“…By contrast, our duality pairs models that share the same "u" perturbations. (Note: the variable called "u" in [16] is called "v" in our paper, in agreement with Mukhanov's convention [23,24].) Thus, whereas our duality connects expanding and contracting models through the substitution ǫ → ǫ = 1/ǫ (which leaves θ ′′ /θ, and hence u, invariant), Wands' duality instead uses the substitution ǫ → ǫ = (2ǫ − 3)/(ǫ − 2) (which leaves z ′′ /z, and hence v, invariant).…”