Assay of total serum protein by the biuret method calibrated with albumin standards according to the reference method provides results with a positive bias approximately 3%-5% exceeding the total error of 3.4% allowable for total protein in serum analysis made by analysers using two-part reagents and short-term procedures. We used two types of two-part biuret reagents utilised in a short-term measurement in analysers with albumin or serum calibrators, in which protein was attested by the Kjeldahl method. Tests with potentially interfering substances proved that serum blanking used in a short-term biuret procedure is not capable of sufficiently eliminating effects of serum interferents. A short-term blanking is evidently capable of suppressing only an absorbance caused by serum-present coloured and turbid interferents, but its capacity to transform them (oxidise, hydrolyse, saponify, etc.) to some other not-interfering substances is very low compared with a long-term blanking. Lipids and bilirubin are responsible for significant positive bias of total protein in normal serum samples (approximately 3%) and even a greater positive offset in lipaemic and icteric sera (approximately 5%). We verified that interference tests based on a normal serum spiked with endogenous lipids and bilirubin give quite false and misleading results in the biuret reaction. A pure albumin, not depending on its bovine/human origin, gives absorbance responding only to its copper complexes with protein with a biuret regent, while its absorbance with a serum also includes the absorbance of interferents present in serum. The simplest way to improve current short-term biuret procedures is the use of a human serum calibrator with total protein attested by the Kjeldahl method. A serum calibrator, behaving analogously to serum samples, compensates for a positive bias in most normal sera. Reagents with a greater concentration of active biuret components (copper and alkali, reference method included) seem to be unnecessarily aggressive to proteins and are responsible for a lower accuracy when used in short-term measurements. Standard Reference Material 927c based on pure bovine albumin is still recommended and used as the primary standard for assays of total protein by colourimetric methods. The albumin calibrator is responsible for a positive bias of approximately 3%-5% in serum total protein assayed by the biuret reaction both in the reference and in current methods. Its substitution by a serum calibrator attested by the Kjeldahl method could solve this drawback. Clin Chem Lab Med 2009;47:91-101.