When interpreting the present results, one should bear in mind that the data, particularly for the small fractions, derive from mixtures of volcanic groundmass with different chemical compositions and, potentially, eruption ages. Indeed, the Antarctic Peninsula area experienced a complex Mesozoic and Cenozoic plate tectonic evolution that was accompanied by igneous activity with different geochemical signatures. Glaciogenic sediments are therefore expected to reflect the complex igneous activity that affected the area in time. An important observation is that volcanic grains show evidence for abrasion resulting from mechanical reworking. This implies that age constraints derived from these samples should be taken as the oldest limit for their deposition.
On the other hand, disturbed age spectra are common even for argon data extracted from a single specimen. Groundmass samples are polymineralic mixtures that may also include glass; each component may retain a distinct argon signature as a function of secondary alteration processes and potential analytical artifacts. Discordant age spectra may therefore arise from a complex interplay of secondary alteration processes (producing 40Ar* loss and/or potassium addition) and possibly from analytical artifacts resulting from the common fine grain size of primary and secondary minerals within the groundmass. Very fine grained mineral phases may experience argon isotope redistribution and/or loss during irradiation because of recoil processes. Discordant 40Ar-39Ar age spectra can occur because the argon step-heating technique, because of the differential thermal stability of mineral phases during in-vacuo heating, has the potential to separate different argon reservoirs hosted in different mineralogical domains. The contribution of each different mineral phase to the age spectra, however, may be monitored by elemental variations (Ca, Cl, and K from neutron-derived 37Ar, 38Ar, and 39Ar, respectively). Meaningful interpretation of groundmass argon data, therefore, requires the support of petrographical and chemical analyses at a microscopic scale. As an example, Lo et al. (1994) and Koppers et al. (2000) have shown that argon step-heating analyses on volcanic rocks commonly exhibit low apparent Ca/K ratios and high meaningless apparent ages in the low-temperature region. This feature derives from the lower release temperature of K-rich alteration mineral phases that may experience 39Ar recoil loss during irradiation. Koppers et al. (2000) also observed low meaningless apparent ages in high-temperature steps that, based on the concomitant increase of the Ca/K ratios, was attributed to preferential degassing at the high temperature of plagioclase and clinopyroxene affected by 37Ar recoil loss. Lo et al. (1994) instead observed high meaningless apparent ages in the high-temperature steps. This was related to the release of argon from phenocrysts that may host extraneous argon (both inherited or excess argon). However, the intermediate region of the age spectra attributed to outgassing of glass and groundmass plagioclase by Lo et al. (1994) and of interstitials (glassy or microcrystalline) by Koppers et al. (2000) may yield meaningful ages. In general, volcanic glasses were dated with variable success. On one hand, both low and high meaningless ages were reported (e.g., Mankinen and Dalrymple, 1972; Walker and McDougall, 1982; Cerling et al., 1985). This is attributable to (1) hydration and devitrification and (2) incorporation of excess argon in quenched basaltic glass in the submarine environment (cf. McDougall and Harrison, 1988). On the other hand, other studies reported K-Ar or 40Ar-39Ar data on volcanic glass yielding internally consistent results (e.g., Sharp et al., 1996; Pinti et al., 2001) or concordant with coexisting alkali feldspar (e.g., Drake et al., 1980). Thus, the validation of a 40Ar-39Ar age derived from volcanic glass requires the support of independent evidence or of different geochronological data.
Examination of the incremental heating analyses on milligram-sized fractions (Table T2) reveals that the sample from Site 1097 is characterized by a significant fraction of gas (37% of 39Ar released) with Ca/K ratios <<1, whereas, in spite of the extended low-temperature heating schedule, no Ca/K ratios <1 were detected in the samples from Site 1103. This, in agreement with SEM-EDS observation, reflects the higher alteration degree of the sample from Site 1097 than the samples from Site 1103. It is worthy of note that the final steps of the milligram-sized and small fractions for all samples do not show high Ca/K ratios attributable to a significant contribution from pyroxene (Ca/K > 300 in the study samples). Most of the Ca/K ratios derived from argon data for samples from Site 1103 are compatible with outgassing of groundmass glass with possible contamination by groundmass plagioclase (Ca/K > 16, based on EDS analyses). Nevertheless, Cl/K ratios derived from argon data show a narrower variation than electron microprobe analyses (0.024-0.061 vs. 0.033-0.095 and 0.021-0.048 vs. 
0.008-0.12 for Sample 176-1103A-36R-3, 4-8 cm, and 31R-2, 0-4 cm, respectively) (Tables T1, T3). The meaning of such disagreement remains uncertain. When comparing argon data with EMP analyses, one should bear in mind that the latter refer to single-grain analysis. By contrast, argon analyses investigated a volumetrically much larger sample, which may approach the true average of the whole population.
Figure F8 shows three-isotope correlation plots whose argon isotope ratios have been converted to age and elemental ratios. The distribution of data points in Figure F8 is compatible for the three samples with multicomponent systems. It is interesting to note that, based on different recoil lengths of argon isotopes (i.e., 37Ar > 39Ar >> 38Ar) (Onstott et al., 1995), it is likely to obtain during sample irradiation of very fine intergrowths (at least for intergrowths >20 nm thick) significant displacement of 37Ar and 39Ar accompanied by an undetectable effect on 38Ar. In Figure F8A this process would produce a distribution of data points along a trend with a positive slope, whereas it would have a less important effect on the Ca/K ratios. Whereas a similar process may explain the argon isotope compositions of the low-temperature region of the milligram-sized fraction of Sample 178-1097A-27R-1, 35-58 cm, and of the first step (age ~66 Ma) of the milligram-sized fraction of Sample 178-1103A-36R-3, 4-8 cm, it cannot explain (1) the distribution of the remaining steps of the milligram-sized fractions, (2) the argon isotope composition of the small fractions, and (3) the overall negative correlation of Figure F8B. The negative correlation between age and Cl/K, such that lower Cl/K analyses give higher ages, is the most striking feature of Figure F8. Based on the Ca, Cl, and K contents, the Cl/K variations correspond to differences in Cl. Given the imperfect agreement between argon data and electron microprobe analyses on glasses (see above) and the lack of Cl determinations on volcanic rocks of the area from literature, the question of whether the Cl variation reflects a true increase with decreasing age of the igneous activity or is an analytical artifact remains unclear and requires further investigation.
Following the interpretation that argon data from Site 1103 are derived mainly from volcanic glass and assuming that analyzed glasses are free from K and Ar mobility, we may speculate that the age intervals defined by the argon data of the small fractions (18-57 Ma for Sample 178-1103A-31R-2, 0-4 cm, and 7.6-50 Ma for Sample 178-1103A-36R-3, 4-8 cm) provide for each sample an estimate of the maximum youngest limit and the minimum oldest limit for the eruption age of the volcanic sequence sampled by the glacial erosion. Because for Sample 178-1097A-27R-1, 35-58 cm, fresh glass was not detected on epoxy-mounted grains, we believe that the age range of 75-173 Ma can be taken only as indicative of the age variation of the volcanic clasts recovered at 220 mbsf at Site 1097. This age interval vastly exceeds the plausible depositional age of sequence group S3. At Site 1103, the hypothesized maximum youngest limit for the volcanic activity as sampled by glacial erosion (7.6 ± 0.7 Ma) and detected close to the bottom of drilling (340 mbsf) is compatible with the age range of the diatom Zone Actinocyclus ingens v. ovalis (6.3-8.0 Ma) determined for the interval 320-355 mbsf at the same site (Shipboard Scientific Party, 1999c). In addition, it agrees with the maximum ages of 7.4 (+1.5/-0.9) and 7.8 (+1.5/-1.1) Ma obtained at 262-263 mbsf from strontium isotope composition of barnacle fragments (Lavelle et al., Chap. 27, this volume). With the above-mentioned assumptions concerning K and Ar mobility, the 7.6-Ma age provides an estimate for the maximum age of sequence group S3 at 340 mbsf, bearing in mind, in addition, that (1) the distribution of the age-steps (only two ages younger than 10 Ma of the 26 analyses) suggests the presence of grains with apparent ages younger than 7.6 Ma (Fig. F9) and (2) analyzed grains show evidence for abrasion supporting mechanical reworking. An additional point deserving attention is that from the distribution of the age steps of the small fractions at Site 1103 (Fig. F9) we note that with decreasing sample depth (340 and 290 mbsf for Samples 178-1103A-36R-3, 4-8 cm, and 31R-2, 0-4 cm, respectively), volcanic clasts show overall increasing apparent ages. This is compatible with glacial erosion that affected with time deeper levels of a volcanic sequence previously deposited on the continent. This observation is compatible with the interpretation of S3 as representing an initial stage of development of the glacial margin that preceded the dramatic advances of the ice sheet to the continental shelf edge represented by S1 and S2 (Barker et al., in press).
