The method of 'linearization via the wave operator' establishes a close connection between scattering theory and complete integrability for nonlinear wave equations in four-dimensional Minkowski space. We review the proofs of complete integrability for the massive φ 4 theory on the space of solutions of finite energy, and for the massless φ 4 theory on the space of solutions all of whose images under conformal transformations have finite energy. We show that the complete integrability is associated to the infinite-dimensional symmetry group C(S 3 , U (1)) in the massive case, and a subgroup of C 1 (S 3 , U (1)) in the massless case. We review the construction of gauge-invariant conserved quantities for suitably regular solutions of the Yang-Mills equations in terms of asymptotic 'in' or 'out' fields, and discuss the prospects for complete integrability in this case. We wish to emphasize that by 'completely integrable' we do not mean 'exactly solvable,' which is a rather vague notion in any event. Instead, we mean the existence of a complete set of conserved quantities with vanishing Poisson brackets, arising from the action of a commutative Lie group of symmetries. Strictly speaking, this is a property not of an equation, but of a one-parameter group of canonical transformations of phase space. There are various ways of formulating this property precisely, especially when the phase space is infinite-dimensional, so we begin with some definitions. In all that follows we assume that the phase space M is a (smooth) Banach manifold. We say that a function from M to another Banach manifold is 'smooth' if it is infinitely Frechét-differentiable. We say that (M, ω) is a 'symplectic manifold' if ω is a smooth closed 2-form on M that is weakly nondegenerate, that is, for any nonzero u ∈ T x M there exists v ∈ T x M with ω(u, v) = 0. Given a symplectic manifold (M, ω), we say that a subspace V ⊂ T x M is 'isotropic' if ω|V = 0, and 'Lagrangian' if it is a maximal isotropic subspace. We say that a diffeomorphism f : M → M of a symplectic manifold is a 'symplectomorphism' (in the language of physics, a canonical transformation) if f * ω = ω. Let (M, ω) be a symplectic manifold. We say a function f ∈ C ∞ (M) is 'nice' if it generates a smooth vector field ζ f on M in the following sense: df = ω(ζ f , •). Since the map v → ω(v, •) need not define an isomorphism of T x M and T * x M , not every smooth function need be nice, but ζ f is unique if it exists. Given nice functions f 1 , f 2 generating vector fields ζ 1 , ζ 2 , the function {f 1 , f 2 } = −ω(ζ 1 , ζ 2) generates the vector field [v 1 , v 2 ]. In other words, the nice functions form a Lie algebra under the Poisson bracket {•, •}, and the map f → ζ f is a Lie algebra homomorphism. Let G be a Banach Lie group and ρ: G × M → M a smooth action of G on M as symplectomorphisms. The action ρ defines a homomorphism dρ from the Lie algebra g of G to the Lie algebra of smooth vector fields on M. We say that the action ρ is 'Hamiltonian' if for each v ∈ g there is a nice functio...