Recent work has suggested that highest Mg/Ca layers are associated with organic components of the test [Kunioka et al., 2006]. Here we find that the shape of the layer with highest Mg/Ca mimics the arrangement of the calcite crystals sometimes revealed by test cleaning (Figure 4a). Additionally, the tests in this study were cleaned of organic material and no difference was observed between the two cleaning techniques. Thus, we conclude that the layer with highest Mg/Ca is calcite located adjacent to the primary organic membrane, rather than the organic membrane itself. Bands of high Mg/Ca have been previously described for symbiont bearing species and attributed to symbiont photosynthetic activity [Eggins et al., 2004], yet neither of the species studied here harbor photosymbionts. Similar banding has been described in benthic species that have much higher Mg/Ca ratios than these planktonic foraminifera [Erez, 2003; Bentov and Erez, 2005]; these authors also conclude that the high Mg/Ca layer is primary calcite precipitated next to the organic membrane.
4.2.1. Environmental Controls
 The range in Mg/Ca observed here within a single test is similar to that reported previously [e.g., Sadekov et al., 2005; Anand and Elderfield, 2005]. The question central to this and previous studies is what controls the variation in Mg/Ca? It is well known that planktonic foraminifera migrate vertically through the water column and add their gametogenic crust at depth in colder waters [e.g., Orr, 1967; Rosenthal et al., 2000]. Given the close relationship between foraminiferal Mg/Ca and temperature [e.g., Anand et al., 2003], this could result in variable Mg/Ca. However, the temperature of the water column at the PAP site when the foraminifera calcified (April–May) is well established and, over the depth range of these species, varies between 9.5 and 12.5°C (Figure 2). Using calibrations [Anand et al., 2003] obtained from measurements of multiple foraminifera tests (no calibrations based on single tests, or individual components of single tests, are available), this corresponds to Mg/Ca values of ∼0.9–1.2 mmol/mol, which is significantly less than the range recorded here (∼0.4–3.6 mmol/mol (Figure 6)). Salinity also varies over the depth range of these species, but only by ∼0.2 units which would have a negligible effect on Mg/Ca [e.g., Kisakürek et al., 2008]. Meanwhile, [CO32−] varies between 130 and 190 μmol/L (Figure 2), but culturing studies of planktonic foraminiferal reveal that, for these values of [CO32−], there is no effect on foraminiferal Mg/Ca [e.g., Russell et al., 2004].
 The B/Ca ratio of the innermost part of the test wall of G. inflata is ∼20 μmol/mol higher than that recorded for the outer part; for G. scitula the difference is even greater (∼40 μmol/mol) (Figure 6). Yu et al.  have demonstrated that B/Ca of G. inflata varies both as a function of pH and also of temperature. pH measurements are not available for the PAP site, but can be estimated from measurements of pressure, temperature, salinity, phosphate, silicate, total alkalinity and TCO2 from a nearby site [Rommets, 2003; Rommets et al., 1991, 2003], together with the CO2sys macro [Pierrot et al., 2006]. In this way, foraminiferal B/Ca is expected to vary between ∼90 μmol/mol in surface waters and ∼60 and ∼40 μmol/mol at depths of 300 m and 1000 m, respectively. These values agree remarkably well with those measured across the test walls, raising the possibility that intratest variations in B/Ca are the result of changes in pH and temperature due to vertical migration of the foraminifera.
 Li is conservative in the oceans [Stoffyn-Egli and Mackenzie, 1984] so variations in foraminiferal Li/Ca cannot be attributed to changes in seawater Li/Ca. Temperature appears to regulate the Li/Ca ratio of inorganic calcite [Marriott et al., 2004], but the size of this effect is small and acts in the wrong direction to explain the variation in the Li/Ca ratio between the innermost (>30 μmol/mol) and outermost (10 μmol/mol) calcite observed here. Furthermore, studies of planktonic foraminifera indicate that the relationship between temperature and the Li/Ca ratio is weak [Hathorne and James, 2006; Hall and Chan, 2004]. Changes in seawater [CO32−] must also be considered as Lear and Rosenthal  have suggested that the Li/Ca ratio of benthic foraminiferal calcite falls by ∼10% as the degree of calcite saturation (ΔCO32−, the difference between seawater [CO32−] and [CO32−] required for calcite to be at saturation) falls from 45 to 5 μmol kg−1. At the PAP site, ΔCO32− ranges from 80 to 130 μmol kg−1 over the depth range for calcification of G. inflata and G. scitula; if the relationship between Li/Ca and ΔCO32− for benthic foraminifera holds for these species then Li/Ca would be expected to decrease by ∼13% through the shell wall which is far less than is measured here. Recently, Bryan and Marchitto  observed a decrease of ∼20% in the Li/Ca of benthic foraminiferal calcite as ΔCO32− increased from 50 to 150 μmol kg−1 in the Florida Straits. Again, if this effect is important for planktonic foraminiferal calcite, then it is too small and acts in the wrong direction to explain the intratest Li/Ca pattern that we observe here.
 Manganese and barium are classified as recycled elements in seawater, as their concentration is low in surface waters and increases with depth [e.g., Bruland and Lohan, 2003]. In this study, we find that the outer part of the test, which formed in deeper waters, has lower Mn/Ca and Ba/Ca than the inner part of the test so vertical migration cannot account for the intratest variation in these elements. Whether changes in Mn/Ca and Ba/Ca are effected by other variables is not clear from the literature, as studies of these trace elements are scarce.
 Sr/Ca ratios show little variability across the test walls of G. scitula and G. inflata. A temperature control on the Sr/Ca ratio of deeper dwelling Globorotalia species has been reported by Elderfield et al.  and Cléroux et al. , but if there is any temperature control on Sr/Ca for the species measured in this study then it is smaller than the external precision of the analyses (∼10% (Figure S2)).
 To summarize, the data that we present here seem to demonstrate that, with the exception of B/Ca, the intratest variation in Mg/Ca, Li/Ca, Mn/Ca and Ba/Ca of G. scitula and G. inflata is not primarily controlled by environmental parameters. Furthermore, because the tests were collected over a very short (2 week) time period, differences between the chemical composition of the individual tests are also unlikely to be due to differences in environmental parameters, such as temperature and salinity. For these reasons, this data set can provide crucial insight as to the biomineralization control on trace element/Ca ratios; such information may be lost in core top studies, because the individual tests will have grown during different seasons and years, stretching often up to thousands of years into the past.
4.2.2. Biomineralization Controls
 Foraminifera can regulate the trace element composition of their tests in a number of different ways. The aim here is to identify which of a number of different biomineralization processes can reproduce the sense and magnitude of the mixing relationships that we observe between Mg/Ca and Mn/Ca, Sr/Ca and Ba/Ca (Table 1 and Figure 7). Only the divalent cations are considered, because these are known to substitute for Ca2+ in the calcite lattice [Reeder et al., 1999].
 First, we consider the effect of precipitation rate. A number of laboratory studies have shown that the partition coefficient, DX, between element X in inorganic calcite and in solution varies as function of calcite precipitation rate [e.g., Lorens, 1981; Tesoriero and Pankow, 1996]. While the precipitation rate of foraminiferal calcite is thought to be slower (by a factor of 2 or more) than the precipitation rate of inorganic calcite [Erez, 2003], a precipitation rate control on the trace element composition of foraminiferal calcite is often implicated in the literature [e.g., Hall and Chan, 2004; Russell et al., 2004]. The variation in DX as a function of precipitation rate can be modeled in terms of ion mobility at the calcite-solution interface; this is a function only of ion size when the charge is the same [Watson, 2004]. Figure 8a shows the effect of changing precipitation rate on DX; note that DMg is independent of precipitation rate [e.g., Morse and Bender, 1990], while the rate dependences of DMn and DBa act in opposite directions. This is in contrast to the trends shown in Figure 7, suggesting that precipitation rate is not responsible for the intratest variation in trace element/Ca ratios, at least for the divalent cations. Furthermore, changes in precipitation rate cannot explain the small variation observed in Sr/Ca relative to other trace elements.
Figure 8. Graphs showing the variation in X/Ca ratio as a result of (a) changes in precipitation rate, (b) Rayleigh distillation of the calcifying solution, and (c) changes in calcite structure. For Figure 8a, data are from Lorens  for DMn and from Tesoriero and Pankow  for DSr and DBa. For Figure 8b equation (vii) from Elderfield et al.  was used. Note that X/Ca is reported relative to the initial value as the absolute value is dependent on the boundary conditions employed by each model.
Download figure to PowerPoint
 Second, it seems likely that the perforate foraminifera (which include G. scitula and G. inflata) incorporate Ca2+ and trace elements via intracellular vacuoles that serve as reservoirs for calcification [Erez, 2003]. The composition of the solution in the vacuoles is similar but not necessarily identical to seawater [Elderfield et al., 1996; Erez, 2003] and trace elements are extracted from it during the biomineralization process. If this is the case, then the composition of the foraminiferal calcite changes as a function of the fraction of Ca2+ remaining in the reservoir. This can be modeled in terms of Rayleigh distillation [Elderfield et al., 1996] and the results are shown in Figure 8b. One of the predictions of this model is that intratest heterogeneity will be large for ions that have DX > 1 (Mn2+) but small for ions that have DX < 1 (Sr2+, Ba2+, Mg2+). Our data indicate that this is true for Sr but not for Ba. The model also predicts that the relationship between DX and the size of the Ca2+ reservoir for elements with DX > 1 is opposite to that for elements with DX < 1. Thus, Mn2+ and Ba2+ should show contrasting behavior (Figure 8b), but this is not what we observe in our data (Figure 7). In addition, it is difficult to reconcile the abrupt changes in the Mg/Ca ratio of the test (as revealed by EMP (Figure 3)) with the smooth, gradual nature of a distillation process.
 Finally, the incorporation of trace elements into foraminiferal calcite can be affected by changes in the structural form of the calcite [Russell et al., 2004]. Figure 4 demonstrates that the structure of the calcite crystals that form the inner and outermost parts of the test wall are distinctly different and, in this connection, it is well documented that calcite growth hillocks have two types of vicinal faces with the positive faces incorporating trace elements in a way that is different from the negative faces [e.g., Paquette and Reeder, 1990, 1995; Reeder, 1996]. It is also recognized that the proportional area of each of these faces can be altered by Mg incorporation [Davis et al., 2004] and the presence of amino acids [Orme et al., 2001]. The average composition of the calcite (Cav) can be calculated in terms of the relative proportion of positive versus negative faces:
where C+ve and C−ve are the trace element concentration of the positive and negative faces, respectively, taken from Paquette and Reeder  and Reeder , and P+ve and P−ve are the fractional proportion of positive and negative faces (P+ve + P−ve = 1). The results of this calculation are shown in Figure 8c. This model successfully accounts for the small variation in Sr/Ca, but the relationship between Mg/Ca and Ba/Ca is in the opposite sense to that which we observe (Figure 7). Moreover, while the model can reproduce the covariation between Mg/Ca and Mn/Ca shown in Figure 7, it accounts for <30% of the range of values that we measure. One complication, however, is that the crystal face preference of Mg can shift depending on whether calcite growth is diffusion or surface reaction limited [Wasylenki et al., 2005]; this is not accounted for in the model presented here.
 Although this exercise is instructive, none of the biomineralization processes discussed above can, by themselves, account for intratest variations in the composition of calcite that we observe. Nevertheless, it is likely that trace element/Ca ratios are affected by more than one process, and there is a need to establish an integrated biomineralization model, as well as models for other biomineralization processes. Much more knowledge is required concerning the effect of precipitation rate on the crystal face preferences of various elements and the influence of specific organic molecules, known to be integral to foraminiferal calcite [Robbins and Brew, 1990; Robbins and Donachy, 1991], on crystal structure and trace element partitioning. Additionally, cellular ion transport also has the potential to alter the chemistry of foraminiferal calcite [e.g., Gussone et al., 2003] but little is known about this process. What is clear, however, is that the Mg/Ca ratios that we have measured in this study imply that the Mg/Ca of the solution from which the foraminifera calcifies is at least an order of magnitude lower than the Mg/Ca ratio of seawater. How the foraminifera regulate the Mg/Ca ratio of the calcifying solution is still debated [Bentov and Erez, 2006], but changes in the solution Mg/Ca will likely impact the crystal size and orientation [Kwak et al., 2005] suggesting that the foraminifera regulate the Mg/Ca ratio of the solution to control crystal growth. The difference in crystal structure between the inner and outer calcite (see Figure 4) could be the direct result of the difference in Mg/Ca between the inner and outer calcite. This implies that DMg is primarily a function of solution Mg/Ca, as regulated by the organism, and not temperature. This is a departure from the model of Bentov and Erez  in which test Mg/Ca is defined by the proportion of high Mg (primary) calcite to low Mg (secondary) calcite, which in turn is regulated by temperature, and it suggests that a detailed examination of the possible physiological impacts on the foraminiferal Mg/Ca paleothermometer is required.