Using a simple Gaussian beam solution to the one-way scalar wave equation, we derive analytical expressions for the evolution of phase and group delay after a wave passes through a Gaussian-shaped heterogeneity of half width L. As a function of distance x, there are two clearly separated regimes, depending upon the wavelength λ of the wave. In regime I, when x/L≪&pgr;L/λ, the absolute magnitude of the phase delay decreases approximately linearly with x, and the anomaly does not widen appreciably except by developing small sidelobes with delays of opposite sign. Tomographic inversions of such delays will be damped but are theoretically well posed. In regime II, when x/L≫&pgr;L/λ, the absolute delay decreases toward zero as 1/x, most markedly on the ray itself, and the cross-path shape of the wave front bears little resemblance to the original anomaly. Tomographic inversions of delay times in this regime are ill posed. Group delay times show a similar behavior in the two regimes. Although their rate of decrease with distance is slower in regime I, they develop more disturbing sidelobe behavior off the central ray. The effects of wave front healing for surface waves traveling in two dimensions are less severe than those for body waves in three dimensions; as a result, surface wave inversions will commonly be in regime I. Short-period body wave group delays are also in regime I; nevertheless, the damping of delays in this regime is likely to contribute significantly to the scatter of observed travel time anomalies. Tomographic inversions of long-period body waves, which fall at the limit of regime I, or even in regime II, face perceptible limitations in theoretical resolving power. Finally, we show that there is an asymmetry in the evolution of positive versus negative travel time anomalies. ¿ 2000 American Geophysical Un