We confront theoretical models for the rotational, magnetic, and thermal evolution of an ultramagnetized neutron star, or magnetar, with available data on the anomalous X-ray pulsars (AXPs). We argue that, if the AXPs are interpreted as magnetars, their clustering of spin periods between 6 and 12 s (observed at present in this class of objects), their period derivatives, their thermal X-ray luminosities, and the association of two of them with young supernova remnants can only be understood globally if the magnetic field in magnetars decays significantly on a timescale of the order of 104 yr.