Water-level/pressure data recorded at a dense network of 16 wells of depths ranging from 23 to 201 m within 400 m of the Tono Mine in Gifu prefecture, central Japan, have been studied in search of possible earthquake-related changes. The study area is traversed by the east-west trending Tsukiyoshi fault, which is ordinarily impermeable, sustaining a significantly higher groundwater pressure on the north side of the fault than the south side. On March 16, 1997, a local earthquake of magnitude 5.8 occurred about 50 km south of Tono. The related water-level/pressure changes recorded at different sites showed different features. At three deeper wells on the north side of the fault the water level and pressure showed large coseismic drops (up to 29 cm) followed by larger rises (of as much as 1.8 m). Two days before the earthquake the water level at another deeper well began to show a total drop of about 3 m during the next 2 months. Most of the shallow wells on the north side of the fault showed water-level drops after the earthquake. On the south side, however, only a small delayed surge and a subsequent gradual increase were recorded at a deeper well. Long-term data since 1989 at one of the deeper wells on the north side showed coseismic drops and recoveries for 25 moderate local and large distant earthquakes up to about 1000 km away. The different observed earthquake-related water-level/pressure changes cannot be explained by the static strain field calculated on the basis of poroelastic dislocation models; the changes require earthquake-related permeability changes at the monitoring sites. The high sensitivity of some of the water wells to seismic shaking may be due to the near-critical condition of the tapped aquifer/barrier system beginning several years prior to the March 16 earthquake, such that the local seismic shaking caused by each of the above mentioned earthquakes was sufficient to produce some quickly recoverable fissures in the system, resulting in the observed water-level changes. The sensitivity appears to be variable, showing an apparent decrease during a 1-year period after the earthquake. ¿ 1999 American Geophysical Union