A short article describes analysis of 20,000 searches of the federated system at the University of Tromsø, Norway. They found that "60% of all the searches ended up with no hits at all", with incorrect spelling and misuse of boolean (e.g. ANDing Norwegian and English versions of the same word) accounting for many of the problem. I'm not sure about the conclusions, though; don't the results show that you can't just rely on allegedly "user friendly" systems? The article concludes as follows: "Should the library provide the actual searching for the patrons, or is it simply more cost efficient to increase focus on user instruction? So far, the debate has leaned heavily on the second solution. With library systems becoming increasingly user friendly (we hope) it is quite difficult to see how we could persuade library patrons to learn more about searching. To justify spending a large part of the library budget on databases that users find difficult or even unnecessary requires us to sit down and properly discuss the library’s role in information searching and how to best assist students and staff in their research process."