A mobile device that adapts its behaviour to complement the user experience has long been a goal for the pervasive and ubiquitous research communities. Known as context-awareness, this enables behaviours such as diverting incoming phone calls to an answer phone or hands-free-kit if the carrier of the phone is currently driving. Another example is the automatic filtering of content to show only relevant data, e.g. the locations of the closest restaurants. Traditionally, positional information has been determined via the use of GPS receivers; everyday activities such as walking and driving have been recognised using machine learning techniques to classify patterns of accelerometer data. Both of these sensors require additional hardware and in terms of power consumption, are computationally expensive to run. In this thesis we demonstrate how a similar level of context-awareness can be achieved without the use of quantitative positioning techniques involving a GPS receiver and without the user of an accelerometer to recognise everyday activities such as walking, travelling in a car and remaining stationary. We show how patterns of signal strength fluctuation can be classified as occurring whilst undertaking activities such as walking and driving, and show how this behaviour enables accelerometer free activity recognition. A qualitative approach is presented for modelling the spatial environment that shields the user from inconsistencies in positioning system performance. We demonstrate how position and activity data can be used to improve the performance of both the activity sensing and positioning services. In conclusion this thesis argues that for many applications this level of context-awareness is sufficient.