This paper presents evidence from two empirical studies which show that forward slices will tend to be smaller than backward slices and discusses the importance of this result for work on source code analysis and manipulation. The difference in forward and backward slice size distributions stems from the way in which control dependence and data dependence affect slice size. The paper explains this claim in detail and supports it with empirical evidence. The empirical evidence comes from a further study which uses a restricted form of slicing that isolates the effects due to control and data dependence. This study shows that it is the inherent difference between forward and backward control dependence that leads to the difference in slice size distributions for forward and backward slicing. The results from all three empirical studies are shown to be strongly statistically significant.