The paleomagnetic investigations carried out in the 70's on Oligo-Miocene volcanics of Sardinia have demonstrated that the island was turned by 35-30 degrees clockwise from 33 Ma up to 21-20.5 Ma and rotated counterclockwise in a few million years [De Jong et al., 1969, 1973; Bobier et Coulon, 1970; Coulon et al., 1974; Manzoni, 1974, 1975; Bellon et al., 1977; Edel et Lortscher, 1977; Edel, 1979, 1980]. Since then, the end of the rotation fixed at 19 Ma by Montigny et al.  was the subject of discussions and several studies associating paleomagnetism and radiometric dating were undertaken [Assorgia et al., 1994; Vigliotti et Langenheim, 1995; Deino et al., 1997; Gattacceca et Deino, 1999]. This is a contribution to this debate that is hampered by the important secular variation recorded in the volcanics. The only way to get out of this problem is to sample series of successive flows as completely as possible, and to reduce the effect of secular variation by the calculation of means. Sampling was performed north of Bonorva in 5 pyroclastic flows that belong to the upper ignimbritic series S12 according to Coulon et al.  or LBLS according to Assorgia et al.  (fig. 1). 40 Ar/ 39 Ar dating of biotites from the debris flow (MDF) has yielded an age of 18.35+ or -0.03 Ma [Dubois, 2000]. Five of the investigated sites are located beneath the debris flow (TV, TVB, TVD, SPM85, SPM86), one site was cored in the matrix of the debris flow (MDF) and one in 4 metric blocks included in the flow (DFC). Another site was sampled in the upper ash flow (PDM) that marks the end of the pyroclastic activity, just before the marine transgression. According to micropaleontological and radiometric dating this transgression has occurred between 18.35 and 17.6 Ma [Dubois, 2000]. After removal of a soft viscous component, the thermal demagnetization generally shows a univectorial behaviour of the remanent magnetization (fig. 2a). The maximum unblocking temperatures of 580-620 degrees (tab. I) and a rapid saturation below 100 mT (fig. 3) indicate that the carrier of the characteristic magnetization is magnetite. The exception comes from the upper site PDM in which were found two characteristic components, one with a normal polarity and low unblocking temperatures up to 350 degrees C and one with a reversed polarity and maximum unblocking temperatures at 580-600 degrees C of magnetite. After calculation of a mean direction for each flow, the mean "A1" direction 4 degrees /57 degrees (alpha 95 = 13 degrees ) computed with the mean directions for the 5 flows may be considered as weakly affected by secular variation. But the results require a more careful examination. The declinations are N to NNW beneath the debris flow, NNW in the debris flow, and NNE (or SSW) above the debris flow. The elongated distribution of the directions obtained at sites TVB and TVD, scattered from the mean direction of TV to the mean direction of MDF is interpreted as due to partial overprinting during the debris flow volcanic episode. The low temperature component PDMa is likely related to the alteration seen on thin sections and is also viewed as an overprint. As NNE/SSW directions occur as well below (mean direction "B": 5 degrees /58 degrees ) as above the debris flow (PDMb: 200 degrees /-58 degrees ), the NNW directions ("C": 337 degrees /64 degrees ) associated with the debris flow volcanism may be interpreted as resulting from a magnetic field excursion. According to the polarity scale of Cande and Kent [1992, 1995] and the radiometric age of MDF, the directions with normal polarity (TV, TVB, TVD, SPM85, SPM86a, MDF, DFC) may represent the period 5En, while the directions with reversed polarity PDMb and SPM86b were likely acquired during the period 5Dr. Using the mean "A1" direction, the mean "B" or the PDM direction (tab. I), the deviation in declination with the direction of stable Europe 6.4 degrees /58.7 degrees (alpha 95 = 8 degrees ) for a selection of 4 middle Tertiary poles by Besse et Courtillot  or 7 degrees /56 degrees (alpha 95 = 3 degrees ) for 19 poles listed by Edel  can be considered as negligible. Using the results from the upper-most ignimbritic layer of Anglona also emplaced around 18.3 Ma [Odin et al., 1994], the mean direction "E" (3 degrees /51.5 degrees ) leads to the same conclusion. On the contrary, when taking into account all dated results available for the period 5En (mean direction "D" 353 degrees /56 degrees for 45 sites) (tab. II), the deviation 13 degrees is much more significant. As the rotation of Sardinia started around 21-20.5 Ma, the assumption of a constant velocity of rotation and the deviations of the Sardinia directions with respect to the stable Europe direction locate the end of the motion between 18.3 and 17.2 or 16.7 Ma (fig. 4). During the interval 18.35-17.5 Ma, the marine transgression took place. At the same period a NE-SW shortening interpreted as resulting from the collision of Sardinia with Apulia affected different parts of the island [Letouzey et al., 1982]. Consequently, the new paleomagnetic results and the tectono-sedimentary evolution are in favour of an end of the rotation at 17.5-18 Ma.