in this paper we present and empirically evaluate a user study using a web search log and eye tracking to measure user behaviour during a query session, that is, a sequence of user queries, results page views and content page views, in order to find a specific piece of information. we evaluate different tasks, in terms of those who found the correct information, and in terms of the query session sequence itself, ordered by serp (search engine result page), number and return visits to the results page for the same query. from this we are able to identify a number of different behaviour patterns for successful and unsuccessful users, and different trends in user activity during the query session. we find that a user behaves differently after the first query formulation, when we compare with the second formulation (both queries being for the same information item). the results can be used to improve the user experience in the query session, by recognising when the user is displaying one of the patterns we have found to have a low success rate, and offering contextual help at that point. the results may also contribute to improving the design of the results page.