DSM-III-R, DSM-IV and ICD-10 definitions of alcohol dependence were all developed from the concept of the Alcohol Dependence Syndrome, and thus have a common theoretical link. This link is not shared by DSM-III, and no link exists between definitions of abuse in the different classification systems. The level of agreement on diagnoses produced by the different systems has practical as well as theoretical implications. We tested this agreement in 962 US household residents randomly sampled and screened for heavier than average drinking in the last 12 months. Agreement for most comparisons involving diagnoses of current dependence ranged from good to excellent, with no clear pattern of lower agreement for DSM-III. However, agreement on past dependence was sharply lower for comparisons involving DSM-III than those involving the other classification systems. This appeared to be due to the DSM-III requirement for physiological dependence and the apparently emerging nature of the disorder in this relatively young, non-treatment sample. Comparisons for abuse were generally poor for current as well as past diagnoses. Implications of the findings are discussed.