Monitoring of telluric current, which is practically a synonym for geoelectric potential difference, was conducted on Kozu-shima Island about 170 km south of Tokyo from May 14, 1997 to June 25, 2000. During the monitoring period, 19 anomalous telluric current changes (ATCs) were observed. Their possible correlation with nearby earthquakes was statistically examined by assuming various lead times for different ranges of magnitude and focal distance. The best correlation may be obtained for earthquakes with a magnitude greater than 3.0 occurring within 20 km of focal distance. There were 23 such earthquakes, of which 11 were preceded by ATCs within 30 d. Of these 11 earthquakes, preceding ATCs of 5 and 6 were positive and negative polarities of telluric current, respectively. Their epicenters were spatially well clustered in the east and west of the island. These facts were clearly beyond those expected by chance and led to a simple speculative model.T he VAN method, named after the initials of Varotsos, Alexopoulos, and Nomicos, for short-term earthquake (EQ) prediction by means of the monitoring of preseismic telluric current signals, called seismic electric signals (SES), has been in successful practice since the 1980s (1). The VAN method, however, has been a contentious subject with regard to the causal relationship between SES and EQs (e.g., refs. 2, 3). One way to get around this would be to accumulate as many case studies as possible on one hand and to build plausible physical models on the other (1, 4, 5, 6).The VAN method is characterized by the notion of "selectivity," which has two aspects (3, 7): (i) there are only selected sites which are sensitive to SES, and (ii) a sensitive site is sensitive only to the SES from some specific focal area(s). A map identifying those focal area(s) is called the "selectivity map" of that site. The selectivity is thought to originate from inhomogeneity of the subterranean electrical structures, implying that SES is transmitted only through subterranean conductive channels (1, 4). Another remarkable property of SES, called the "VAN relation," is expressed by Eq. 1 in terms of the focal distance r, the magnitude M, and the amplitude E of SES.where a is a constant ∼0.34-0.37 and b is a site-dependent constant. The selectivity and VAN relation are the backbone of the VAN method. In Japan, VAN-type telluric current monitoring was conducted and significant preseismic signals were observed in the late 1990s and early 2000s (6,8,9). The site in Kozu-shima Island was unique in showing SES-like telluric current changes before many EQs (9), whereas other stations showed such signals only for one EQ (6,8). In addition, the EQs occurred in the east and west of Kozu-shima Island after the positive and negative polarities of the SES-like telluric current changes, respectively (9). So, in this paper we have tried to statistically examine the correlation between the SES-like telluric current changes and the EQs for Kozu-shima Island. In the following, the term "geoelectric potential dif...