APL still deals poorly with some aspects of real world problems. For instance, frequently to mark entries in an otherwise numeric database as “not applicable” or “not entered,” one picks arbitrarily one or more numbers to represent the attributes. But this and other approaches suffer difficulties ranging from ambiguity to inefficiencies in both programming and execution. Further, APL errors which might logically be associated with individual elements of data arrays are not reported by existing APL systems in any detail. This paper proposes an enhancement to APL which would overcome both these difficulties, discussing considerations of both use and implementation.