Terrestrial massif anorthosites have gained new interest for the understanding of the deep zones of the crust and for the reconstitution of its history in Proterozoic time. The purpose of this paper is to show how trace elements can enlighten two controversial questions in the problem of anorthosites, namely the nature of the parental magma and the process which gives rise to related acidic rocks. Data obtained on rocks and minerals coming from the Rogaland anorthositic province, South Norway, are presented together with those available in the literature. The Sr and Ba in plagioclase and their relationship with Ca and K, K and Rb in rocks and plagioclases, rare earth elements (REE) in cumulate minerals and in various liquids, 87 Sr/ 86 Sr initial ratios on rocks and minerals as well as a few data on transition elements and on 18 O/ 16 O ratios are discussed in the different sections. Quantitative modelling of the behaviour of trace elements is realized mainly by graphical methods in the Bjerkrem-Sogndal layered lopolith and in the Hidra body, both andesine-type massifs. The major conclusions are as follows:(1) The parental magma of the andesine-type massifs has a jotunitic (hypersthene-monzodioritic) composition characterized by variable K/Rb ratios (from 480 to 1700), by the absence of an Eu anomaly, by variable REE contents (from 50 to 220 for La chondrite-normalized content) and by La/Yb ratios almost constant (from 6 to 12), as well as by high Ti and Fe contents, by transition elements indicating calc-alkaline affinities and by 87 Sr/ 86 Sr initial ratios similar to those of rocks derived from the upper mantle or the deep crust.(2) A jotunitic composition appears not to be compatible with the parental magma of the labradorite-type massif anorthosite. The relationship between andesine anorthosite and labradorite anorthosite cannot be described simply in terms of fractional crystallization.(3) Fractional crystallization can explain the succession of rocks from andesine anorthosite to leuconorite, to norite and finally to acidic rocks. In some cases however, namely the Bjerkrem-Sogndal lopolith in Rogaland, contamination by supracrustal material must be invoked and seems to have superimposed its effects on those of fractional crystallization.(4) Deformation, granulation and recrystallization do not appear to fractionate both the major and trace elements of the plagioclase, except Ti which is lowered.