Psychological testing requires an instrument that is both 'valid' and 'reliable.' A valid instrument actually measures what it intends to measure, and a reliable instrument measures consistently. Sentencing as an instrument that measures criminal behavior has often been more concerned with reliability, i.e., consistency, than with validity, i.e., whether it actually does measure criminal behavior. Construct validation focuses on whether a theory proves true in practice. Valid sentencing under the retributive theory is that which successfully matches sentencing severity to offender culpability. This is not likely to occur in the use of descriptive sentencing guidelines. Sentencing guidelines calibrate raw judicial agreement about relative offender culpability in the same way that standardization calibrates raw test responses, so that the normal responses in relation to particular target groups can be known and compared. Sentencing guidelines are more valid under retributive theory than sentencing derived from appellate case law, the current system in England, since appellate case law does not attempt to reflect consensus judicial assessments of culpability. 3 notes and 9 references.