The onset of sustained Antarctic glaciation across the Eocene-Oligocene transition (EOT) marks a pivotal change in Earth's climate, but our understanding of this event, particularly the role of the carbon cycle, is limited. To help address this gap we present the following paleoceanographic proxy records from Ocean Drilling Program Site 1218 in the eastern equatorial Pacific (EEP): (1) stable isotope (δ18O and δ13C) records generated in epifaunal benthic foraminifera (Cibicidoides spp.) to improve (double the resolution) the previously published records; (2) δ18O and δ13C records measured on Oridorsalis umbonatus, a shallow infaunal species; and (3) a record of benthic foraminifera accumulation rate (BFAR). Our new isotope data sets confirm the existence at Site 1218 of a two-step δ18O increase. They also lend support to the hypothesized existence of a late Eocene transient δ18O increase and early Oligocene Oi-1a and Oi-1b glacial maxima. Our record of BFAR indicates a transient (∼500 kyr) twofold to threefold peak relative to baseline Oligocene values associated with the onset of Antarctic glaciation that we attribute to enhanced biological export production in the EEP. This takes the same general form as the history of opal accumulation in the Southern Ocean, suggesting strong high-to-low-latitude oceanic coupling. These findings appear to lend support to the idea that the EOT δ13C excursion is traceable to increased organic carbon (Corg) burial. Paradoxically, early Oligocene sediments in the EEP are extremely Corg-poor, and proxy records of atmospheric pCO2 indicate a transient increase associated with the EOT.
 Evidence that Earth's carbon system changed dramatically during the EOT comes from a variety of geochemical and sedimentary sources and includes: (1) an approximately 1‰ positive excursion in benthic δ13C lasting ∼1 million years (Myr), (2) an up to 1.3 km deepening of the calcite compensation depth (CCD), and (3) changes in biogenic marine sediment accumulation [Diester-Haass and Zahn, 1996; Salamy and Zachos, 1999, Vanden Berg and Jarrard, 2004; Coxall et al., 2005; Ravizza and Paquay, 2008]. A decrease in atmospheric pCO2 is also suggested [Pearson et al., 2009], but the timing and structure of this change is still poorly constrained. These observations imply that the carbon system played a crucial role in the abrupt switch in climate mode but the mechanisms linking these changes are not well understood.
 Outside the Southern Ocean, productivity increases across the EOT have been reported in the South Atlantic [Hartl et al., 1995], the equatorial Atlantic [Diester-Haass and Zachos, 2003], NW Australia [Diester-Haass and Zahn, 2001] and a number of sites close to continental margins [Ravizza and Paquay, 2008]. The majority of these studies, however, are of low resolution and lack sufficient age control to assess the relationship with the intensification of Antarctic glaciation. Moreover, the role of the eastern equatorial Pacific (EEP), which today supports a significant proportion (18%–56%) of global new production and is responsible for about two thirds of ocean-to-atmosphere CO2 release [Chavez and Toggweiler, 1995; Takahashi et al., 2002, 2009] is virtually unknown. Existing data sets from the EEP over the EOT interval are sparse and point to different interpretations of change in productivity. Patterns of bulk sediment accumulation have been interpreted as signaling no relative increase in EEP export production and burial [Schumacher and Lazarus, 2004, Vanden Berg and Jarrard, 2004; Griffith et al., 2010], yet a spike in diatom abundance close to the E-O boundary recorded in several equatorial Pacific sites [Fenner, 1984; Baldauf, 1992; Barron et al., 2004] suggests at least a short-lived increase in surface production, involving increased silica supply, associated with widespread Antarctic glaciation. Chronologic and temporal resolution of these existing diatom records, which are derived from discontinuous rotary-drilled deep-sea cores, however, is insufficient for making direct comparisons to high-resolution proxy records of glaciation.
 Scientific ocean drilling (ODP Leg 199 [Lyle et al., 2002]; IODP Expedition 320 [Pälike et al., 2009]) has greatly improved spatial and stratigraphic coverage of EOT sections in the EEP. Our study capitalizes on the high-quality sequence recovered at ODP Site 1218 (Leg 199) to improve on the epifaunal benthic foraminifera stable isotope record published by Coxall et al.  and add infaunal stable isotope and BFAR records to gain further insight into carbon cycling across the EOT.
 Here we report three new data sets based on benthic foraminifera for the EOT from ODP Site 1218: (1) epifaunal benthic stable isotope (δ18O and δ13C) records based on Cibicidoides spp. that double the resolution of those published by Coxall et al. , (2) accompanying δ18O and δ13C records for the shallow infaunal species O. umbonatus, and (3) a record of benthic foraminifera accumulation rate (BFAR) in two size fractions. Our principal aim is to test the hypothesis that the onset of widespread Antarctic glacial activity in the earliest Oligocene was associated with a change in primary productivity and export production in the EEP.
2. Oceanographic and Geological Setting
 ODP Leg 199 drilled a series of holes along a latitudinal transect in the equatorial Pacific to obtain climate records through the Paleogene [Lyle et al., 2002]. Site 1218 is situated in the EEP (Figure 1) on a basement swell ∼3°N of the Clipperton Fracture Zone [Shipboard Scientific Party, 2002]. Sediments of EOT age are not deeply buried at this site (∼215 m below seafloor) because northeastern motion of the Pacific plate carried the site away to abyssal depths where red clay accumulation was slow (∼0.5 cm kyr−1). The recovered cores, therefore, contain mostly pelagic oozes, typically radiolarites in the Eocene, switching to nannofossil ooze in the early Oligocene. A fixed hot spot model indicates that the site remained within the equatorial high-productivity zone (within 2° of the paleo-equator) during the E-O interval [Moore et al., 2004; Parés and Moore, 2005].
 Site 1218 has proved to be an excellent reference for E-O climate studies for a variety of reasons: (1) shipboard physical property measurements and the results of downhole logging reveal excellent stratigraphic coverage with no obvious hiatuses through the EOT in the three holes drilled; (2) Site 1218 is critically positioned at a paleo-depth of ∼3800 m and thus captures the ∼1300 m CCD deepening (from ∼3300 to ∼4500 m) associated with the EOT [Lyle et al., 2002; Rea and Lyle, 2005; Coxall et al., 2005]; (3) biostratigraphy, magnetostratigraphy, and cyclostratigraphy provide excellent stratigraphic control that have permitted the construction of an astronomically tuned age framework [Coxall et al., 2005; Pälike et al., 2006]; and (4) from ∼35.2 million years ago onward, sufficient numbers of deep-sea benthic foraminifera are present to allow generation of deep-sea paleoclimate proxy records.
3.1. Samples and Isotopic Analysis
 Twenty cubic centimeter samples were obtained for isotopic and BFAR analysis. These were washed over a 63 μm sieve using water purified by reverse osmosis and dried in an oven at 50°C. Dry weights of the whole samples were recorded to calculate benthic accumulation rates. The coarse fraction >150 μm was examined under a reflected light microscope and benthic foraminifera were isolated. The choice of species was restricted by assemblage composition. Species of Cibicidoides are widely used in paleoclimate studies because they are considered to include epifaunal species and appear to have consistent and predictable offsets from seawater “equilibrium values” [Shackleton and Hall, 1997; Katz et al., 2003a]. The abundance of individual species of Cibicidoides at Site 1218 varies considerably between samples and there were insufficient numbers of a single species in a consistent size range to permit a continuous monospecific record. To standardize analysis we generated records by picking two individuals each of C. grimsdalei, C. havanensis and a third common morphotype Cibicidoides praemundulus (termed C. subspiratus in the work by Coxall et al. ; see Figure S1 in the auxiliary material) from the 150–300 μm size fraction. In a few late Eocene samples, where calcitic microfossils are less abundant, this mix was modified to the specimens available (typically a greater weighting to C. grimsdalei).
 The O. umbonatus record was generated for two reasons: First, to evaluate the fidelity of the published Cibicidoides δ18O and δ13C records and second, as a means to test for changes in δ13C gradients across the sediment water interface that might shed light on deep sea organic carbon cycling across the EOT. O. umbonatus is more abundant than any individual species of Cibicidoides spp. and a monospecific record based on specimens from the 150–250 μm size fraction was constructed. The δ18O and δ13C measurements were made on 304 samples of Cibicidoides spp. and 392 samples of the infaunal species O. umbonatus. O. umbonatus were taken from the same samples used in the study of Coxall et al.  while the Cibicidoides spp. represent additional samples that double the resolution of the Coxall et al.  record (now every 2.5 cm) through the earliest Oligocene.
 Isotopic analysis was performed at the National Oceanography Centre Southampton, University of Southampton, United Kingdom, using a Europa Geo 20-20 mass spectrometer equipped with a “CAPS” automatic carbonate preparation system. Results are reported relative to Vienna Peedee belemnite standard (VPDB). External analytical precision, based on replicated analyses of in-house standards calibrated to NBS-19 is ±0.08‰.
3.2. Site 1218 Eocene-Oligocene Chronology
 One of the primary benefits of the E-O section at 1218 E-O is the excellent age control afforded by the combination of well-defined magnetostratigraphic and biostratigraphic records and clear lithologic and geochemical cycles in cores, downhole logging data and geochemical proxies. Construction of a detailed stratigraphic correlation between Site 1218 and nearby Site 1219, provided opportunities to assess the “completeness” of the records, construct a quasi-continuous composite section and to constrain the position of magnetochron C13n within the XCB-cored interval of Site 1218. The latter was achieved by detailed stratigraphic mapping from Site 1219 (with an error of ±10 cm), where the basal Oligocene magnetochron was recovered, to Site 1218 where it was not. The detailed chronology was generated by matching Site 1218 data with a back-projected astronomical model [Laskar et al., 2004; Pälike et al., 2005; Coxall et al., 2005]. The age model used here represents a refinement of the chronology used by Coxall et al.  and involves an additional “automatic” tuning step [Pälike et al., 2006]. On the Site 1218 time scale the estimated age of the E-O boundary is ∼33.8 Ma (see Text S1 in the auxiliary material).
 On the Site 1218 age model, 5 cm sample spacing of Coxall et al. , provides a maximum resolution of ∼4 kyr. We have doubled the sampling density for the early Oligocene Cibicidoides spp. record, resulting in a theoretical maximum resolution of ∼2 kyr in the early Oligocene where sedimentation rate is ∼1–3 cm kyr−1. While bioturbation acts to place an upper limit on the resolution achievable at sedimentation rates such as these, the higher fidelity of the new epifaunal record compared to that of Coxall et al.  bears testimony to the value of the higher-density sample set (section 4).
 So far there have been no attempts to produce a 3He normalized age model in the Paleogene EEP sediments. It is possible that this approach, which assumes a constant flux of extraterrestrial 3He (3HeET) to the seafloor, might help further constrain fluxes [e.g., Murphy et al., 2010], however, such methods come with their own uncertainties because of poor model calibrations that can overestimate sediment focusing [Lyle et al., 2005]. Moreover, our cycle-based orbital age model yields a more finely resolved record than all previous E-O accumulation rate-based Paleogene paleoproductivity reconstructions and thus goes a long way to minimizing chronological uncertainties at the 40 kyr cycle scale.
3.3. Benthic Foraminifer Accumulation Rate: A Paleoproductivity Proxy
 Changes in paleoproductivity are not straightforward to reconstruct from sediments because of variable preservation of biogenic components and chemical markers at the seafloor. Among the proxies available to Paleogene studies is the accumulation rate of deep-sea benthic foraminifer (BFAR). The BFAR proxy is based on the postulated relationship between surface primary production, organic carbon flux to the deep sea and the number of individual benthic foraminifers deposited [Herguera and Berger, 1991]. It must be applied carefully because taphonomic processes, especially calcite dissolution and bottom water oxygenation also influence assemblage composition [Herguera, 1992; Jorissen et al., 2007]. These considerations have hampered attempts to develop a quantitative proxy of organic matter flux but BFAR appears to correlate predictably with export production in a variety of modern locations [Herguera, 1992; Loubere, 1994; Jorissen et al., 2007] and the method has been used widely as a qualitative/semiquantitative proxy of paleoproductivity in the Pleistocene [e.g., Schmiedl and Mackensen, 1997; Gooday and Rathburn, 1999] and early Cenozoic [Smart, 2008; Alegret and Thomas, 2009]. The approach has been applied to E-O sequences from the Southern Ocean and the equatorial Atlantic [Diester-Haass and Zahn, 1996, 2001; Diester-Haass and Zachos, 2003; Schumacher and Lazarus, 2004] but only at low resolution in comparison to the isotope stratigraphies now available. Critical to accurate reconstructions of BFAR is a robust and detailed age model because changes in sedimentation rate, together with sediment density measurements, act as multipliers, and therefore have a large impact on calculated values of BFAR. The tight age control available at Site 1218 therefore, provides an unprecedented opportunity to apply the BFAR technique to the problem of EOT export production.
 Benthic foraminifera were counted independently in both the >250 μm and the 150–250 μm size fractions in 785 samples taken from the original Coxall et al.  sample set. Two size fractions were counted to assess for preferential dissolution of smaller thinner walled specimens. The >250 μm residues, which were typically <5 cm3, were counted in entirety while the >150 μm samples were typically counted on a split (half) sample. Whole-sample dry weights were used to calculate the number of individuals per gram of dry sediment. BFAR was calculated following the methods of E. Thomas (personal communication, 2008) [after Herguera and Berger, 1991] by multiplying the number of benthic foraminifera per gram by the sample dry bulk density (DBD) and then by the linear sedimentation rate (LSR):
Constraints on DBD were derived by direct measurement on discrete samples and by regression from whole-core analyses of physical properties that are found to be highly correlated with measured values (Figure S2). Minor deviations in the physical properties data calibration may limit power to resolve the highest-frequency cyclic variations in sediment composition, however, the greatest source of uncertainty in BFARs propagate from the age model.
3.4. Cross-Correlation Analysis
 Cross-correlations of Site 1218 geochemical proxy records were carried out to explore lead-lag relationships between glaciation and carbon system changes. This method provides a measure of the similarity of waveforms as a function of a time lag applied to one of them and allows us to more confidently identify potential leads and lags during a fixed interval. The analysis was performed using AnalysSeries 22.214.171.124 [Paillard et al., 1996]. Data were evenly sampled before analysis. The time step used was 2.5 kyr for Cibicidoides spp. and bulk data pairs and 5.0 kyr for O. umbonatus, reflecting the maximum mean sampling resolution of data sets. Analysis was restricted to the second isotope shift of the transition (Figures 2 and 3), because it appears to be complete whereas the base of the first step is likely truncated by dissolution.
4.1. Benthic Foraminifera Assemblages and Preservation
 The most common species in Site 1218 early Oligocene benthic foraminifer assemblages belong to the cosmopolitan, long-ranging genera Cibicidoides, Oridorsalis, Nuttallides, Gyroidinoides, Pullenia and Globocassidulina that are typical of lower bathyl to abyssal water depths above the CCD [Thomas, 1992; Katz et al., 2003b; Takata et al., 2007]. The scarcity of agglutinated species suggests well-oxygenated conditions at the seafloor [Shipboard Scientific Party, 2002]. Detailed assemblage analysis was not part of this study. Takata et al.  observed only minor changes in benthic assemblage composition in their study of early Oligocene sediments from Site 1218, however, their records are of low resolution and do not cover the critical E-O interval. Oligocene benthic foraminifera show signs of mild etching (Figure S1) on their shell surfaces but no evidence of severe corrosion, or shell breakage expected in response to extensive dissolution [Corliss and Honjo, 1981; Nguyen et al., 2009].
4.2. Cibicidoides spp. and Oridorsalis umbonatus Stable Isotope Records
 The most prominent features in the isotope records of Coxall et al. , are a precursor hypothesized dissolution horizon and then two-step increase in δ18O and δ13C close to the E-O boundary with Cibicidoides spp. attaining maximum δ18O and δ13C in the basal Oligocene. The isotope series that we now present (Figure 2) substantiate this basic pattern, but their higher fidelity also brings to light important new structure.
 The two step pattern of benthic δ18O step increase seen in our new infaunal record is very similar to that recorded in our improved epifaunal record with both series describing an initial increase of ∼0.6‰, an intermediate plateau lasting ∼300 kyr, followed by a further increase of ∼0.9‰ in <40 kyr. The base of the first increase in δ18O, here referred to as the EOT-1 for consistency with Katz et al. , is not tightly constrained in our records because of the dissolution event between 34 and 34.1 Ma. The second δ18O step, referred to as Oi-1 and also after Katz et al. , appears to be complete. Using the youngest Eocene samples before the dissolution horizon as a baseline, the overall increase to peak δ18O in the earliest Oligocene has an amplitude of about 1.5‰ for Cibicidoides spp. and 1.6‰ for O. umbonatus equivalent to the global average as measured in different ocean basins (Table 1). The five-point moving averages plotted in Figure 2 show that superimposed on the intermediate plateau are two well-defined cycles of ∼100 kyr duration each in both δ18O series. Sedimentation rates in Site 1218 are low in the late Eocene (∼0.5 cm kyr−1), but our improved epifaunal record captures basic structure on the runup to the EOT with a particularly marked ∼0.4‰ amplitude peak in δ18O centered on ∼34.37 Ma.
Table 1. The Magnitude of Deep Sea δ18O and δ13C Increases During the Eocene-Oligocene Transition Measured in Foraminifera From Locations Across the Globea
Maximum δ18O Shift (‰ VPDB)
Maximum δ13C Shift (‰ VPDB)
All measurements are for benthics except T. ampliapertura, which is a mixed-layer planktonic species.
 Oligocene δ18O reaches maximum values (∼2.2‰ for Cibicidoides spp., 2.5‰ for O. umbonatus) around 33.65 Ma early in magnetochron C13n. The Oligocene record shows elevated δ18O, on average 1.0‰ above the Eocene baseline, with marked, up to 0.5‰, δ18O fluctuations. Phases of elevated δ18O, comprising clusters of cycles are focused early in magnetochron C13n, upper C13n and around 32.68 Ma and this structure is more clearly defined in the higher-resolution epifaunal than in the infaunal record.
 Our new δ13C records also substantiate the two step structure and amplitude of δ13C increase described by the published records of Coxall et al. . The O. umbonatus record is more severely truncated by the precursor dissolution horizon than the epifaunal record making comparison for EOT-1 difficult to assess but the overall amplitude of increase from the late Eocene baseline to peak δ13C in the earliest Oligocene is ∼1.2‰ for Cibicidoides spp. and 1.0‰ for O. umbonatus. Following the abrupt increase in δ13C there is a well-defined relaxation in both δ13C series such that the overall pattern described is a characteristic “overshoot” [cf. Zachos and Kump, 2005]. In the δ18O series the Oligocene baseline is distinctly offset to a new (higher) state than the Eocene baseline but for δ13C, both data series return to near–late Eocene values in the early Oligocene.
 Visual inspection of the isotope records reveals similar overall structure for δ18O and δ13C but there are some differences in phasing of these signals across the transition. Cross-correlation of carbon system proxies with δ18O indicate that both the epifaunal and infaunal δ13C records lag benthic and bulk carbonate δ18O by 10 kyr (±1.25 kyr) on the build up to Oi-1 (Figure 3 and Table 2). In contrast %CaCO3, Cibicidoides spp. δ18O, and bulk sediment δ18O are in phase.
Table 2. Cross-Correlation Statistics for Proxies During the Oi-1 Step of the E-O Isotopic Shift, 33.63–33.7 Ma
Sample Spacing (kyr)
Maximum Correlation Coefficient
Time Lag (kyr)
Revised Cibicidoides δ18O
O. umbonatus δ13C
O. umbonatus δ18O
 An obvious difference between our epifaunal and infaunal records is the large isotopic offsets with O. umbonatus registering higher δ18O (by ∼0.5‰) and lower δ13C (by ∼1.0‰). Calculation of epifuanal-infaunal Δδ13C values during four phases of our Site 1218 Eocene-Oligocene interval reveals a small increase (by ∼0.4‰) in the mean Δδ13C value in samples after the Eocene dissolution event, and virtually no change in infaunal-epifaunal Δδ18O (Figure 4), but there is no obvious correlation between changes in BFAR and Δδ13C or Δδ18O. The calculated offsets between species during the Oligocene, where the sample set is largest (n = 285), based on ordinary least squares linear regression are δ18O Cibicidoides spp. = O. umbonatus −0.34 and δ13C Cibicidoides. spp. = O. umbonatus +1.3 (Figure S3). These values are similar to those reported previously [Shackleton et al., 1984; Katz et al., 2003a; Billups et al., 2004], but they will likely yield over/under “corrections” because of the assumption that the slope equals 1 [Katz et al., 2003a].
4.3. Benthic Foraminifera and Carbonate Accumulation Rates
 Our BFAR records show a similar pattern in the >250 μm and 150–250 μm size fractions (Figure 5). For both, BFAR was low during the late Eocene but increased dramatically with CCD deepening. Early Oligocene BFAR shows a fluctuating pattern with high-frequency variation and an interval of elevated values between 33.4 and 33.8 Ma. The data describe two separate peaks of >3000 BF cm2 kyr−1 in the 150–250 μm fraction and ∼2000 BF cm2 kyr−1 in the >250 μm fraction, falling to <200 BF cm2 kyr−1 in between. These maxima are around 2 to 3 times greater than modal Oligocene values (500–600 BF cm2 kyr−1 for both fractions), with the extremes reaching 4 to 5 times the mode. From ∼33.35 Ma, the BFAR records decrease to a low of ∼100 BF cm2 kyr−1 before increasing to a new baseline of ∼400 BF cm2 kyr−1, fluctuating between 100 and 600 BF cm2 kyr−1. This pattern is interrupted at 32.9–33.1 Ma and 31.4–31.6 Ma, when mean BFAR values increase to modes of about 1200 and 900 BF cm2 kyr−1 in the 150–250 μm and >250 μm fractions, respectively.
 Early Oligocene bulk sediment CaCO3 accumulation rate (CaCO3 AR) increased abruptly at ∼33.70 Ma to a maximum (overshoot) of >1.0 g cm−3 kyr−1 in the basal Oligocene. Elevated values 2 to 3 times the Oligocene base line of ∼0.4 g cm−3 kyr−1 were maintained over two pulses before decreasing at ∼33.36 Ma to a ∼40 kyr long minimum (Figure 5f). The BFAR increase that is coincident with the CCD shift is driven by high accumulation rate estimates. Subsequently, CaCO3 AR increased again, settling into a pattern of strong, high-frequency fluctuations with similar amplitude variation to the earliest Oligocene maximum.
5.1. Structure of the Eocene-Oligocene Glaciation
 Our new Site 1218 climate proxy records provide the most detailed view to date of the structure of the EOT from the World's largest ocean basin, confirming its two-step structure at Site 1218. In Figure 6 we compare our new isotope records with published data sets from other drill sites. This exercise serves to strengthen the case for an EOT precursor late Eocene ice growth event [cf. Katz et al., 2008; Peters et al., 2010; Pusz et al., 2011] but better resolved records are needed from the Late Eocene to determine whether or not this transient represents one in a series of comparatively low-amplitude cycles in δ18O. Regardless, our records show that any glacial advances that might have occurred on Antarctica on the immediate runup to the EOT must have been comparatively small and transient in nature and that large ice sheets were not sustained until the EOT itself. Our records also emphasize that these large ice sheets must have grown extremely rapidly in keeping with the hypothesized role of powerful nonlinear feedbacks in ice sheet mass balance [Weertman, 1976; Oerlemans, 2002; DeConto and Pollard, 2003; Pollard and DeConto, 2005; Coxall et al., 2005]. Our new records reveal more variance during the interval between the two pronounced δ18O steps than seen in the records of Coxall et al.  thereby serving to strengthen correspondence between Site 1218 data and the results of ice sheet coupled climate model experiments [DeConto and Pollard, 2003]. Our results do not, however, strengthen the case for a third δ18O step, indicating that “EOT-2” interpreted from the U.S. Gulf Coast shelf record [Katz et al., 2008] corresponds to an interval of amplitude variation change not seen in the deep sea records.
 Based on δ18O records from the South Atlantic and Southern Ocean [Miller et al., 1991; Zachos et al., 1996] a case has been made for a “zone” of maximum δ18O that represents an early Oligocene glacial maximum. This has been referred to as Oi-1 (also the Early Oligocene Glacial Maximum) [Liu et al., 2004; Coxall and Pearson, 2007], modifying the original concept of Oi-1 as an interval of abrupt δ18O increase (and applied here) as defined by Miller et al.  for lower-resolution records. In fact, the four records presented in Figure 6 lend strong support to the concept of two distinct glacial maxima within magnetochron C13n Oi-1a and Oi-1b [Zachos et al., 1996] and both of these maxima are associated with peaks in δ13C that lag δ18O by about 10 kyr (Figure 7b). Nevertheless, the main conclusion to be drawn from the new Site 1218 record is that the EOT signal cannot really be adequately understood as a sequence of discrete “events,” because it is clearly bears the imprint throughout of strong variability at the 10–100 kyr (orbital) scale (Figure 6).
 Oi-1a and Oi-1b glacial maxima appear to be less well defined in our infaunal than our epifaunal record (Figure 3), a pattern also seen in records from Walvis Ridge, South Atlantic (Zachos et al. , Site 522; Riesselman et al. , Site 1262), although δ13C shows two well-defined maxima. The cause of this is unknown but may reflect a habitat difference.
5.2. CCD Deepening
 Cross correlation analysis confirms that the deep sea and near-surface (bulk) δ18O increase are in lock step with increasing carbonate preservation over Oi-1 (Figure 3). This finding is consistent with the suggestion [Coxall et al., 2005] that CCD deepening was closely associated with Antarctic glaciation. One potential mechanism for explaining the link between the onset of sustained Antarctic glaciation and the basification of the deep Pacific Ocean across the EOT is a shelf-to-basin shift in marine carbonate sedimentation in response to glacioeustatic sea level fall [Coxall et al., 2005]. The results of nutrient capable geochemical box model experiments have lent support to this suggestion [Merico et al., 2008]. In these experiments shelf basin fractionation achieves the observed temporary increase in seawater δ13C and “overshoot” in CCD deepening by a one off “dump” into the global ocean of isotopically “heavy” weathered carbon from newly exposed shelf carbonates while the permanent shift to a deeper CCD state is induced by the requirement of the ocean to maintain the input/output balance of carbonate in response to the reduction in shelf habitat for neritic calcifiers [cf. Zeebe and Westbroek, 2003; Ridgwell and Zeebe, 2005]. This explanation, however, has been called into question on a number of grounds.
 One question concerns the glacioeustatic history of the EOT. The shelf basin fractionation model for explaining EOT CCD deepening has been criticized as being incompatible with the best available sequence stratigraphic reconstructions of sea level change across the EOT [Rea and Lyle, 2005; Miller et al., 2009]. Yet the considerable geophysical uncertainties associated with the sequence stratigraphic method on even “passive” continental margins means that more globally distributed records of this type are needed before we can confidently extract the eustatic signal [Moucha et al., 2008]. Similarly, attempts to separate the ice volume (eustacy) and temperature change components from the benthic δ18O signal using Mg/Ca paleothermometry have met with difficulties because of the competing effects of changes in deep sea [CO3]2− chemistry that appears to mask the temperature effect in Mg/Ca [Lear et al., 2004, 2008, 2010; Lear and Rosenthal, 2006; Coxall et al., 2005; Yu et al., 2007; Pusz et al., 2011]. Where this problem has been circumvented in shallow shelf sequences, results indicate that the overall δ18O signal is attributable to both cooling and ice growth with increasing contribution of ice volume as the EOT progressed [Lear et al., 2008, 2010; Katz et al., 2008; Pearson et al., 2008]. At Site 1218, however, the records suggest tight coupling of δ18O, CaCO3 and lagged δ13C for both EOT-1 and Oi-1, implying that the two δ18O steps share a common origin. Increases in benthic Li/Ca at Site 1218 [Lear and Rosenthal, 2006; Lear et al., 2010] and in the south Atlantic (ODP Site 1263) [Peck et al., 2010], which are interpreted to indicate significant increase in [CO3]2− associated with both steps, is consistent with this hypothesis.
 Our BFAR record provides evidence for increased numbers of benthic foraminifera in the EEP associated with the EOT. Depth-dependent preservation of calcite has undoubtedly influenced the sedimentary record of calcite deposition at Site 1218 and so before we attempt to interpret the BFAR record we must first separate the CCD signal.
 For the most soluble sedimentary carbonate fraction (planktonic foraminifera and calcareous nannofossils), the relationship between the CaCO3 sediment dissolved and that remaining is highly nonlinear such that, when the bulk CaCO3 fraction is large, small percentage loss in bulk CaCO3 indicates substantial CaCO3 dissolution [Berger, 1971; Thunell et al., 1981; Broecker, 1995; Stap et al., 2009]. Oligocene variations in bulk carbonate accumulation at Site 1218, therefore, are likely responses to fluctuations in lysocline depth even once the CCD had deepened [Pälike et al., 2006]. In contrast, the tests of hyaline deep sea benthic foraminifera, which live and calcify at depths as great as 4000 m post-Eocene and are much more refractory than bulk carbonates, typically showing little evidence of dissolution in carbonate-rich sediments [Berger, 1971; Corliss and Honjo, 1981; Thunell et al., 1981; Nguyen et al., 2009]. Thus, while preservation is the dominant control on BFAR in the latest Eocene, once the CCD had deepened, such that bulk sediment CaCO3 at Site 1218 increased to ∼75% and benthic foraminifera show minimal signs of dissolution, we interpret our BFAR record to represent an abundance signal. This interpretation, and >75% criterion, is supported by nannoflora preservation studies [Blaj et al., 2009] that record full Oligocene diversity in samples with >75% CaCO3, and by the similarity in form of the BFAR records generated on both the smaller (150–250 μm) and larger (>250 μm) size fractions.
 Restricting interpretation to the interval of >75% CaCO3, which occurs from ∼33.7 Ma (equivalent to the Oi-1 step) onward, our record suggests a transient (∼500 kyr) twofold to threefold increase in BFAR relative to baseline Oligocene values during magnetochron C13n, in two principal pulses (Figures 5 and 7). We interpret these peaks to represent increased surface biological production and/or increased efficiency of organic carbon export in the EEP associated with early Oligocene Antarctic glaciation. A third more modest increase in BFAR at around 33 Ma suggests a subsequent pulse after peak glacial conditions.
 The δ13C differences between epifaunal and infaunal benthic foraminiferal species have been used to evaluate organic matter remineralization at the seafloor, with high Δδ13C corresponding to high organic matter remineralization rates at times of maximum organic fluxes [e.g., Schmiedl and Mackensen, 2006; Jorissen et al., 2007]. The small (0.4‰) offset to higher Δδ13C in the early Oligocene relative to the late Eocene (Figure 4), therefore, may be a consequence of the early Oligocene EEP productivity changes. In the Eocene, Corg may have been more efficiently recycled above the sediment water interface whereas in the Oligocene, overall more Corg made it into pore waters to be remineralized there. Decreasing ocean temperature associated with the EOT could have had a role in this process [Olivarez Lyle and Lyle, 2006]. The lack of correspondence between Δδ13C and maximum BFAR may be related to the shallow infaunal habitat of O. umbonatus that may not capture the full magnitude of pore water DIC signal. Additional isotope data from a deep infaunal species, and expansion of the assemblage-based data set after Takata et al.  might provide additional insights into this question.
 Records of opal accumulation rate in the Southern Indian Ocean (ODP Site 744) [Salamy and Zachos, 1999] demonstrate a marked increase in the early Oligocene (Figure 7). Controlled partially by an increase in diatom accumulation [Salamy and Zachos, 1999], the increase in opal suggests the initiation of significant seasonal production in the Southern Ocean facilitated by deep mixing and upwelling supplying silica to the photic zone [Salamy and Zachos, 1999; Berger, 2007]. In this context the modest pulse in diatom abundance that occurs in the early Oligocene at Site 1218 and several other equatorial Pacific sites may take on added significance [Fenner, 1984; Baldauf, 1992; Barron et al., 2004; J. Bauldauf, unpublished data, 2010].
 On the other hand, low-resolution barite and CaCO3 accumulation records from several equatorial Pacific ODP Sites, including Site 1218, have been used to argue against increased equatorial productivity during the E-O, instead suggesting a decrease in the rain ratio of organic to inorganic carbon from the late Eocene to early Oligocene [Schumacher and Lazarus, 2004, Griffith et al., 2010]. Interpreting organic export accumulation at E-O transition time, however, is complicated by major changes in the processes of ocean biomass production, transport, preservation and storage, which, like calcite preservation, appear to have undergone fundamental state changes across the EOT. Such changes include the possible influence of decreased ocean temperature on the production and preservation of organic carbon [Olivarez Lyle and Lyle, 2006] and a hypothesized increase in the dependence of deep sea benthos on surface supplied food (development of bentho-pelagic coupling), compared to in situ and laterally transported food, which are considered to have been important for Mesozoic and early Cenozoic benthos [Thomas and Gooday, 1996]. Increases in diversity and mean size of siliceous phytoplankton, especially diatoms [Falkowski et al., 2004; Finkel et al., 2005] at this time, may have also played a role in global patterns of export production and preservation of biogenic remains, although the effects of reduced ballasting associated with a shift toward siliceous primary production at this time is unknown [e.g., Francois et al., 2002; Iversen and Ploug, 2010]. The potential impact of these factors on Corg distribution and preservation in the deep sea equatorial Pacific across the EOT, are still largely unconstrained [Archer et al., 1993; Olivarez Lyle and Lyle, 2006].
 On the balance of evidence it seems that E-O climate change did not trigger a major state shift toward increased production in the EEP, but there is evidence for a short-lived increase in productivity and export production associated with the onset of sustained Antarctic glaciation. Furthermore, the first-order similarities between our BFAR record from the Equatorial Pacific Ocean and the Southern Ocean record of opal accumulation suggests that the productivity increases at the two sites are in some way linked. One potential mechanism to explain this inferred high-to-low-latitude link is through the development of deep mixing in the Southern Ocean, strengthening of ocean ventilation and the development of a common (nutrient/silica-rich) chemistry in waters outcropping at the surface at both Site 744 and 1218, as recognized over more recent times scales [Toggweiler and Carson, 1995; Dugdale et al., 2002; Loubere et al., 2003; Sarmiento et al., 2004]. This is supported by sedimentary and geochemical data that indicates an intensification of thermohaline circulation in the early Oligocene including increased influence of Southern Ocean deepwater in the tropical Pacific [Miller and Tucholke, 1983; Wright and Miller, 1993; Davies et al., 2001; Scher and Martin, 2008; van de Flierdt et al., 2004; Via and Thomas, 2006; Thomas et al., 2008; Cramer et al., 2009; Pusz et al., 2011].
 A second possibility is that the inferred productivity teleconnection between these two sites is driven by increased atmospheric circulation, that is, increased strength of trade winds [Moore et al., 2002; Huber, 2002; Lyle et al., 2008], and a focusing of equatorial sedimentation attributable to changes in the pattern of equatorial divergence [Lyle, 2003; Moore et al., 2004, 2008; Vanden Berg and Jarrard, 2004; Rea and Lyle, 2005]. In reality, as in the modern, productivity variation in the equatorial Pacific is probably controlled by multiple factors [Loubere et al., 2003; Anderson et al., 2009]. Our ability to test these hypotheses is currently limited by the lack of detailed spatial reconstructions of E-O sediment flux patterns and the complications of changes in carbonate preservation associated with CCD deepening.
5.4. The Cause of the δ13C Excursion
 The prominent excursion in δ13C, extending 400–500 kyr into the early Oligocene, indicates a major perturbation of the global carbon system associated with the onset of Antarctic glaciation but the cause of this perturbation, as well as the link with CCD deepening, is debated. Arguably, the most obvious explanation for the >1.0‰ δ13C increase is a global increase in export and burial of organic carbon in the deep sea. Our BFAR records, together with records of opal burial in the Southern Ocean provide some support for a productivity increase (section 5.2). However, our deep-sea records appear to be symptomatic of Corg export to the seafloor but not carbon sequestration (burial) because Paleogene EEP sediments exhibit Corg contents of <0.1 wt% (an order of magnitude lower than Holocene EEP values) [Lyle et al., 2002; Olivarez Lyle and Lyle, 2006; Pälike et al., 2009]. Elsewhere, for example, the Paratethys, southern Indian and west African margins, Corg exported was not remineralized but buried [Schulz et al., 2002; Sarkar et al., 2003; Ravizza and Paquay, 2008]. The “missing carbon,” therefore, may be accounted for in regions such as these that are close to continents, and where 90% of Corg burial takes place today [Berner and Canfield, 1989; Hedges and Keil, 1995]. Regardless, the processes that result in such low Corg in the Equatorial Pacific require further study.
 A second problem associated with the Corg burial hypothesis is that, although geochemical modeling studies can simulate the δ13C excursion by increased organic carbon burial, thus far they have been unable to simulate the observed CCD response by the same mechanism [Zachos and Kump, 2005; Merico et al., 2008]. In contrast, simulations forced by sea level–led increases in weathering inputs of isotopically heavy neritic limestone reproduced both a permanent CCD deepening and a transient δ13C perturbation over comparable time scales to that observed [Merico et al., 2008]. That said, the way in which the shelf basin fractionation hypothesis is instantiated into the modeling of Merico et al.  needs to be improved to consider Corg (not just CaCO3) fractionation.
 One critical question to establish is the extent to which δ13C increase and CCD deepening are forced by the same mechanism. Our detailed records reveal cycle-level similarities in the form of δ13C, δ18O and %CaCO3 but there they are not synchronous. The ∼10 kyr lag of δ13C with respect to δ18O and %CaCO3 over Oi-1 (Figure 5), which has been identified at several sites [Zachos et al., 1996; Riesselman et al., 2007], implies an indirect or delayed response of the driver of δ13C change. One explanation is that δ13C and δ18O were forced by the same climatic mechanism, for example, ice sheet growth (shelf to basin transfer), but the long residence time of carbon in the ocean sediment reservoir (∼100 kyr) [Broecker and Peng, 1982] created a delay in benthic δ13C change [Pälike et al., 2006]. Alternatively, if the δ13C increase was associated with changes in biomass production and burial the delay might signal a “nutrient lag” reflecting shelf to basin delivery of biologically important nutrients such as phosphorous [Broecker, 1982; Filippelli et al., 2007]. Moreover, the carbon system likely responded to different periodic forcing mechanisms at different times during the EOT, because δ13C and δ18O appear to be synchronous over the highest-frequency cycles during the E-O intermediate plateau (Figure 2), yet the spectral analyses of Coxall et al.  show that δ13C leads δ18O (by ∼8 kyr) in the 40 kyr band from ∼33.5 Ma onward.
5.5. Eocene-Oligocene Glaciation and Atmospheric Carbon Dioxide
 The most recently published pCO2 proxy record for the EOT, which is based on boron isotopes in well-preserved planktonic foraminifera from Tanzania [Pearson et al., 2009], is consistent with the basic idea that initiation of Antarctic glaciation was associated with pCO2 decline during the late Eocene (Figure 7). This decline (∼300 ppm from 34.5 Ma to 33.7 Ma), which is essentially constrained by a single data point, is followed by a temporary reversal to late Eocene values roughly in step with rising δ18O. pCO2 subsequently decreased to a new lower Oligocene baseline after peak glacial conditions were initiated, a pattern implied by longer-term records [Pearson and Palmer, 2000; Pagani et al., 2005].
 The data suggest that Antarctic glaciation was closely associated with a temporary increase in pCO2. Age control in the Tanzanian records is weak in comparison to our records but if the structure and timing of this pCO2 increase is substantiated and shown to be globally representative then it is fundamentally inconsistent with a Corg-burial-led explanation for δ13C increase. There is some suggestion based on nannofossil that the lower Oligocene Tanzanian sequence records evidence of eutrophication [Dunkley Jones et al., 2008] suggesting that the site is unlikely to have been at air-sea equilibrium with respect to CO2. Yet, the magnitude of local ocean-atmosphere disequilibrium required to explain the observed signal in this way is large. One alternative to consider is that the CO2 increase was driven by ocean “venting” if, as proxies suggest, overturning circulation increased during the early Oligocene, a mechanism that (1) has been proposed to help explain increases in pCO2 at the end of the last glacial cycle [Anderson et al., 2009; Skinner et al., 2010] and (2) yields a short-lived ventilation spike in CO2 in the model results of Merico et al. . The residence time of CO2 in the subsurface ocean, however, is short (∼1000 years) compared to geological time scales and is unlikely to have caused a 300–400 kyr long pCO2 excursion suggested by the proxy record. Other possible explanations for a CO2 increase at the EOT are a reduction in silicate weathering [Broecker and Peng, 1982; Zachos et al., 1999; Zachos and Kump, 2005; Lear et al., 2004; Vance et al., 2009], or a release of organic carbon from terrestrial or continental shelf reservoirs [Pearson et al., 2009].
 We present new deep sea paleoceanographic proxy records from eastern equatorial Pacific ODP Site 1218 that provide insight into the role of the ocean carbon system in Eocene-Oligocene climate change.
 1. Our new isotope records support a rapid two-step structure of δ18O increase across the EOT described in the records of Coxall et al.  but reveal more variability in the intervening “plateau” interval. This result serves to only increase the convergence between the Site 1218 data series and the model reconstructions of DeConto and Pollard , consistent with threshold crossing behavior and the operation of powerful positive feedbacks associated with ice sheet height and coalescence.
 2. Integration with global deep-sea records provides new support for the hypothesized existence of a precursor Late Eocene δ18O increase event [Katz et al., 2008] on the runup to the onset of full blown Antarctic glaciation and for two early Oligocene glacial maxima (Oi-1a and Oi-1b) [Zachos et al., 1996]. Nevertheless, the improved fidelity afforded by our new records emphasizes that the EOT cannot be adequately understood as a series of discrete events because it is clearly imprinted by orbitally paced variability throughout.
 3. Our record of BFAR record is the best resolved data set of its type available for this time interval and unique in being the first to include the effects of orbital-scale changes in sedimentation rate and sediment density. BFAR appears to reveal temporary (∼500 kyr), pulsed, increase in export production in the EEP associated with maximum Oligocene glacial conditions. First-order similarities in structure to Southern Ocean opal records suggests strengthening of a connection between high and low latitudes through ocean circulation, effecting chemistry of water upwelling to the EEP thermocline. Intensification of equatorial winds associated with E-O climate change could also have played a role.
 4. The offset in δ13C between our epifaunal and shallow infaunal benthic foraminifera are higher in the early Oligocene than late Eocene (by ∼0.4‰), consistent with an overall increase in sediment flux and seafloor remineralization but there is no correspondence to individual BFAR peaks, perhaps reflecting the shallow infaunal habitat of O. umbonatus.
 5. The inferred increase in biological pumping and well-documented increase in benthic δ13C are not readily reconciled with the timing of the transient pCO2 rise observed in a recent proxy reconstruction of the E-O transition from Tanzania [Pearson et al., 2009].
 This research used samples provided by the Ocean Drilling Program (ODP). Funding was provided by the UK (I)ODP and Natural Environmental Research Council grants to P.A.W and a Royal Society University Research Fellowship to H.K.C. We thank Helen Farhall and Heather Birch for laboratory assistance and Mike Bolshaw for analytical support. Thanks also to Jim Zachos, Paul Pearson, and Steve Bohaty for sharing data and providing insight into site-to-site age correlation; to Chris Smart and Ellen Thomas for help with benthic foraminifera taxonomy; and to Ted Moore, Jan Backman, and Jack Baldauf for useful discussion. We thank Greg Ravizza and Mimi Katz for providing insightful reviews.