BACKGROUND: While it is generally expected that libraries should work as an effective system to fill gaps between information rich and information poor, another side of the issue should be considered that gaps are also induced by libraries themselves. Toward further recognition for the library as a secondary system of knowledge mediation, a quantitative and structural analysis of this mechanism is required. METHODS: "Gaps" among 87 Japanese medical university and college libraries in 1990s were considered and measured in three levels: 1) Gap in condition, measured by number of journals held; 2) Gap in accessibility, measured for 12 libraries using three samples of 333 articles respectively; 3) Gap in the scholarly behavior of their patrons, i.e. variety of cited journals by them, measured by analyzing 249,958 cited articles obtained from Science Citation Index Expanded using SPSS 11.5J. RESULTS: Condition gap was 50 times among the libraries. On the contrary, no behavior gap correlated to the condition gap was detected. An accessibility gap of maximum 1.7 times was observed in 12 libraries, with structural differences between private and national universities. ILL services and e-journals were presumed to be well filling the gap. Usage tendency of articles has two vectors: diffusion toward those in widespread journals and concentration into those in limited core-journals, the latter of which cancels the gap strongly. CONCLUSIONS: The gaps observed among the libraries are effectively suppressed by their ILL activity as well as by concentration of researchers' needs into core-journals.