Conversed with Rick on this. You have lots of choices. For false discovery rate q-values, you can use "3dFDR -qval" to generate a q-value dataset that you can threshold directly. Alternatively, "fdrval -qinput" should return the statistical threshold you want. For family wise error correction, you will have to parse the output of 3dClustSim to get the uncorrected p-value. Then use cdf to find the value of the statistic in which you are interested. This is likely to be done only at a group level analysis, so I'm not sure how much awking and greping is needed.