This aim of this thesis is to measure the effect of the user interface in real searches on the Internet through a small experiment. Theories on web and search usability and psychological factors concerning usability were first explored. Then three interfaces were constructed, one "good" that followed usability guidelines and had multiple search functions, one "bad" that completely disobeyed those guidelines and had only a simple and small search field, and one "intermediate" interface that was a mix of the two. The search field was set to different sizes, with the shortest in the "bad" interface, a medium field in the "intermediate" interface and several long fields in the "good" interface. All three interfaces were then attached to the same web search engine, Altavista. 15 students divided into 3 groups were given one interface each. They performed a list of searches and gave a judgement on the relevance of their result. Thus, precision was recorded. The time to formulate a query and the length of the queries were also recorded. At the end of the experiment a short questionnaire followed to measure the students satisfaction on a Likert scale. The results were analysed with a Mann-Whitney test and the conclusion was that the precision measure showed no significant differences between the interfaces. However, the time was significantly longer in the "good" interface compared to the other two and the length of queries was significantly longer in the "good" interface. Satisfaction was also significantly higher in the "good" user interface.