Various neuroscientific techniques are increasingly being used in criminal courts causing a vivid debate on the way that this kind of techniques will and should be used as scientific evidence. The role of experts in this context is important, since it is them that analyse, present, interpret and communicate the results of these techniques to the judges and the jury. In an attempt to contribute to the discussion about the role of the experts in criminal cases where neuroimaging evidence was introduced, we examined twenty seven cases from the US and Europe. Focusing on the role of experts and their presentation of neuroscientific evidence, we aimed to examine the extent to which neuroimaging data can contribute to the construction of a solid and more objective, "scientifically - based" case. We found that neurobiological information introduced through experts' testimony is generally used in order to demonstrate some physical, organic base of a psychiatric condition, or/and in order to make visible some brain lesion, (structural or functional), susceptible to have affected the capacity to reason and to control one's impulses. While neuroimaging evidence is often presented by the defence as a scientific method able to offer a precise diagnosis of the pathology in question, our case analysis shows that the very same neurobiological evidence can be interpreted in different - sometimes diametrically opposed - ways by defence and State experts. Conflicting testimony about the same empirical evidence goes against the hypothesis of neuroscientific techniques constituting "objective and hard evidence", able to reach solid, scientific and objective conclusions. Frequent conflicts between neuroimaging experts require the courts to deal with the resulting uncertainty. As the law changes with technology, it is necessary for legal professionals to train and be prepared for the new issues they may encounter in light of new developments in neuroscience, so that they become more vigilant as to the interpretation of neuroscientific data.