It is noted that research on the psychological effects of disaster, particularly with regard to rates of impairment, has turned up confusing and sometimes contradictory results. Two sets of dimensions salient to such investigations are noted: those which are aspects of disasters per se and affect actual rates of impairment ("true scores") and those which could be expected to affect estimates of impairment rates ("error variance") following disaster. Dimensions of disasters per se suggested by others are reviewed, and an additional dimension is proposed. Four methodological dimensions affecting reported impairment rates are described (sampling of subjects, level of data, case identification, and time of follow-up). Studies of long term psychological effects of disaster where some estimate of impairment was given are reviewed in order to demonstrate the noncomparability of findings from study to study due to methodological differences.