We present an empirical evaluation of three context-sensitive slicing algorithms and five context-sensitive chop-ping algorithms, and compare them to context-insensitive methods. Besides the algorithms by Reps et al. and Agrawal we investigate six new algorithms based on variations of k-limited call strings and approximative chopping based on summary information. It turns out that chopping based on summary information may have a prohibitive complexity, and that approximate algorithms are almost as precise and much faster.