This paper examines the sensitivities of interwell tracer and transient-pressure response to spatial distribution of permeability heterogeneity. On the basis of these sensitivities, we describe a formalism to quantify the resolution and spatial averaging associated with estimates of permeabilities derived through inversion of tracer and/or pressure data. The resolution is a measure of the effectiveness of the data in estimating local-scale (gridblock) permeabilities. The spatial averaging kernels quantify the inherent averaging associated with our estimates because of limited data or sampling. By examining the resolution and averaging kernels as a function of various data types, we can evaluate quantitatively the relative importance of tracer vs. pressure data for heterogeneity characterization and the improvement in estimates obtained by combining the data types. We illustrate the concepts by application to transient-pressure and tracer response from five-spot patterns and also to an experimental tracer response from a well-characterized slab of Antolini sandstone. Tracer data is found to yield much better resolution compared to transient-pressure response. Also, both transient-pressure and tracer data appear to resolve barriers to flow better rather than channels to flow.