SUPPLEMENTAL MATERIAL This supplement contains three sections: S.1 Analytical Techniques S.2 Interlaboratory Comparisons S.3 Trace Element Experiment with Leached Whole-Rock Samples The reported Nd, Sr, Pb, and Hf isotope data (Table 3) were analyzed in three different laboratories: the Vrije Universiteit Amsterdam (VU), Lamont-Doherty Geological Observatory (LDGO), and San Diego State University (SDSU). In the following sections, we provide a detailed description of the analytical methods and an inter-laboratory comparison. In addition to the trace element analyses on sample glasses, we also analyzed trace element concentrations for leached whole-rock samples. Although the actual concentrations and elemental ratios are significantly affected by this process, the overall normalized trace element patterns are similar to the glasses in a very general sense. Therefore, these data can still provide a rough idea of the level of enrichment of each sample compared to other samples in the dataset. S.1 ANALYTICAL TECHNIQUES S.1.1 Nd isotope composition The majority of Nd data were collected at LDGO, with some duplicates and additions collected at both VU and SDSU. Thermal ionization mass spectrometry (TIMS) techniques from VU and LDGO are described in Valbracht et al. [1991] and Park [1990] respectively. At SDSU a technique modified after Schilling et al. [1994] was used, separating the bulk REEs using an Eichrom Ln-Spec resin. The Nd was analyzed in multidynamic mode on a Nu Instruments HR (High Resolution) MC-ICP-MS (Multi- Collector Inductively-Coupled Plasma Mass Spectrometer). The central detectors were 1 amu apart, while the outer detectors were 2 amu apart. In a typical run, three blocks of 20 ratios were measured, where each ratio consisted of an off-peak baseline measurement followed by three peak switching positions (143, 144, and 145 in the axial collector). Standard procedure consisted of sets of two samples bracketed by a Nd standard. All samples were run dry, i.e. using a desolvating nebulizer. Instrumental mass fractionation was normalized to 146Nd/144Nd = 0.7219 using an exponential fractionation law. The in- house Ames standard gave 143Nd/144Nd = 0.512101 +/- 4 (2sigma/sqrt(n), n=16) for this analytical session of about one week, while the SDSU long-term value for Ames 143Nd/144Nd = 0.512118 +/- 3. This long-term value corresponds to an SDSU lab value of La Jolla 143Nd/144Nd = 0.511832. All reported numbers have been normalized to La Jolla 143Nd/144Nd = 0.511868. Uncertainty of Nd isotope ratios are 2sigma/sqrt(n) of internal run statistics. S.1.2 Sr isotope composition A large number of the Sr data were obtained at LDGO, with a few additions and duplicates from SDSU. The TIMS techniques used at LDGO are described in Park [1990], while several reruns at VU were carried out following Valbracht et al. [1991]. Reruns at SDSU follow the technique of Hanan et al. [2004]. SDSU Sr isotope data were obtained with a Sector 54 TIMS, and were corrected for mass fractionation with 88Sr/86Sr = 0.1194. The reported isotope data are all given relative to SRM 987 = 0.710238 (corresponding to E&A = 0.708000 at VU and LDGO). Uncertainty on Sr ratios are 2sigma/sqrt(n) of run means. S.1.3 Pb isotope composition The Pb isotope compositions of most samples were determined at VU, following Valbracht et al. [1991], augmented by two samples ran at LDGO (technique Park [1990]). Several duplicates and additions were run at SDSU. The SDSU Pb technique has been described by [Hanan and Schilling, 1989]. SDSU Pb isotope data were obtained with a Sector 54 TIMS. Pb errors are +/- 0.05 % per amu, unless quoted otherwise. Where quoted, the numbers represent 2sigma/sqrt(n), based on propagating standard reproducibility into the internal run statistics. In order to do this, these samples were corrected through the SDSU laboratory average for NBS 981 of 206Pb/204Pb=16.880 +/- 0.0037 (2sigma/sqrt(n)), 207Pb/204Pb=15.423 +/- 0.0036 (2sigma/sqrt(n)), 208Pb/204Pb= 36.496 +/- 0.0096 (2sigma/sqrt(n)) to the values of Todt et al. [1984]. S.1.4 Hf isotope composition All Hf chemistry and analyses were carried out at the Ecole Normale Supˇrieure in Lyon (ENSL) following the protocol of Blichert-Toft et al. [1997]. Hafnium isotope compositions were measured by MC-ICP-MS using the VG Plasma 54 at ENSL and the technique of Blichert-Toft et al. [1997]. To monitor machine performance, the JMC-475 Hf standard was run systematically after every two samples and averaged 0.282160 ± 0.000010 (2_) for 176Hf/177Hf throughout this study. 176Hf/177Hf was normalized for mass fractionation relative to 179Hf/177Hf = 0.7325. Hafnium total procedural blanks were less then 25 pg. Uncertainties reported on Hf measured isotope ratios are in-run 2sigma/sqrt(n) analytical errors in last decimal place. S.2 INTERLABORATORY COMPARISONS Even though we did not carry out a rigorous inter-laboratory comparison for all laboratories involved, all laboratories reference their analytical data to some common reference materials and we carried out a sufficient number of duplicate analyses. This comparison suggests that none of our geochemical variations are systematically tainted by inter-laboratory analytical biases. In the following we report all data of duplicate analyses. All of these data are normalized to the same reference samples. S.2.1 Nd Three samples were run at two laboratories and we have at least one duplicate run for every laboratory involved (Table S1). This data set shows that the analyses for one sample are not statistically identical (Figure S1). The difference for that one sample is about 80 ppm, even though this difference is far smaller than the overall data range observed (see Table 3). For this reason, our interpretations are not affected by this relatively minor discrepancy. S.2.2 Sr comparison Although Sr analyses were carried out at all three laboratories, we report the most recently collected data (SDSU) augmented by data from LDGO, where no SDSU samples were available, in Table 3. As shown in Figure S2, only some samples are reproducible within the stated analytical uncertainties (data in Table S2). This could be due to varying degrees of efficiency in leaching as many of these samples were run on separately leached sample splits. Samples 12 and 13 show the largest differences, and we speculate that these samples may show some internal heterogeneity, in addition to differences in leaching efficiency. Such heterogeneity within a whole rock has been documented in previous studies (e.g. Bryce and DePaolo [2004]). In conclusion, caution should be taken when using Sr for quantitative modeling, compared to Sr isotope fingerprints in young and unaltered samples. S.2.3 Pb comparison Similarly to Nd, the Pb dataset shows differences in some duplicate runs that are larger than the analytical uncertainties but they are minor in the context of the overall variation in the data set (Figure S3, Table S3). Moreover, most of the Pb data were collected at one laboratory, making the difference an upper bound to any possible bias. Furthermore, the data for FAS forms a very tight cluster, while showing data from more than one laboratory for this eruptive series. This close correspondence suggests samples are reproducible to at least the accuracy necessary for our discussion, and there may be one sample (Sample 9) that is simply an outlier. S.3 TRACE ELEMENT EXPERIMENT WITH LEACHED WHOLE-ROCK SAMPLES In addition to trace element concentrations on glasses, we also analyzed leached whole-rock powders where glass was not available (Supplemental Table S4). Overall, these data confirm the more enriched nature of the FGST samples compared to the more depleted near-ridge seamounts. The more enriched samples also have the highest Na2O+K2O (Figure 3). However, due to the differential effects of leaching on different elements, we cannot perform quantitative petrogenetic modeling based on the obtained results. Instead, we investigate whether the data can be used as a rough gauge of relative enrichment within the data set. S3.1 Analytical Technique Although it is well known that leaching never perfectly restores fresh rock compositions (e.g. Koppers et al. [2003]), we tested a less aggressive procedure for trace element compositions of whole rocks (cold for up to thirty minutes). Trace element compositions of leached powders were analyzed at Universitˇ Paul Sabatier in Toulouse. Powders were dissolved in an HF-HNO3 mixture and analyzed following Benoit et al. [1996]. Analyses were carried out by ICP-MS (PerkinElmer-Sciex ELAN 5000) with In and Re added as internal standards, and using a cross flow nebulizer together with a Scott spray chamber. Molecular interferences on intermediate REEs were corrected for by using the procedure described in Aries et al. [2000]. Analytical errors for this method are ~5 % for most elements in the ppm range, and ~10 % for elements in the ppb range [Benoit et al., 1996]. The standard BE-N was used for monitoring. The data are reported in Table S4. S3.2 Results Applying a mild leaching technique to our samples allows us to obtain some data on samples for which glass was not available. Since we have both glass and whole rock data for several samples, we can evaluate the effect of leaching and the usefulness of this procedure. Surprisingly, we find that in a broad sense the normalized trace element patterns are similar for both methods giving us confidence that overall patterns are reliable (Figure S4). However, some elements differ by up to 60 % in concentration between both methods, likely due to selective removal of groundmass material during leaching. So, as we anticipated, petrogenetic modeling is not advisable for leached whole rock abundance data. Therefore, the use of these leached whole rock data is limited to rough comparisons in terms of overall enrichment and perhaps abundance ratios of elements that might not be as sensitive to leaching (e.g. HFSE). Furthermore, samples are normalized only to compare the patterns of the samples to each other, since a comparison with the normalizing reservoir would disregard the differential extraction of trace elements during leaching and should thus be avoided. The values here are normalized to CI chondrite to avoid direct comparisons with Figure 4. A broad-stroke comparison between samples shows that the entire province of seamounts defines two groups of patterns (Figure S4b-c): Positive slopes, a sign of incompatible element depletion, are mainly associated with the near-ridge seamounts, while negative slopes, due to enrichment, are mainly found in the FGST. The most enriched samples include both FAS and SAS as well as samples from Hoke and Opal (2 and 10). The majority of near-ridge seamounts are relatively depleted, which is expected for MORB compositions (e.g. Batiza and Vanko [1984]). As an exception, three lavas from the near-ridge seamounts are slightly more enriched than the transitional basalts of FTS. However, alkali basalts do occur in a MOR setting [Batiza and Vanko, 1984], and in the extension re-activated seamounts of Davis et al. [1995; 2002] REFERENCES Aries S., M. Valladon, M. Polvˇ, and B. Duprˇ (2000), A routine method for oxide and hydroxide interference corrections in ICP-MS chemical analysis of environmental and geological samples. Geostandards Newsletter, 24, 19-31. Batiza, R., and D. Vanko (1984), Petrology of Pacific seamounts, J. Geophys. Res., 89, 11235-11260. Benoit, M., M. Polvˇ, and G. Ceuleneer (1996), Trace element and isotopic characterization of mafic cumulates in a fossil mantle diapir (Oman ophiolite), Chem. Geol., 134, 199-214. Blichert-Toft, J., C. Chauvel, and F. Albar¸de (1997), Separation of Hf and Lu for high-precision isotope analysis of rock samples by magnetic sector-multi collector ICP-MS, Contrib. Mineral. Petrol., 127, 248- 260. Bryce, J.G., and D.J. DePaolo (2004), Pb isotopic heterogeneity in basaltic phenocrysts, Geochim. Cosmochim. Acta, 68, 4453-4468. Davis, A.S., D.A. Clague, W.A. Bohrson, G.B. Dalrymple, and H.G. Greene (2002), Seamounts at the continental margin of California: A different kind of oceanic intraplate volcanism, Geol. Soc. Am. Bull., 114, 316-333. Davis, A.S., S.H. Gunn, W.A. Bohrson, L.-B. Gray, and J.R. Hei (1995), Chemically diverse, sporadic volcanism at seamounts offshore southern and Baja California, Geol. Soc. Am. Bull., 107, 554-570. Dieu, J.J. (1995), On the formation of cumulates, characteristics of oceanic lithosphere, and the process of carbonatite metasomatism: a study of ultramafic xenoliths from South Pacific islands, Ph.D. thesis, University of California, San Diego. Hanan, B.B, and J.G. Schilling (1989), Easter microplate evolution: Pb isotope evidence, J. Geophys. Res., 94, 7432-7448. Hanan, B.B., Blichert-Toft, J., Pyle, D., Christie, D, 2004. Contrasting origins of the upper mantle MORB souce revealed by Hf and Pb isotopes from the Australian-Antarctic Discordance, Nature, 432, 91-94. Koppers, A. A. P., H. Staudigel, M. S. Pringle, and J. R. Wijbrans (2003), Short-lived and discontinuous intraplate volcanism in the South Pacific: Hot spots or extensional volcanism?, Geochem. Geophys. Geosyst., 4(10), 1089, doi:10.1029/2003GC000533. Park, K.-H. (1990), Strontium, neodymium, and lead isotope studies of ocean island basalts: Constraints on their origin and evolution, Ph.D. thesis, Columbia University, New York. Schiffman, P. and S. Roeske, (2002), Electron Microprobe Analysis of Minerals, in Encyclopedia of Physical Sciences and Technology, 293-306. Schilling, J.G., B.B. Hanan, B. McCully, R.H. Kingsley, and D. Fontignie (1994), Influence of the Sierra Leone mantle plume on the equatorial Mid-Atlantic Ridge; a Nd-Sr-Pb isotopic study, J. Geophys. Res., 99, 12005-12028. Todt, W., R.A. Cliff, A. Hanser, and A.W. Hofmann (1984), 202Pb and 205Pb double spike for lead isotopic analyses. Terra Cognita, 4, 209. Valbracht, P.J. (1991), Early Proterozoic continental tholeiites from western Bergslagen, Central Sweden: II. Nd and St isotopic variations and implications from Sm-Nd systematics for the Svecofennian sub- continental mantle, Precambrian Res., 52, 215-230. SUPPLEMENTAL FIGURE CAPTIONS Figure S1. Comparison of Nd isotope compositions. Only one sample is not reproduced, but the difference is minor compared to overall variations in the FGST. Figure S2. Comparison of Sr isotope compositions. The compositions of several samples are not reproduced within stated uncertainties, suggesting potential sample heterogeneities of non-systematic interlaboratory biases. This effect is small when compared to the overall variation but it also shows that the data should not be used for quantitative modeling. Figure S3. Comparison of Pb isotope compositions. Similarly to Nd, only one sample is not reproduced, but the difference is smaller than the range in eruptive series. Figure S4. Leached whole rock trace element patterns. (A) Comparison between ICP-MS measurements on whole rocks (dashed lines) and IMP measurements on glasses (solid lines). The general, overall patterns are very similar, but absolute concentrations differ significantly for the most incompatible elements. (B) The near-ridge seamounts show two groups with positive and negative slopes respectively. The samples from Echo (4), Jasper Satellite (5) and Bonanza (7) show a positive, enriched slope, while all others are depleted. (C) The late stages of Jasper (SAS/FAS: samples 13,14,17) and Opal Seamount (10) are distinct from the shield stage, which is more depleted (FTS: sample 16). Flint (11) is part of the trail and falls in the FTS field. Hoke (2) is the most enriched sample.