Ten plants per group, dried weights. The t-test compares the statistic t with the t distribution, which is exact if the data are independent normal draws with a common variance. The permutation test compares the same statistic with its values over all ways of relabelling the plants, which is exact if the labels are exchangeable under the null. Change the comparison, the statistic, the side, and see what happens when every measurement is entered twice.
Exact enumeration uses every split of the pooled sample into two groups of the original sizes (184,756 for 10 + 10). The Monte Carlo p-value is (1 + b)/(B + 1), where b counts random relabellings at least as extreme as the data; it is a valid p-value for every B.
code/ch06_testing.R (functions t_stat and perm_t_all, Figures 6.8 to 6.10). This page reproduces the values (ctrl versus trt2: t = −2.134020, p = 0.046851, exact permutation p = 0.048334).