Churchill and Doerge (1994) describe a method to calculate the threshold values for quantitative trait mapping that we have implemented in this program. Basically, it does a permutation of the trait values and the genotypes and redoes the analysis. Over the number of replicates, two types of thresholds are defined: "experimentwise" and "comparisonwise". We calculate the experimentwise thresholds, but only give p values for the comparisonwise values to save on storage space. The p values give the proportion of permuted replicates that have loglikelihood ratios larger than the observed ratios.
If you choose to do permutation tests, you need to run Zmapqtl with the model of choice prior to doing the permutation test. Also, if the program terminates prematurely, you can restart it from where it left off to complete the permutation test.