An Icelandic hyaloclastite, mostly composed of millimetric fragments of basaltic glass, that is fresh at the bottom of the unit but largely palagonitized in the upper part, has been studied by petrologic, mineralogical, and magnetic means, with the aim of determining the nature and characteristics of the natural remanent magnetization (NRM). The NRM was generally found to consist of two components: a thermodetrital remanent magnetization (thermo-DRM) and a crystallodetrital remanent magnetization (crystallo-DRM). Thermo-DRM and crystallo-DRM are defined here as the remanences acquired as a result of the deposition of magnetic particles of detrital origin individually carrying either a TRM (thermoremanent magnetization) or a CRM (crystallization remanent magnetization), respectively. Regardless of the chemistry and size and of these particles, Thellier experiments carried out on samples carrying a thermo-DRM provide apparent paleointensities close to the expected geomagnetic paleointensity, which suggests that in the present case the fractional alignment of individual magnetic moments is similar for DRM and TRM. In the upper part of this outcrop, grain growth CRM was acquired by individual grains of magnetite which crystallized as a result of palagonitization of basaltic glass at low temperature (<100 ¿C). No systematic difference could be observed between the directions of characteristic remanence in the layers richest in magmatic magnetic grains and the layer where secondary magnetite is the main remanence carrier. In both cases, the overall remanence exhibits a large inclination error (~20¿), and the samples have a marked anisotropy of magnetic susceptibility which is typical of sedimentary fabrics. Thus, secondary magnetite probably formed prior to the deposition of particles, and the bulk remanence in the palagonitized layers is a crystallo-DRM rather than a CRM. Throughout the entire stratigraphic thickness, Thellier paleointensity data are of good or excellent quality regardless of the nature of the primary remanence. In agreement with previous theoretical inferences and experimental results the layers carrying a crystallo-DRM provide a much lower (by a factor of 2) apparent paleointensity than the layers where the remanence is a thermo-DRM. This suggests that palagonitized basaltic glasses should not be used for paleointensity determinations. Our study shows that application of the Thellier thermal paleointensity method to sedimentary rocks can be a useful tool for distinguishing crystalline versus thermal blocking of the magnetic moments of the individual particles which, after deposition, carry a DRM. More generally, discrepancies between relative paleointensities obtained from sediments can be expected if rocks with different proportions of thermo-DRM, crystallo-DRM, or CRM are compared. ¿ 1999 American Geophysical Union