Small but significant differences exist among stable carbon and oxygen isotopic excursions measured in coccolith-dominated bulk carbonate and planktic foraminifera during the Paleocene-Eocene thermal maximum (PETM). One hypothesis suggests that the bulk carbonate isotopic record is compromised by changing nannofossil assemblages, since modern nannofossils show a large (5 permil) range of interspecific vital effects. New techniques are employed here to separate different size fractions of coccoliths from PETM sediments at ODP Site 690 for isotopic analysis, removing a major portion of the variation in nannofossil assemblages. Isotopic compositions of coarse and fine coccolith fractions dominated by coccoliths of genus Chiasmolithus and Toweius, respectively, differ by less than 0.5 permil for both oxygen and carbon. The near-monogeneric Toweius record closely parallels the main trends in the bulk carbonate isotope records, including multiple steps in the negative carbon isotopic excursion, suggesting that the trends in the bulk carbonate record are not artifacts of changing species assemblages. Because both coccolithophorids and symbiont-bearing foraminifera like Acarinina must inhabit the photic zone, it is unlikely that the 103 year lags in isotope event onset between coccoliths and Acarinina reflect true time-transgressive invasion of isotopically depleted CO2 into the water column. The small range of vital effects among Paleocene coccoliths is unlikely to result from diagenetic homogenization, and instead may reflect more similar carbon acquisition strategies of Paleocene coccolithophorid algae due to larger and/or more similar cell sizes and higher atmospheric carbon dioxide. The small range of vital effects suggests that bulk carbonate records are likely reliable for other early and pre-Cenozoic sediments where foraminifera are often scarce.
 Paleoceanographic reconstructions of surface ocean conditions from stable oxygen and carbon isotopic measurements have long relied on analysis of planktonic foraminifera because individual specimens can be individually picked and most species have a small and limited range of isotopic vital effects. However, in sediments where planktonic foraminifera are sparse or very high sample resolution is desired over long intervals, bulk carbonate has also been used for stable isotopic analysis. Depending on the age and setting of sediments, coccoliths may contribute the great majority (>90%) of the carbonate. Because coccoliths grown in culture experiments exhibit a nearly 5 permil array of interspecific vital effects in oxygen isotopes [Dudley and Goodney, 1979; Dudley et al., 1986] and in carbon isotopes [Ziveri et al., 2003], changes in coccolith species assemblages could introduce significant artifacts in bulk carbonate isotopic records in sediments. Bulk carbonate isotopic records can faithfully reproduce trends in single species foraminiferal isotopic records over Quaternary glacial cycles [Shackleton et al., 1993] where there are no major shifts in species assemblages. However, there are often major changes in the nannofossil assemblages over longer time intervals or during major oceanographic or climatic events and the possible influence of such assemblage changes cannot be constrained for extinct species for which the vital effect offsets are unknown. Consequently, the validity of bulk carbonate isotopic results has been called into question for many such records.
 However, differences in the details of the isotopic records have important implications for the source and triggering of the isotopically light carbon addition as well as the nature of potential feedbacks on methane hydrate release. In coccolith-dominated bulk carbonate isotopic records from ODP Sites 690 and 1051, the carbon isotopic excursion occurs in three stepwise δ13C decreases over the course of 60 cm, with intervening plateaus in which δ13C values level off [Bains et al., 1999] (Figure 1a). These plateaus were interpreted as periods of stasis in the carbon cycle between multiple discrete injections of isotopically light carbon [Bains et al., 1999]. In contrast, single specimen planktonic foraminiferal records from ODP Site 690 show no intermediate carbon isotopic compositions during the excursion, and are interpreted to reflect a single, instantaneous addition of isotopically light carbon to the system [Thomas et al., 2002] (Figure 1a).
 Bulk carbonate and single specimen foraminiferal oxygen isotope profiles also yield conflicting records of high latitude temperatures just prior to the CIE. In ODP Site 690, bulk carbonate oxygen isotope compositions shift toward heavier values in the ∼10,000 years leading up to the CIE, interpreted by Bains et al.  to reflect a brief cooling episode (Figure 1b). This positive shift is not apparent in the single specimen foraminiferal oxygen isotopic record, which instead shows a slight decrease in δ18O in some individual foraminifera just prior to the carbon isotopic excursion, interpreted as a slight warming in the ∼1,000 years prior to the CIE [Thomas et al., 2002] (Figure 1b). These temperature changes just prior to the PETM are within the range of longer term variability [Kennett and Stott, 1991], but have been invoked as potential triggers for methane dissociation [e.g., Bains et al., 1999; Thomas et al., 2002].
 Because the PETM is characterized by major shifts in the assemblages of calcareous nannofossils at ODP Site 690 (Figure 1c), Bralower  and Thomas et al.  suggest that the details of the bulk carbonate record, including the plateaus in carbon isotopes, are artifacts of changing coccolith assemblages over this interval.
 The current study employs a new technique to separate different size fractions of coccoliths from PETM sediments at ODP Site 690 for isotopic analysis. This approach removes a major portion of the variation in nannofossil assemblages. Consequently, it is possible to assess the extent to which the bulk carbonate isotopic records are biased by the assemblage changes and whether there are large magnitude interspecific vital effects among the dominant Paleocene nannofossil species in this record. This comparison allows us to more precisely resolve some of the oceanographic changes during the PETM. More generally, it provides a perspective on the fidelity of bulk carbonate records for other early and pre-Cenozoic sediments where foraminifera are often scarce.
 Twenty-four samples were processed from core depths of 173.8 to 167.9 mbsf in ODP Hole 690B from the Weddell Sea, the most expanded and complete PETM record examined to date. Latest Paleocene and earliest Eocene sediments at ODP Site 690 are composed of carbonate-rich (70–90% CaCO3) nannofossil ooze.
 For each sample, compositionally distinct coccolith subfractions were obtained by a combination of microfiltering [Minoletti et al., 2001] and repeated decanting methods with settling times modified from Stoll and Ziveri  to adjust for the larger coccoliths in the Paleocene material. Initial test separations of coccoliths into coarse, intermediate, and fine fractions showed that the intermediate size fraction was a mixture of many species and for subsequent samples primarily coarse and fine fractions were recovered for analysis. Nannofossil assemblages in all fractions were counted by scanning electron microscopy and light microscopy. Counts and size measurements were converted into volume and mass contribution of each species using the approach of Young and Ziveri , with shape factors derived from volume rotations of cross sectional sketches of Romein  using the NIHImage macro written by Young and Ziveri . The estimation of the shape factor is the greatest source of uncertainty in these calculations, but any biases in estimated volume contributions are consistent from sample to sample.
 Stable isotopic measurements were made on approximately 150 μg samples reacted with 100% H3PO4 at 90°C in a Multiprep device with evolved CO2 analyzed on a Micromass Optima gas source mass spectrometer at the University of Albany, State University of New York. Duplicate samples had a precision (1s) of 0.05 permil for oxygen and 0.02 permil for carbon.
 Nannofossil assemblage counts indicate that while the separated fractions are not purely monogeneric, there are fewer variations in their species makeup than in the bulk carbonate. In the fine coccolith fraction most carbonate is from Toweius (typically 70–90% of the carbonate) with minor carbonate contributed by Chiasmolithus, Zygrhablithus and Sphenolithus throughout the record and additionally by Fasciculithus during the PETM (Figure 2a). In the coarse fraction, Chiasmolithus dominates in the lower part of the record, Discoaster becomes a major contributor in the middle part of the record, and Zygrhablithus becomes a major contributor in the upper part of the record (Figure 2b). For some intervals, additional intermediate size fractions were recovered; these are mixtures of more even contributions by a larger number of species and higher abundances of Fasciculithus (Figures 2c and 2d).
 Although the species makeup of the fine and coarse fractions are significantly different, these fractions have surprisingly similar carbon and oxygen isotopic values that typically differ by less than 0.5 permil (Figures 3a and 3b). In both fractions the coccolith isotopic records are quite similar in trends and absolute values to the bulk carbonate records published by Bains et al. .
 Bulk δ13C values generally fall in between those of the coarse and fine coccolith fractions, as expected by mass balance since the bulk carbonate is a mixture of both. The Toweius-dominated fine fraction has δ13C that is slightly lower than that of the bulk carbonate throughout the record. Toweius δ13C is also lower than the coarse coccolith fraction regardless of whether the coarse fraction is dominated by Chiasmolithus (lower part of the record) or Zygrhablithus and Chiasmolithus (upper part of the record). Both the Toweius-dominated record and the coarser coccolith record replicate the plateaus in carbon isotopic composition present in the bulk isotopic record.
 Coccolith δ18O values are also similar to those of bulk carbonates although there are minor differences in trends. The δ18O values of the Toweius-dominated fraction are similar to those of the bulk carbonate throughout the record. When the coarse coccolith fraction is dominated by Chiasmolithus or by Chiasmolithus and Discoaster, its δ18O values are lower than both the bulk carbonate and Toweius fraction. In contrast, when the coarse coccolith fraction is dominated by Chiasmolithus and Zygrhablithus, δ18O values are comparable to or higher than those of the Toweius fraction. Although the resolution of the separated coccolith record is much lower than that of the bulk record, both the coarse and fine coccolith fractions show a slightly longer duration warm period with a steeper recovery than the bulk carbonate record. Neither the coarse or fine coccolith fractions show an increase in δ18O in the 50 cm preceding the CIE, as is observed in the bulk carbonate record.
 In several aspects, the isotopic records of the coccoliths and bulk carbonate are more similar to those of single and multispecimen analyses of the foraminifer Subbotina than to those of the foraminifera Acarinina [Thomas et al., 2002; Kennett and Stott, 1991] (Figure 3). The major negative shifts in δ13C and δ18O in the bulk carbonate and coccoliths at 170.66 cm coincide with the negative δ13C and δ18O shifts in Subbotina and both lag the δ13C and δ18O shifts observed in foraminifera Acarinina by about 10 cm. The long duration of negative δ13C values in the coarse and fine coccolith records is most similar to that of Subbotina. In both coccolith records and Subbotina single specimen records, δ18O values reach their minimum values by 170.66 m, whereas the Acarinina records show a second drop in δ18O at 170.25 m. Absolute δ13C values are also similar between the Toweius coccolith fraction and the multispecimen Subbotina record. However, neither foraminiferal record shows evidence for intermediate plateaus in δ13C seen in the coccolith and bulk records. At the resolution examined, there are no coccolith analyses showing intermediate δ18O values prior to intermediate δ13C values as observed in single specimen Acarinina analyses.
4.1. Do Interspecific Vital Effects Influence the Isotopic Trends in Bulk Carbonate Records?
 The similar isotopic values in the coarse and fine fractions with significantly different species composition, suggest a limited range of interspecific isotopic vital effects in the major Paleocene coccolith species. While the shifts in assemblage composition in separated fractions coincide with changes in the carbon isotopic ratios in the bulk carbonate, these species shifts cannot cause the principal features of the isotopic variations. For example, Bralower  and Thomas et al  suggest that the plateaus in the coccolith record are artifacts of changing species assemblages and that the true isotopic shift is a single instantaneous δ13C drop. However, for the first step in the separated fine coccolith carbon isotopic record to be an artifact of the increased abundance of Fasciculithus (reaching 20% by mass), the δ13C of Fasciculithus would need to be about 4 permil higher than the composite δ13C of the other species in the fine coccolith fraction. However, the similar isotopic ratios for the multiple separated fractions from 170.45 (Figure 2c) constrain the isotopic composition of Fasciculithus to be quite similar to that of the other species. In fractions from 170.45, about 75% of the intermediate fraction species makeup can be simplified as a 50/50 mixture of the species in the fine (3–5 μm) and coarse (8–12 μm) fractions, with the remaining 25% comprised of extra Fasciculithus not predicted by a simple mixture. The isotopic composition of the intermediate fraction is thus 75% defined by the average of that of the coarse and fine fractions. By mass balance, the δ13C of Fasciculithus would be about 0.1 permil lighter than that of the other species and the δ18O of Fasciculithus about half a permil heavier than that of the other species. The one caveat in assumptions of this calculation is that the 50/50 mixture of fine and coarse fractions slightly overestimates the abundance of Discoaster in the intermediate size fraction. If Discoaster had an isotopic composition which differed from the composite of the other species in the opposite sense from Fasciculithus, a larger difference in isotopic composition between Fasciculithus and other assemblages might be obscured. However, the coarse and fine fraction time series records taken together show the same δ13C step offsets even though the coarse fraction shifts significantly to Discoaster dominance with the onset of the PETM while the major change in the fine fraction is a shift to Fasciculithus. This similarity is not consistent with Discoaster and Fasciculithus differing from the other species in contrasting ways.
 Consequently, the similarity of stable isotopic values from coccolith fractions with significantly different species makeup, the coherency of isotopic trends among distinct coccolith fractions and bulk carbonate, and the detailed mass balance comparisons in specific intervals, indicate that the major features of the bulk carbonate record cannot be attributed to artifacts of interspecific coccolith vital effects with changing species assemblages. Other processes must account for the later onset of the δ13C excursion in bulk (coccolith) records compared to single specimen Acarinina, and for the stepped nature of the bulk (coccolith) δ13C excursion compared to a single excursion in Acarinina and Subbotina.
4.2. Coccolithophorids in the Planktic Succession
 While both photosymbiont-bearing Acarinina and coccolithophorids must dwell within the photic zone, they need not occupy the same seasonal or vertical niche. In fact, the similarity of isotopic trends in the thermocline-dwelling Subbotina and the photic-zone dwelling coccolithophorids suggests a greater similarity in habitat among coccolithophorids and Subbotina. If the mixed layer was very shallow the pycnocline may have been within the photic zone and coccolithophorids and Subbotina may have occupied the same habitat. Such a situation would be analogous to the modern Southern Ocean where maximum carbonate fluxes occur during the light-induced summer bloom event (January–February) under highly stratified conditions with a shallow mixed layer [Honjo et al., 2000]. Alternatively, the similarity in isotope values among Subbotina and coccolithophorids may have arisen because maximum coccolithophorid production occurred during seasonal convective events (very deep mixed layer) in a more homogeneous water column. This would also explain the fact that coccoliths have carbon isotopic ratios which are lower than Acarinina but oxygen isotopic ratios that are higher than Acarinina. Thus coccolith results should be interpreted as reflecting a photic zone signal which may or may not be identical to a mixed layer signal or annual average photic zone signal.
4.3. Onset and Rate of the Isotopic Excursions
4.3.1. Why a Different Onset for Carbon Isotopic Excursions?
Thomas et al.  proposed that the lead in the δ13C excursion in the symbiont-bearing photic zone dweller Acarinina relative to that of the thermocline dweller Subbotina represents the actual time transgressive invasion of δ13C-depleted CO2 from the atmosphere down through the photic zone and into the thermocline waters. According to the most recent age models for ODP Site 690 [Farley and Eltgroth, 2003; Rohl et al., 2000], the 8–12 cm lead of Acarinina δ13C excursion would imply a 9–13 kyr lead. The Acarinina excursion also leads the coccolith-dominated bulk carbonate signal. Regardless of whether the coccolithophorid signal represents a deeper part of the photic zone or conditions during seasonal convective events, and even if thermohaline circulation slowed during the PETM, it is not possible that ∼10 kyr would be required for “top down” mixing of an atmospheric δ13C signal down through the photic zone and into the thermocline. Given seasonal mixing timescales of the photic zone, neither does it seem plausible that the age models are so in error that the open ocean pelagic site would have had accumulations of 8–10 cm/yr during this particular time window in this event.
 Differential bioturbation could potentially affect the timing of events in different sediment components like coccoliths and foraminifera. Bard  showed that for sediments with an accumulation rate of 1–2 cm/kyr (comparable to those during the onset of the CIE), deeper bioturbation of fine particles can shift the event onset in the fine (coccolith) fraction back in time by 0.5 to 3 kyr relative to the onset in the coarse (foraminiferal) fraction. However, this mechanism is the opposite sense for what is observed in the bulk-coccolith and Acarinina records from PETM sediments at 690, and cannot explain offsets among like-sized foraminifera Acarinina and Subbotina. Consequently, alternative explanations for this offset must be evaluated.
 Bioturbation might contribute to shifts in the isotopic signal is if there were drastically different preservation potential of different shells under corrosive conditions. The onset of the Acarinina δ13C excursion coincides with the minimum in CaCO3 content [Rohl et al., 2000] and drop in CaCO3 mass accumulation rates [Farley and Eltgroth, 2003] and the appearance of heavily calcified acarinids [Kelly, 2002], all of which may be expressions of a dissolution event triggered by the release of methane to the ocean-atmosphere system [e.g., Dickens, 2000]. Models of the effect of methane hydrate release predict that this dissolution event should affect not only synchronous sediments but should “burn down” into sediment predating the methane hydrate release [e.g., Dickens, 2000]. Consequently, if the onset of the drop in CaCO3 content and accumulation reflects the base of the burn down event, it should predate the true timing of the δ13C excursion. Such a phase relationship is found in sediments in ODP 1209B, where increased foraminiferal fragmentation from dissolution begins about 6 cm below the carbon isotope excursion [Zachos et al., 2003]. If the preexcursion Acarinina shells were extremely susceptible to dissolution and were extensively removed during the burn down, then post excursion Acarinina could have been bioturbated down into the burn down sediments and might be the only phase there for analysis, effectively shifting back the onset of the Acarinina δ13C excursion. This explanation would require that the majority of preexcursion coccoliths and Subbotina shells were resistant and survived burn down so that their signal was not shifted. Given that coccoliths contribute the great majority of carbonate in these sediments, it is possible that proportionally they might experience more complete preservation. However, among the foraminifera, Acarinina comprised 75% of planktic assemblages before the Acarinina δ13C decrease and still comprised 50% during the 10 cm interval between the onset of Acarinina and Subbotina/coccolith isotopic excursion [Kelly, 2002]. It is hard to imagine such a large proportion of the Acarinina foraminiferal population being reworked while the Subbotina (comprising the other 50% of the foraminiferal population) would be in situ. Likewise, it seems unlikely that all of the fifty eight individuals of Acarinina picked for analysis in the sample intervals between 170.77 and 170.64 mbsf actually represent postexcursion samples mixed down into the burn down zone in which a comparable number of Subbotina in that same depth range represent in situ individuals. These concerns are particularly relevant if the number of foraminifera in the sediment is the same in preburndown, burndown, and excursion sediments.
 Intraspecific vital effects on isotopic fractionation are known from foraminifera but such effects like the carbonate ion effect are small and have relatively comparable effects among different species [Spero et al., 1997]. Intraspecific vital effects on isotopic explanation are less well studied in coccoliths. Major effect attributable to light- or temperature-induced growth rates have been ruled out [Ziveri et al., 2003]. The similarity of coccolith isotopic trends to those of foraminifera Subbotina also suggests that the major coccolith trends record primary changes in seawater isotopic ratios and temperature. Furthermore, the shift in δ13C of surface carbon reservoirs must be recorded in all foraminifera and coccoliths so it seems highly unlikely that a comparable magnitude major shift in δ13C would happen due to vital effects and that an equal and opposite major shift in δ13C could obscure the true shift in δ13C of surface dissolved inorganic carbon.
 Consequently, given available constraints on age models and sedimentology, neither true time lags in the invasion of isotopically depleted CO2 nor sedimentological processes readily explain the lag between carbon isotopic excursions in the Acarinina record relative to Subbotina and bulk carbonate records. It will be important to ascertain whether similar lags are present among symbiont bearing foraminifera, nonsymbiont bearing foraminifera, and coccolith-dominated bulk carbonate at other PETM sections.
4.3.2. Why Plateaus in the Coccolith/Bulk Carbonate δ13C Excursion?
 The persistence of plateaus in the δ13C excursion from near-monogeneric coccolith fractions rules out the explanation that that the plateaus are artifacts of interspecific vital effects from changing nannofossil assemblages in the bulk carbonate [Bralower, 2002; Thomas et al., 2002]. The other two explanations provided to explain the stepped nature of the carbon isotopic excursion, multiple releases of isotopically negative carbon to the ocean-atmosphere system [Bains et al., 1999] or superposition of a background precessional cyclicity in the δ13C of dissolved inorganic carbon on a monotonic PETM excursion, imply global changes in the carbon isotopic composition of dissolved organic carbon which should affect foraminifera as well as coccoliths.
 For either of these latter two explanations to be correct, some process must preclude the recording of these higher resolution variations in the foraminiferal record. One possible explanation is that foraminifera are too scarce a fraction of the sediments to record the brief durations of the intermediate plateaus. Bulk carbonate and coccolith isotopic records are capable of capturing high resolution carbon isotopic variations. Paleocene-Eocene orbital scale rhythms in bulk carbonate δ13C are reproduced at multiple sites [Cramer et al., 2003] and the coccolith PETM records provided here strongly suggest that these rhythms are true oscillations in the carbon cycle and not artifacts of changing coccolith assemblages. However, the resolution of existing foraminiferal records at high accumulation rate Site 690 is not yet sufficient to identify this type of background cyclicity in δ13C. One test of the resolution of the foraminiferal record would be to evaluate whether the pre-PETM bulk carbonate δ13C cyclicity could be replicated in multiple specimen foraminiferal records in Site 690. If the foraminifera replicate this cyclicity, then a foraminiferal record at comparable resolution would be capable of resolving plateaus at the PETM, as long as the abundance of foraminifera in the sediment did not drop significantly during the δ13C excursion.
 An alternative explanation is that the plateaus in the bulk carbonate and coccolith record during the PETM are artifacts of mixing of different age material, whereas single specimen foraminiferal records do not contain such mixing. However, it seems unlikely that the dominant carbonate component could be mixed so completely on such long length scales (24 cm from preexcursion values to the first δ13C minimum and 65 cm for the full magnitude of the excursion). The apparent coherence of variations in planktic foraminiferal assemblages with the series of plateaus in the bulk and coccolith isotopic records also suggests that differential mixing has not decoupled the foraminiferal and coccolith signals on this scale [Kelly, 2002] and that there is some environmental significance to these plateau features.
 A final possibility is that the plateaus in the coccolith/bulk carbonate δ13C excursion arise from vital effects within a given coccolith species. In this case, if the δ13C variation in dissolved organic carbon were a single instantaneous decrease, then the coccolith isotopic record would require intraspecific vital effects that initially enriched the coccolith carbon isotopic composition relative to equilibrium, with the amount of this vital effect successively decreasing. While intraspecific vital effects on coccolith stable isotopes have not been extensively investigated, Ziveri et al.  hypothesized that the large range of interspecific coccolith vital effects might result from different strategies of active and passive carbon acquisition by different species of coccolithophorids. It is possible that the large increases in pCO2 and perhaps decreased ocean pH accompanying the initial methane hydrate release could have altered the efficiency or strategies of carbon acquisition of individual taxa and hence the magnitude of isotopic vital effects. The existence or magnitude of such an intraspecific CO2 vital effect has not been examined in modern coccolithophorids. On the basis of analogy with interspecific relationships where species with high surface area/volume ratios have highest isotopic values, enhanced CO2 availability might be expected to produce more enriched isotopic values. However, if the plateaus in bulk carbonate δ13C reflected such a vital effect, it is surprising that the effect is more short lived than the warming which is presumably also a symptom of elevated CO2 [Zachos et al., 2003].
4.3.3. Gradients in Carbon Isotopic Values
 Following the negative δ13C excursion, δ13C values of Acarinina recover from their excursion minimum starting at 170.26 cm, 66 cm below the initial recovery in Subbotina or bulk carbonate (coccolith) δ13C values. Kennett and Stott  and Thomas et al.  interpreted this initial rise as the restoration of the gradient in water column δ13C of dissolved inorganic carbon as productivity and organic carbon export were restored and local thermal and chemical stratification were reestablished. The timing of the rise in Acarinina δ13C values does correspond to the start of a peak in productivity inferred from Sr/Ca of Toweius coccoliths [Stoll and Bains, 2003]. Since the δ13C gradient between coccolithophorids and Acarinina also increases although both must inhabit the photic zone, the steepest part of the δ13C gradient must be between the depth habitat of Acarinina and that of coccolithophorids, or there must be a large seasonal contrast in surface δ13C. A part of the magnitude of the Acarinina excursion, which exceeds the magnitude of the excursion in Subbotina and coccolithophorids, may be due to the reduction in or loss of the water column or seasonal δ13C gradient at the onset of the PETM.
4.3.4. Onset and Timing of Oxygen Isotopic Variation
 Since there is no evidence for a rise in δ18O of coccoliths from either fraction prior to the CIE, the coccolith record provides no support for a high latitude cooling prior to the CIE onset [e.g., Bains et al., 1999]. Neither does the coccolith record support any pre-PETM high latitude warming as was interpreted from the Acarinina isotopic record. Instead, like the Subbotina record, the coccolith and bulk carbonate record indicates simultaneous change in temperature and the carbon cycle. Although warming may have occurred in the surface habitat of Acarinina [Thomas et al., 2002] there is no evidence that such warming propagated through the rest of the photic zone and thermocline and hence affected the bathyal depths of the ocean where methane hydrates occur. Consequently, these results do not provide support for the thermal dissociation mechanism for PETM triggering.
4.4. Origin of Interspecific Isotopic Variations in Paleocene Coccoliths
 Different isotopic compositions in different species of coccoliths may arise from different depth habitats within the water column or from species-specific vital effects in isotopic fractionation. If differences among isotopic values in different species were due solely to differing depth habitats, then we would expect deeper dwelling species to have higher oxygen isotopic ratios typical of colder waters but lower carbon isotopic ratios typical of lower δ13C of dissolved organic carbon in deeper waters. In contrast, interspecific vital effects in modern coccolithophorids tend to produce positively correlated fractionations in carbon and oxygen isotopes in coccoliths [e.g., Paull and Thierstein, 1987; Ziveri et al., 2003] with increasing isotopic enrichment with smaller cell diameters [Ziveri et al., 2003].
 In the lower sediments of ODP Site 690 where the coarse and fine coccolith fractions are dominated by Chiasmolithus and Toweius, respectively, the isotopic signals (lower δ13C, higher δ18O) of Toweius would be consistent with a deeper depth habitat for Toweius compared with Chiasmolithus. However, a portion of the higher δ18O of Toweius could potentially arise from a smaller coccosphere size of Toweius compared to Chiasmolithus if cell size and coccolith size are correlated. In the upper part of the record where the coarse coccolith fraction is dominated by Chiasmolithus and Zygrhablithus, the coarse fraction is enriched in both δ13C and δ18O relative to Toweius which may be more consistent with interspecific vital effects than depth stratification.
 The relatively small differences in isotopic compositions between different coccolith fractions are surprising given large differences in isotopic composition of modern species grown in culture. The small isotopic differences among different Paleocene species is unlikely to reflect homogenization of coccolith chemistry from diagenetic overgrowths because different species of coccoliths preserve a very high degree of minor element heterogeneity which would also have been homogenized during diagenesis [Shimizu and Stoll, 2004]. End member isotopic compositions can be estimated directly for only two genera, Chiasmolithus and Toweius, but even when the coarse fraction includes other genera like Discoaster and Zygrhablithus, there are not major changes in isotopic fractionation between the coarse and fine coccolith fractions. Consequently, although the coccolith fractions separated here represent only a small fraction of the global diversity of Paleocene coccolithophorids, they do sample species from across the range of coccolithophorid phylogeny and suggest a more limited overall range of vital effects among Paleocene coccoliths compared to modern.
 If cell diameter were an important control over isotopic fractionation in the Paleocene as in the modern, then a limited range of coccolith vital effects might be explained by a more limited range of cell sizes among different coccolith species. While there are significant differences in the sizes of coccoliths analyzed in this study (3 to 5 μm for Toweius, and 8 to 12 μm for Chiasmolithus and Discoaster), and it would be unusual for large placoliths to occur on a cell smaller than the coccoliths, it is possible that numerous small Toweius coccoliths could have occurred on large cells comparable to those of Chiasmolithus. Recovery of late Paleocene fossil coccospheres of Toweius and Chiasmolithus could test this hypothesis.
 It is also possible that isotopic fractionation becomes less sensitive to cell diameter at larger cell diameters. In culture experiments, the relationship between cell size and carbon isotopic fractionation flattens out at cell diameters larger than 10 μm [Ziveri et al., 2003]. Overall, Paleocene coccoliths and presumably coccolithophorid cells, were larger than modern coccoliths and therefore might be expected to have a smaller range of interspecific vital effects. Also, the large range of vital effects in modern species is hypothesized to reflect different carbon acquisition strategies and efficiencies among different species. The smaller range of vital effects in Paleocene coccoliths could also reflect more similar carbon acquisition strategies among different species, perhaps in response to higher atmospheric CO2 concentrations.
 Stable isotope records in restricted coccolith fractions from ODP Site 690 closely reproduce the major trends in carbon and oxygen isotopes observed in coccolith-dominated bulk carbonate. Consequently, trends in bulk carbonate isotopic ratios are not artifacts of changing phytoplankton communities and must reflect true environmental changes For example, the plateaus in the coccolith carbon isotope excursion likely represent real events, though with existing data we cannot confirm if they reflect a change of δ13C of dissolved inorganic carbon due to multiple injection of methane or periodic variations in the carbon cycle, or an environmentally driven intraspecific vital effect in coccoliths. The relatively small differences in isotopic compositions between different coccolith fractions suggest that bulk carbonate isotopic analyses are unlikely to be significantly biased by changing nannofossil assemblages and likely represent reliable paleoceanographic records for the Early Cenozoic.
 Stephen Howe of SUNY Albany provided expert assistance with stable isotopic analyses, and Katharina Billups gave close reading and discussions which improved this manuscript. Samples were provided by the Ocean Drilling Program, and Gar Esmay of the East Coast Repository is appreciated for assistance in sampling the core. Santo Bains shared splits of samples used in his original study [Bains et al., 1999], and Jackie Lees offered guidance with nannofossil taxonomy. Financial support for this work came from startup funds provided by Williams College. Undergraduate research assistants Alcia Jackson, Alison Gaby, Nina Trautman, Susie Theroux, and Alicia Arevalos assisted in the laboratory separation of coccolith fractions and in scanning electron microscopy.