The distraction detection apparatus includes: an electroencephalogram detection section for detecting an electroencephalogram signal of a driver; an arousal level estimation section for retaining a first rule of mapping parameter values of an electroencephalogram signal to arousal levels, and estimating an arousal level based on the detected electroencephalogram signal and the first rule; an attention allocation estimation section for retaining a second rule of mapping parameter values of an electroencephalogram signal to attention allocations, and estimating an attention allocation based on the detected electroencephalogram signal and the second rule; a driver state estimation section for retaining a third rule of deriving an amount of attention from an arousal level and an attention allocation, estimating an amount of attention of the driver paid to driving based on the estimated arousal level and attention allocation and the third rule, and classifying a state of the driver into a normal state, a state of reduced attention, or a state of reduced arousal level; and an output section for performing an intervention for the driver based on a result of classification by the driver state classification section.