This study aims at finding out which attributes (e.g., names, location etc.) people actually recall about their own documents (electronic and paper), and how they recall them, in order to provide recommendations on how to design tools that allow users to retrieve their electronic files more effectively and more easily. A quasi-experiment was conducted with fourteen participants at their workplace. They were asked first to recall features about one (or several) of their own work document(s), and secondly to find these documents. First of all, when participants were asked to find their document(s), they only used a small subset of the attributes they were able to recall, and often they had difficulties in finding their document(s), with current available tools. This correlates with the need for better tools for retrieval. In addition, the results indicate which document attributes are more often recalled (e.g., keywords) but also which ones are best recalled, that is with less errors (e.g. type). Thus, results indicate which attributes are candidate for facilitating file retrieval. Results also provide potential recommendations on: how the retrieval system should allow users to express these attributes; how the retrieval system should include a tolerance in the results to overcome the unreliable nature of memory; and what particular information in relation to the use of files should be tracked by the system. Finally, results suggest that the design of search tools could exploit the fact that the recall of document attributes depend on the type of user and the frequency of use of the files.