ERROR is NOT an internal SAS variable used to mark errors within a dataset. The correct automatic variable is ERROR (with one underscore on each side), which is set to 1 when an error occurs during DATA step execution. However, the question asks about it being used to mark errors 'within a dataset', which is misleading - it marks DATA step execution errors, not dataset-specific errors.