Suppose we have a model
$Y = \beta_1 X_1 + \beta_2 X_1X_2 + \epsilon$
Hence $X_1$ has an effect on $Y$ regardless of the value of $X_2$, but $X_2$'s effect on $Y$ is mediated or enabled by $X_1$.
Please fill in the blank: $X_1$ is a ******** variable for the relationship between $X_2$ and $Y$?