The reliability of five simple techniques for the estimation of blood haemoglobin levels was studied under laboratory conditions. The methods tested were the Sahli technique, the Tallqvist method, and the Lovibond comparator with undiluted blood, blood diluted in Drabkin solution, and blood diluted in modified Drabkin solution. The Lovibond comparator technique using undiluted blood was satisfactory in terms of precision and accuracy, and seems suitable for the monitoring of individual patients and for screening for anaemia. The other techniques were less reliable.A detailed procedure was worked out for estimation of the number of screening errors that can be expected with each of these techniques in any epidemiological situation.