A probabilistic method has been devised to assess the geologic realism of subsurface well-to-well correlations that entail the lateral tracing of geologic bodies across well arrays with constant spacing. Models of geo-body correlability (based on the ratio between correlatable and penetrated geo-bodies) are obtained from total probabilities of penetration and correlation, which are themselves dependent on the distribution of lateral extent of the geo-body type. Employing outcrop-analog data to constrain the width distribution of the geo-bodies, it is possible to generate a model that describes realistic well-to-well correlation patterns for given types of depositional systems. This type of correlability model can be applied for checking the quality of correlation-based subsurface interpretations by assessing their geologic realism as compared with one or more suitable outcrop analogs. The approach is illustrated by generating total-probability curves that refer to fluvial channel complexes and that are categorized on the basis of outcrop-analog classifications (e.g., braided system, system with 20% net-to-gross), employing information from a large fluvial geo-body database, Fluvial Architecture Knowledge Transfer System (FAKTS), which stores information relating to fluvial architecture. From these total-probability functions, values can be drawn to adapt the correlability models to any well-array spacing. The method has been specifically applied to rank three published alternative interpretations of a stratigraphic interval of the Travis Peak Fonuation (Texas), previously interpreted as a braided fluvial depositional system, in terms of realism of correlation patterns as compared to (1) all analogs recorded in FAKTS and considered suitable for large-scale architectural characterization, and (2) a subset of them including only systems interprethd as braided.