In an asymmetric colour matching experiment, eleven observers adjusted computer displays to colour-match surface samples in a viewing booth. We found systematic discrepancies between the observers' judgments and the predictions of the CIE 1964 Standard Colorimetric Observer. The features of the discrepancies are consistent with previous reports on adaptation in colour matching and on failures of colorimetric additivity, but have never been confirmed to be significant in practical colorimetry. We attribute the discrepancies to post-receptoral adaptation mainly of the blue-yellow chromatic channel, and report a framework of an adaptation transform based on the MacLeod-Boynton chromaticity diagram which can compensate for them without abandoning traditional colorimetry and the use of tristimulus values. (C) 2009 Wiley Periodicals, Inc. Col Res Appl, 34, 182-193, 2009; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/col.20492