The measurement of frequency shifts for light beams exchanged between two test masses nearly in free fall is at the heart of gravitational wave detection. It is envisaged that the derivative of the frequency shift is in fact limited by differential forces acting on those test masses. We calculate the derivative of the frequency shift with a fully covariant, gauge-independent and coordinate-free method. This method is general and does not require a congruence of nearby beams' null geodesics as done in previous work. We show that the derivative of the parallel transport is the only means by which gravitational effects shows up in the frequency shift. This contribution is given as an integral of the Riemann tensor -the only physical observable of curvature-along the beam's geodesic. The remaining contributions are: the difference of velocities, the difference of non-gravitational forces, and finally fictitious forces, either locally at the test masses or non-locally integrated along the beam's geodesic. As an application relevant to gravitational wave detection, we work out the frequency shift in the local Lorentz frame of nearby geodesics.Introduction-The exchange of light beams between (almost) free falling test masses and the measurement of the corresponding frequency shifts (see Fig. 1) is at the heart of any thought (real) experiment devised for measuring in principle (in practice) space-time curvature. It is in fact the key for the direct observation of gravitational waves (GW) by interferometer detectors on the ground [1, 2] and in space [3], and pulsar timing arrays [4]. Particularly at low frequency, GW detectors are limited by differential forces acting on the test masses [5] and, as such, the derivative of the frequency shift may be a good observable of space-time curvature and, in general, a means to separate true gravitational forces from spurious effects. Recently, two different approaches [6,7] have put forward a formalism that allows the frequency shift to be computed in terms of an integrated measure of curvature. This is distinct from earlier attempts that often relied on simplifying assumptions, e.g. metric expansion, geodesic deviation, choice of a preferred coordinate system, or fixing an a priori gauge (see the introduction of Ref.[6] and references therein).The only physical covariant quantity that unambiguously describes the effect of curvature in vacuum is the Riemann tensor [8]. In effect, other general relativistic variables, such as the Ricci tensor and the Ricci scalar are identically zero in vacuum, even in a curved spacetime. Additionally, the Christoffel symbols can be set to zero by a proper change of reference frame and the metric itself is in general gauge dependent [9]. Consequently, all those quantities are not good observables of the true gravitational effect and, as such, they might eventually lead to ambiguous results. Previous work has already pointed out, although in different formulations, that the frequency shift is sensitive to an integrated measure of the Riemann tensor...