In a paired sample t-test, if the differences $d_i = x_{1,i} - x_{2,i}$ are calculated for $n=15$ pairs, what is the null hypothesis $H_0$ being tested regarding the mean difference $\mu_d$?