In this paper, we evaluate how the performance of a wearable context recognition system is affected by the sampling frequency and the resolution of the sensor signals used for the classification. We introduce our method for this evaluation and present the results for a widely studied activity recognition task: the classification of human modes of locomotion using body-worn acceleration sensors. With this example we show that both the sampling frequency and the resolution can be significantly reduced without much impact on the recognition performance. While many of the published approaches in this domain rely on higher sampling frequencies and signal resolutions, we show that good recognition performance can already be achieved with 20Hz and 2 bit resolution.