Turbulence modeling has traditionally relied heavily on the eddy viscosity concept relating turbulent fluxes to local gradients. Important deficiencies of the eddy viscosity concept include its inability to capture non-local and memory effects of turbulence. This has motivated recent interest in exploring how fractional derivatives (derivatives with non-integer order) may be helpful in generalizing the eddy viscosity concept to include non-local effects. Building on the speculative work of Chen (Chaos 16, 023126, 2006), a recent study by Epps & Cushman-Roisin has provided a systematic derivation for a fractional Laplacian model of the Reynolds stress. Meanwhile, unpublished work by Song & Karniadakis applies numerical optimization to determine the fractional order for a variable fractional RANS model in wall-bounded turbulence using DNS results, finding universal results across a range of Reynolds numbers. This talk will evaluate the general claims made in support of fractional turbulence modeling, the assumptions in the Epps & Cushman-Roisin derivation, as well as the relative success of results in these works, with an eye toward clarifying the promise and difficulties of such an approach to turbulence modeling.