The meta-analysis of homeopathy trials that appeared in the Lancet in 1997 seemed to endorse the experience of practitioners and patients that homeopathic medicines have specific clinically relevant effects. However, results from later unsuccessful trials, and negative inferences from a review of trials for a condition excluded from the meta-analysis--delayed onset muscle soreness (DOMS)--have since been presented to suggest that the meta-analysis may well have overestimated the positive effects of homeopathy, and that the "placebo question is still not resolved." This article reviews the evidence underlying this challenge to the meta-analysis and homeopathy, and demonstrates that it would be valid if it were based on: a comprehensive literature search; appropriate classification of primary studies; clear discrimination between clinical effectiveness and placebo questions; sound and transparent review methods; and a reliable and unconfounded clinical treatment model for testing the ultramolecular hypothesis. It is suggested that different models are needed to answer different questions.