The accuracy of location methods based on received signal strength (RSS) measurements is highly variable, but the underlying causes are not well known. In this paper, narrowband measurements at five VHF frequencies are used to evaluate the accuracy of RSS-based location algorithms. Least-squares algorithm is the most accurate if the propagation model is unbiased. Location error increases linearly with mean distance from transmitters and with RSS variation caused by shadow fading or varying antenna gain. Since shadow fading increases with frequency, location accuracy decreases with frequency. A satisfactory approximation of location accuracy in a given network is obtained by the error estimates that are based on linearization, but modified when the receiver is close to one transmitter. Kalman filtering performed on the estimated coordinates removes the largest location errors, but does not improve the median accuracy significantly due to correlated shadowing process.