately severe and severe Alzheimer's disease. (3) The clinical evaluation dossier on memantine is poor. Marketing approval was obtained thanks to only one placebo-controlled trial. It included only 252 patients treated for 28 weeks. Patients with moderately severe Alzheimer's disease were not analyzed separately from those with severe forms, even though the response criteria are different for the two categories of patients. (4) According to the chosen endpoint, 5% to 19% of patients were clinically improved by memantine. It is not known whether this benefit persists beyond six months. (5) The report of a trial comparing memantine + donepezil with placebo + donepezil does not analyse the response rate. Use of this combination is not currently justified. (6) In clinical trials the main adverse effects of memantine were neurological (dizziness and headache). Fairly lengthy pharmacovigilance data from Germany are relatively reassuring. (7) In practice, donepezil remains the reference option for moderately severe Alzheimer's disease. Memantine is a second-line option, as its adverse effects differ from those of anticholinesterases. There is still no drug offering a clear benefit for patients with severe forms of the disease.