Functional magnetic resonance imaging has been used extensively to study the underlying mechanisms of acupuncture. More than 80 studies and a dozen review articles have been published on this topic in the past 10 years. Emphasising the statistical quality of the studies, this critical review evaluates the results obtained so far by studies that have used a hypothesis-driven approach. A minimum sample size of 12 subjects, the application of random effects (or mixed effects) analyses, and a threshold corrected for multiple comparisons were defined as inclusion criteria for a meta-analysis of the reported cortical activations. Fourteen out of 71 studies met the inclusion criteria. The majority of these studies reported activations in cortical areas relevant to the processing of somatosensory, motor, or pain signals, as well as areas related to the special senses. From this type of study, it is so far impossible to say if cortical activations under acupuncture are part of an underlying mechanism, or if they simply reflect the brain's processing of the somatosensory or pain stimulus from the acupuncture needle stimulation. In the future, the application of data-driven methods may lead to a more comprehensive understanding of neuronally mediated acupuncture effects.