An improved method for the semi-quantification of stress protein biomarkers by western blotting is described. The construction of a calibration curve on each gel (using a standard) to calculate relative stress protein levels improved reproducibility of results both within and among blots. Variability in relative stress protein levels of replicate samples run on separate gels was reduced from ca. 24% to 13% when related to a calibration curve rather than being based on direct comparison of absorbance units. Within blot variability was reduced from 5.2-11.7% to 3.4-6.8% by calibrating stress-70 levels. The potential significance of this methodological development is discussed with regard to the literature.