Immunology Today, vol. 4, No. 10, 1983 271 One man's answer to immunological information overload: microcomputer management of the personal reprint collection Eric Martz The subject of information overload in immunology has been addressed recently by Waksman, who proposed some innovative solutions for the afferent pathways ~'2, (see also reactions by Sanderson 3, Turk 4 and FranklinS). Here Eric Martz describes his solution for efferent overload. In the mid-1970s I had been an immunologist for a few years and had accumulated a couple of thousand reprints in my office file drawers. Often, I would ransack my brain and my office for some time looking for a particular reference to be cited in a manuscript. How frustrating it was to find an earlier citation to the reference which lacked the title when I needed to include it. Then, there was further ransacking of the brain while I tried to reconstruct which of five or so possible subject categories I would have used to file the reprint. As a computer afficionado, it seemed clear to me that great benefit could be derived from storing references in a computer.. At first, my idea was to store only those references which I cited in my own publications. I wrote a small pro- gram in BASIC on a minicomputer at the Health Sciences Computer Facility of the Harvard School of Public Health. Once this began to work smoothly, it seemed logical to program the computer to print bibliographies. The varying style requirements of different journals seemed like something which a computer should be able to handle with ease, freeing the author from these annoying details and from proof-reading many of the same references over and over in different bibliographies with different styles. The tedium of resequencing the citation numbers in a manuscript was a task suited for computerization. Indeed, the entire bibliography of a manuscript can be constructed automatically by having the computer find the citations in the text and then find the corresponding references and print them. These capabilities were programmed into the computer and have continued to evolve towards greater ease of use and reliability. By the mid-1970s I had entered the references to my entire reprint collection into the computer. I found that the cost of maintaining several thousand references avail- able 24 hours per day (on hard disk) and accessing them was about $1000 per annum. This seemed well worth- while, even to a small lab like mine with a single grant. With the advent of inexpensive microcomputers (see below), the cost has been further reduced. I added a sorting routine which enabled me to order alphabetically and print my entire reprint collection. I discovered that having such a list on my desk was invalu- able. As long as I could remember the first author, I could find a complete reference and the reprint itself as quickly as looking up a phone number. Since I could not always remember the first author, I also made a list sorted by senior co-authors. I went to the computer terminal only when I needed to search by subject or remembered only fragments about a paper (such as the year or journal and a phrase in the title). I included space for keywords and other personal notes with each reference. Here I noted 'classic' on papers which especially impressed me, 'classuse' on those I might later assign to my students, and 'PFEX' for papers suggesting possible future experiments. Thus I could search for all' classics', all papers which pertained to my teaching duties, or, when I felt I had run out of ideas, the PFEX papers. One of the advantages of a personal system is that the keywords can be as personal and arbitrary as you like; they need not be public or objective as in the MEDLINE keyterm index. Thus, a cell biologist and a surgeon may use 'adhesion' as a keyword, each meaning different things. I numbered the subject file folders in which I had my reprints filed using an expandable library-type decimal system. Thus, each subject was numbered, and cross- referencing a paper to several subject categories became as simple as putting the corresponding numbers in the keyword space. Thus, '515.5' designated mast cells and TABLE I. Three levels of computer-aided bibliographic systems Worldwide Local Personal Example Type of computer Number of references in database Appointment needed for search Location of computer access Person conducting search Availability of references found Usefulness of references found Automatic extraction from text, resequencing, and restyling of bibliographies MEDLINE PAPER CHASE BIBLIOFILE Large mainframe Minicomputer Microcomputer ~106 ,~I05 103-10 ~ Yes No No Typically remote Local library Personal office Trained specialist End user End user Often unavailable All in library All in personal file drawers Varies widely Varies widely All chosen earlier as useful No No Yes © 1983, Ehevier Science Publishers B.V., Amsterdam 0167 - 4919/83/$01 6 0