Abstract : Recently new algorithms appeared for updating the Burrows-Wheeler transform or the suffix array, when the text they index is modified. These algorithms proceed by reordering entries and the number of such reordered entries may be as high as the length of the text. However, in practice, these algorithms are faster for updating the Burrows-Wheeler transform or the suffix array than the fastest reconstruction algorithms. In this article we focus on the number of elements to be reordered for real-life texts. We show that this number is related to LCP values and that, on average, L_ave entries are reordered, where L_ave denotes the average LCP value, defined as the average length of the longest common prefix between two consecutive sorted suffixes. Since we know little about the LCP distribution for real-life texts, we conduct experiments on a corpus that consists of DNA sequences and natural language texts. The results show that apart from texts containing large repetitions, the average LCP value is close to the one expected on a random text.