We show that Mutual Information between word pairs can be successfully used to discriminate between word senses in the query translation step of Cross Language Information Retrieval. The experiment is conducted in the context of Amharic to French Cross Language Information Retrieval. We have performed a number of retrieval experiments in which we compare the performance of the sense discriminated and non-discriminated set of query terms against a ranked document collection. The results show an increased performance for the discriminated queries compared to the alternative approach, which uses the fully expanded set of terms.