Several studies in the literature have indicated that interchanges are the most crash-prone areas within the motorway system in number and severity of accidents. The reason is the high level of turbulence as a result of vehicle lane changes and speed variability. To understand the safety consequences of an interchange design (e.g., type of connecting ramps, radii and superelevation of curves, and lane and shoulder widths), an in-depth investigation of driving speed behavior is needed. Such an investigation requires the collection of detailed trajectory data on vehicles on different interchanges. These types of data are rarely available, and as a result, such studies are scarce in the literature. The main objective of this present study was to analyze driver speed behavior on different ramps at interchanges, and to develop an operating speed prediction model as a function of the road design elements. Trajectory data on free-moving vehicles were derived from stabilized video images taken from a camera mounted underneath a helicopter, which hovered over the road areas studied. Data were collected from 29 curves at six freeway-freeway interchanges in the Netherlands. The sample included nine direct connections, 12 semidirect connections, and eight indirect connections. The findings showed that speeds were affected by several road geometric characteristics of the curves, by driver expectancy and design consistency, and by the percentage of trucks in traffic. The operating speed prediction models developed in the study will provide designers with tools to estimate the operating speed during the design process.