Eickhoff et al (2016) reported an implementation error in GingerALE meta‐analysis software older than version V2.3.6. This error affected the code for sorting P‐values as a step towards false discovery rate (FDR) thresholding, resulting in a more lenient threshold by failing to adequately correct cluster‐level P‐values. The effects of this sorting error depend on the specific distributions within a given data set, potentially resulting in non‐uniform errors in cluster size, as well as over‐estimation of the significance of smaller clusters.
The analysis reported in Morrison (2016), which had been run using V2.3.2, was re‐run with V2.3.6 in order to confirm that the implementation error had not appreciably affected the results. All clusters reported in Table 2 were still present in the new analysis, with the sizes of five of the 24 clusters (20.8%) smaller than originally reported (in cubic mm). These five clusters ranged from 1% to 50% smaller than their originally reported size, with a mean of −19%. The most affected cluster (50% smaller) was in the precentral sulcus for the “discriminative touch” map. The effect of the implementation error was therefore negligible with respect to the overall robustness of the results reported in Morrison 2016.