Seek and you shall find?

Unless you read EMBO reports cover to cover, or were alerted by this issue’s table of contents, chances are that you found this article by using a search engine. So, what were you looking for? Information on search engines? Google? PubMed? Literature searches? Web services, SOAP, WSDL, text retrieval or text mining? Systems biology? If so, any of these keywords or phrases was probably enough for a search engine to put this article high on its ‘hit list’. If this article is not exactly what you were after, you will have to start again, using a new selection of search terms that may give you a better fit to your expectations. In either case, you will probably not look beyond the first page of results, considering it easier and faster to reiterate this trial-anderror process than to sift carefully through hundreds, if not thousands, of other hits. So how smart is your favourite search engine? Can you usually rely on its results? Given the enormous reach of the worldwide web, do we need better search engines? Do you really care? Most scientists quite sensibly expect that search engines, similar to many other ‘under the hood’ technologies, should work quickly, efficiently, accurately and above all, unobtrusively—not an unreasonable demand in today’s research environment where you may feel that you need to outrun Lewis Carroll’s Red Queen simply to stay where you are. Nevertheless, in the same way that the quality of reagents is crucial to the success or failure of an experiment, the quality and limitations of a search engine, as well as the skill with which it is used, largely determine the success or failure of a web search. I still recall the early days when literature searching was almost the exclusive domain of the trained librarian, who needed to be fully briefed before embarking on a series of detailed searches that often lasted for several days. Today, most life scientists will not bother their librarian, but will instead use the National Library of Medicine’s PubMed service at one stage or another during their literature search. However, an increasing proportion of young researchers prefer Google or Google Scholar as, at least, a starting point and, more worryingly, also as an endpoint. Furthermore, they expect it to produce significant results from only a few relevant keywords. The average query length in Google is approximately 2.2 words (O’Reilly, 2004) and most users seem to think that this is enough to find the most recent and most notable literature within the top ten search results. Imagine what the result would be if the same two-word question was fired at a trained librarian.