Objective - To identify similarities and difference in the questions asked at the virtual and physical refernece desks of a helath scienmces library, in order to better undertand user needs and highlight areas for service improvement. Also to retrospectively analyze reference statistics collected over the previous six years. Design - Use study; retrospective study of reference statistics for the period July 1997 to June 2003; literature review. Setting - Large academic helath sciences library in the United States. Subjects - All questions asked at the reference and information desks, plus questions submitted to the University-wide virtual reference service and answered by a health sciences librarian, over a period of one month. The questions were asked by faculty, staff, students and members of the public. Methods - A literature review was carried out to examine the types of information/reference questions typically asked in health sciences libraries both before and after the mass introduction of remote end-user searching of online resources and the establishment of virtual reference services. Next, the reference statistics collected at the University of Illinois at Chicage (UIC) Library of the Health Sciences between July 1997 and June 2003 were examined. For most of this period a digital reference service was offered using a listserv address to which patrons would submit email queries. Beginning in March 2003, a formal virtual reference service (chat and email) was provided using commercial software. Finally, data was gathered on questions answered by a health sciences librarians, and clients who asked the question, at either the physical or cirtual reference desk, during the month of November 2003 at the UIC Library of the Health Sciences. Library staff completed an online survey form for each question, and if a client asked more than one question, each question was coded individually. Data included: status of client using the service (faculty/staff. undergarduate student, graduate student, non0UIC, unknown); mode of submission (email, chat, phone, in person); and type of question asked (directional, ready reference, in-depth/mediated, instructional, technical, accounts/status and other). In subsequent analysis, the original seven types of questions were further broken down into 19 categories. Main results - It was not possible to undertake a meta-analysis or systematic review of the studies identified in the literature review because of differences in time frames, settings and the categories used to code reference questions. However the following trends emerged: directional questions accounted for between 30 and 35% of questions asked at both the physical and virtual reference desks; the remainder of questions were generally about known item searched, library policies and services, research, dadabase use and quick reference. The statistics collected at UIC Library of the Health Sciences over the period July 1997 to June 2003 were analyzed. Coded reference questions fell into one of four categories; ready reference, in-depth reference, mediated searched and digital reference. There was a noticeable drop in the number of reference questions received in 1999/2000 which reflects trends reported in some of the studies identified in the literature review. The number of mediated searches decreased from 154 in 1997-98 to 4 in 2002/2003, but the number of digital reference questions increased from 0 to 508 in the same period. Statistics were collected over the month of November 2003 for 939 questions asked at the reference and information desks which included: 38 email; 48 chat; 156 phone; and 697 in person. The major findings were as follows: - appoximately 55% of questions were reference questions (33.5% ready reference, 9.7% in-depth/mediated, 9.7% instructional); 30% were directional; and 10% were technical; it is not stated what the remaining 5% of questions were; - library clients who asked the questions comprised graduate students (26%), faculty (24%), undergraduate students (24%) and non-UIC patrons (22%); - all groups were most likely to ask their reference questions in person; - graduate students were the largest group of email reference (34%), undergraduates were the largest group to use chat (35%) and faculty/staff were the largest group to use the telephone (39%); - 28% of the questions were answered from the library's staff's general knowledge of the library; 22% using the online catalogue or electronic journals list; 22% by referrning the patron to, or using an online database or resource; 13% by referring the patron to another department in the library (such as circulation); 7% by consulting another individual; and 5% by using print resources. Conclusion - The results of the three parts of this study - literature review, study of 1997-2003 statistics, and in-depth collection of statistics for November 2003 - are difficult to compare with each other. However, the general trend emerging from the results is that some kinds of questions asked by health sciences library clients have remained very much the same regardless of the rise of remote end-user searching and the mode of transmission of the questions. These include questions about library policies and services, journal and book holdings, database searching and instructional support. Questions about citation verification and consumer health appear to be decreasing while technical questions and questions about accessing remote databases and online journals are increasing. The majority of reference questions are still asked in person.