======================================================================
Dear Yehouda Harpaz
Thank you very for your draft. I found it interesting reading but I must tell you right from the start that I think that you are making a major mistake. I think that you have confused greatly replicability in studies where exactly the same experiment has been performed in cognitive terms and replicability where there are similar processes postulated but the tasks are different. In such cases it is perfectly reasonable for there to be a lack of replicability because our notion of processes is not mechanistic but based on metaphorical representation and unperfect models of behavioural data. It is therefore perfectly possible for two so called processes to actually vary and be composed of different sub-components. The variability introduced by biology is nowhere properly discussed by you. You also completely fail to discuss the statistical framework in which imaging data are analyzed. The analytical equations developed by statisticians for assessing the significance of change in the simple experiments that you concentrate upon (i.e. those which contrast two different tasks) are now mathematically well established. If the analytical framework is common and the experiment is common then the results converge. There is a trivial lack of convergence due to machines which have different sensitivities or which have different capacities to collect data from large or small areas o brains. It's like comparing micrometeres to rulers, as the Poline paper clearly shows. Finally, there is the issue of whether the reported results refer to subjects, groups or groups reperesentative of the population at large. The threshold necessary for each level of inference are very differet - another statistical issue.
In summary, therefore, I think that the conceptual background to your paper is fatally flawed. I do hope that you take this as constructive criticism. I do not wish to appear negative.
Youe sincerely
Professor Richard Frackowiak