All fourteen REE and Y have been determined for eight GSJ reference rocks including granitic and rhyolitic samples by ICP-AES with a preconcentration method. The two rhyolites (JR-1 and JR-2) and one granite (JG-2) exhibit convex tetrad effects of the M-type even in their chondrite-normalized REE patterns. Their tetrad effects, however, are modified from ideal ones by some step-like variations in their abundances from Tm to Lu. Similar step-like irregularities are seen commonly in chondrite-normalized patterns of other samples: granodiorites (JG-1, JG-la, and JG-3), a fresh water lake sediment (JLk-1), and an alkali basalt (JB-1). We normalized the REE analyses for JR-1, JR-2, and JG-2 by those for JB-1, and then found that they clearly indicate quite regular convex tetrad effects of the M-type without apparent irregu larities in heavy REE. The other samples, when similarly normalized by JB-1, also display analogous but less marked convex tetrad effects except for JG-3. Our present results are strong evidence for the tetrad effect in magmatic processes relevant to granitic and rhyolitic rocks formations. There are common chemical characteristics among JG-2, JR-l, and JR-2: (1) large negative Eu anomalies, and (2) rather high F contents of about 1,000 ppm. The chemical characteristics and the theoretical basis of tetrad effects may suggest that the granitic and rhyolitic rocks with the convex tetrad effects of the M-type are residual silicate melts having partitioned REE with fluids in late stages of differentiation processes. The basalt-normalization appears to be so effective for extracting tetrad effects of granitic and rhyolitic rocks. It filters off the REE characteristics common in magmatic products like the step-like irregularities in heavy REE abundances when normalized by chondrite. This makes it easier to identify lanthanide tetrad effects in magmatic rocks.