Methods and systems of semiconductor testing where reference dice and non-reference dice in a wafer and/or lot are tested differently. In one embodiment of the invention, geography, lithography exposure, other characteristics, performance and/or behavior are taken into account when selecting reference dice, thereby improving the likelihood that the response of reference dice to testing is well representative of the wafer and/or lot. In one embodiment, based on data from the testing of reference dice, the test flow for non-reference dice and/or other testing may or may not be adjusted.