A comparison of two sequences may uncover multiple regions of local similarity. While the significance of each local alignment may be evaluated independently, sometimes a combined assessment is appropriate. This paper discusses a variety of statistical and algorithmic issues that such an assessment presents.