As of early 2016, 550,000 patent applications submitted to the USPTO are unexamined. The considerable number of unexamined applications is likely to reduce the amount of time that patent examiners can spend on the examination of an application. Fast-paced technological progress, on the other hand, implies that patent examiners need to invest more time in patent examination. In light of these facts and with sometimes millions at stake the importance of conducting patent searches efficiently is obvious. This thesis analyzes various characteristics of more than 15,000,000 Boolean search queries submitted by professional patent examiners to the EAST patent search engine at the USPTO. Search queries generated during the examination of a patent application are available from the USPTO as SRNT ("search related notes") documents. In order to assemble a rather large dataset of search query logs, more than one million patent applications had to be retrieved and processed. The set of obtained SRNT documents was split into two groups. For one group (SRN T 892 ) relevant patents had been cited by the patent examiner. For the other group (SRN T no892) ), no relevant patent documents had been found during the search. Query Expansion (QE) is a popular and well-studied technique for improving search results by adding relevant terms to an user query. Professional patent searchers apply QE manually. Either by providing lists of related terms within a Boolean query, or by using the truncation operator as instruction for the patent search engine to consider variations of a word. Contrary to my expectation it is shown that there is no difference between "successful" and "unsuccessful" searches in terms of quantity by which lists of alternate terms are provided. However, it is also shown that the use of the truncation operator is more popular in "successful" searches. Most of the examined search features, such as the average query length, the use of parentheses, the use of Boolean operators, yield relatively similar results for both document sets. Noteworthy differences have been found in the average document length (= number of queries per search), in the use of patent database specific search fields (e.g. to search in the "claims" section of a patent) and in the use of references (to address former queries).