Deep drilling was carried out of two sedimentary sequences considered as representative of reservoir formations; one at Soultz-sous-Forets in Alsace and the other at Balazuc in the Ardeche, France. The natural fracturing in these predominantly sandstone sequences has been analysed both from the continuous cores collected during the drilling and from wall imagery obtained by acoustic (BHTV) and electric (FMS) methods. This core-imagery comparison shows, quite logically, that the core analysis gives more complete results than the interpretation of wall imagery. With the BHTV it was possible to characterize 50% of the fractures seen in the cores, as against only 6% using the FMS. In the Bundsandstein at Soultz the fractures that form clusters (i.e. very close to one another) are not correctly recorded by the BHTV. This bias is even more significant with the FMS since it applies to the entire fracture distribution, regardless of the fracture spacing. This fracture-detection filtering that we find with the imagery techniques are even more evident where the fractures are filled (little physical contrast) or small (below the detection threshold). The core-imagery comparison also shows that the horizontal resolution of the imaging tool is directly proportional to the number of fractures detected. The BHTV, which has a resolution of between 1 and 2 mm, does not detect fractures of this thickness or less, i.e. 45% of the population. This percentage is similar to the amount of fractures recognized by BHTV compared to that recorded in core. The FMS, with a much coarser resolution of about 7.5 mm, can only detect fractures with at least this thickness of infill material. In the Balazuc sequence, 6% of the fractures were detected by the FMS, which corresponds exactly to the proportion of core fractures with a thickness greater than the tool＇s resolution. Despite this sampling bias, the major fracture directions were correctly detected with both the imaging techniques used.