I've been using a tool called Stuart Tactile maps - a table top test that has a high correlation with people's propensity to get lost or remain orientated during functional orientation and mobility. Everyone does the Stuart Tactile Maps test under blindfold, so it takes vision out of the equation. See http://lildeverell.net/roam/stm/ for more information.
Observing real navigation may be problematic. Normally VR is used for measuring navigation in computer generated situations. Depthmap is a tool for simulating navigational behavior and has been proven to have high correlation with normal movements. For measuring behavioural factors structured questionnaire may be used.
Spatial memory assessment in humans is a very interesting since spatial memory methods used to be very sensitive to brain alterations. But (there’s always a “but”) although spatial orientation has been studied in the last 40 years in humans, only in the last 17 years the number of studies increased. The explanation: introduction of virtual reality based tasks. As pointed by Dr. Zhong, virtual reality based task are very effective. But I’m afraid if you use old people with very low educational level, the use of such tests could provide no reliable results.
I attached a couple of links that could help you to understand how we proceed.
Virtual reality contributes greatly to the understanding of the neural substrates involved, but it seems currently limited when I think of the clinical utility of Spatial Navigation in the elderly (especially in developing countries).
Numerous methods have been employed in the evaluation of this measure in humans, all indisputably important.
But conducting clinical research with different methods to evaluate the Spatial Navigation, could lead us to problems of comparability among the findings..
If you assess spatial navigation you have to take into account the kind of strategy used by participants. So be sure that methods used demanded the same kind of strategies and you'll be sure that they're measuring the same "thing". I.e. we applied two different virtual reality based task to the same sample. Both tasks showed the same results, but very different from classical neuropsychological test. Thus, spatial navigations tasks revealed memory impairments whereas the other tasks did not disclose any difference between groups.