The study of neutrino oscillation physics is a major research goal of the worldwide particle physics program over the upcoming decade. Many new experiments are being built to study the properties of neutrinos and to answer questions about the phenomenon of neutrino oscillation. These experiments need precise theoretical cross sections in order to access fundamental neutrino properties. Neutrino oscillation experiments often use large atomic nuclei as scattering targets, which are challenging for theorists to model. Nuclear models rely on free-nucleon amplitudes as inputs. These amplitudes are constrained by scattering experiments with large nuclear targets that rely on the very same nuclear models. The work in this dissertation is the first step of a new initiative to isolate and compute elementary amplitudes with theoretical calculations to support the neutrino oscillation experimental program.,Here, the effort focuses on computing the axial form factor, which is the largest contributor of systematic error in the primary signal measurement process for neutrino oscillation studies, quasielastic scattering. Two approaches are taken. First, neutrino scattering data on a deuterium target are reanalyzed with a model-independent parametrization of the axial form factor to quantify the present uncertainty in the free-nucleon amplitudes. The uncertainties on the free-nucleon cross section are found to be underestimated by about an order of magnitude compared to the ubiquitous dipole model parametrization.,The second approach uses lattice QCD to perform a first-principles computation of the nucleon axial form factor. The Highly Improved Staggered Quark (HISQ) action is employed for both valence and sea quarks. The results presented in this dissertation are computed at physical pion mass for one lattice spacing. This work presents a computation of the axial form factor at zero momentum transfer, and forms the basis for a computation of the axial form factor momentum dependence with an extrapolation to the continuum limit and a full systematic error budget.