This work develops a methodology for converting data from the Global Positioning System (GPS) into observed routes (routes actually taken) to characterize intraindividual and interindividual variability in route choice and to compare observed and minimum-cost routes. Exploration of observed route choice behavior is crucial because the underlying decision-making process is more complex and dynamic for route choice than for other travel choice dimensions. Furthermore, the difficulties associated with collection of data on route choice are reflected in the scarcity of studies on observed behavior and the major simplifications made in traffic assignment models developed for the most common commercial software. The present study analyzes a GPS-based database of 679 routes, collected by a personal probe system called the activity locator over a period of 2 weeks for a sample of 12 students from the University of Cagliari in Italy. In particular, variability in the daily route (for the same individual and for several individuals) and observed deviations of the route from minimum-cost routes have been examined in depth. The results indicate that higher levels of intraindividual variability are found for discretionary trips, whereas higher levels of interindividual variability, as well as greater deviation from minimum-cost routes, are associated with work or study trips.