This study investigates the efficiency of the Google search engine at retrieving items from 26 UK Institutional Repositories, covering a wide range of subject areas. One item is chosen from each repository and four searches are carried out: two keyword searches and two full title searches, each using both Google and then Google Scholar. A further search tests the retrieval of the item from the repository interface. Known information-seeking behaviour was taken into account when interpreting the results. These show that the Google and Google Scholar user would retrieve items from the repositories, particularly when the full title was known. However, some skill would be needed to evaluate the different versions of texts retrieved, and accessing the repositories through more sophisticated harvesting services might prove a better option for the scholar. (Contains 3 tables.)