Measurements of the activity of transketolase in human erythrocyte lysates by an assay coupled to NADH oxidation indicate that interactions of assay substrates with hemoglobin can give rise to overestimations of transketolase activity. Three potential sources of error are identified. Thus, in lysates containing methemoglobin, NADH oxidation can be due firstly to methemoglobin reductase activity or secondly to the monooxygenase activity of methemoglobin, for which the substrate can be ribose 5-phosphate, a substrate also of transketolase. Thirdly, the addition of high concentrations of the transketolase cofactor, TDP, to an insufficiently buffered reaction mixture can cause the aggregation and precipitation of hemoglobin: a phenomenon that may be misconstrued as an enhanced increase in absorbance at 340 nm and hence as additional transketolase activity. Although the present study concentrates on these potential artefacts in assays of transketolase activity, the findings may well be relevant to the measurement of other enzyme activities in hemolysates by procedures based ultimately on the rate of consumption or production of NAD(P)H.