Measuring cytokine responses to infection has proven to be invaluable for the understanding of immunity to tuberculosis in the laboratory. However, far less data are available from studies in humans and these have often produced conflicting results. Here we describe a comprehensive multi-center comparison of the most commonly used protocols for cytokine analysis: ELISA, ELISPOT and RT-PCR, in cohorts of TB patients, their household contacts and community controls. In particular, we have studied the effect on these protocols of conditions that commonly prevail in field studies, such as delays between sample collection and analysis, or different source material, such as whole blood or frozen PBMC. The results clearly show that while there is good correlation between the methods under optimal conditions, each method has strengths and weaknesses that render them more or less suitable for particular types of analyses. Researchers should carefully consider these factors when planning human field studies.