Current INDOT policy requires that culvert-like structures with spans greater than 20 ft be treated for purposes of hydraulic analysis as a bridge, and hence mandates the use of software such as HEC-RAS for predicting the headwater, rather than the culvert-specific software, HY-8. In this context, culvert-like structures are assumed to have a standard inlet geometry (e.g., such as those already modeled in HY-8) and a constant barrel geometry. The present study examines the technical basis of this policy, and whether the policy could be revised to allow the application of simpler culvert-hydraulics analysis and HY-8 to culvert-like structures with spans greater than 20 ft. Laboratory experiments were performed with model box culverts of span 1.5 ft and two streamwise lengths, 2.1 ft and 8 ft, and performance curves describing the variation of headwater with discharge were obtained. The effects of bed roughness, the presence or absence of a cover (if present, the rise was 0.5 ft), and a range of tailwater levels, were investigated. The laboratory observa­tions were compared with predictions by HY-8 and HEC-RAS models, and the model performance assessed. In general, HY-8 predictions were found to be as good as, and in some cases superior to, the HEC-RAS predictions, for both long and short culvert-like structures. This was attributed to the empirical information in HY-8 being more tailored to the specific standardized geometry of culvert-like structures, and the automatic inclusion of roughness effects, whereas HEC-RAS, at least when used with default coefficients and settings, relied on generic coefficients and neglected roughness effects. It was therefore recommended that a change in INDOT policy allowing large-span culvert-like structures to be analyzed using conventional culvert hydraulics would be technically justified for problems where the structure could be considered in isolation and accurate input data are available.