A reducible expression (redex for short) is an expression which matches the left-hand-side of some reduction rule or definition.
For example, given the definitions:
fac 0 = 1 fac n = n * fac (n-1)
Operationally, a redex is any expression whose evaluation requires work to be done. For example, a function call with all of its arguments supplied is a redex, but a constant is not. This is useful, for example, in common subexpression elimination, which only saves work if the common subexpression is a redex.