- at 
max) is recalculated to rebound by dividing by the maximal testing stress (
max) and multiplying by the in situ stress (Table 2), following the method of Hamilton (1976).
The compaction curves follow the classical pattern for constant rate of strain tests: following a stress interval with elastic deformation, each sample reaches in situ stress or maximal prior stress, yields, and enters the normal-consolidated interval, where it undergoes plastic pore collapse (Fig. 2A, Fig. 2B). If the test is continued to high strains, work hardening sets in (Fig. 2A). To compare the textural changes in the studied samples, the compaction curves are recalculated to strain vs. porosity, and different consolidation curves are obtained (Fig. 3, Fig. 4).
In the following sections the compaction curves are discussed with regard to three aspects: (1) rebound of core material after removal from its natural position, (2) natural stress history of the material cored at Sites 999 and 1001, and (3) textural interpretation of the form of the compaction curves.
In the present study, unloading curves were recorded for a few samples (Fig. 3,
Fig. 4), but a possible rebound effect is discussed by comparing the final porosity measured from the dimensions of the sample when removed from the test chamber to the porosity calculated from the maximal strain observed during testing. The difference in porosity (Final - at max) is recalculated to rebound by dividing by the maximal testing stress ( max) and multiplying by the in situ stress (Table 2), following the method of Hamilton (1976).
The estimated rebound in porosity has similar depth trends for Sites 999 and 1001 and ranges from zero near the seafloor to 1.9% at a depth of 459 mbsf. These values are larger than those reported by Lind (1993a) following the Hamilton method for samples from the Ontong Java Plateau, but smaller than those reported by Hamilton (1976), Moran (1997), and Masters and Maghnani (1993). These data suggest that no significant rebound effects are present in the most porous upper section of Sites 999 and 1001, whereas a rebound in porosity of ~2% may be expected in the cemented, deeper sections below 572 mbsf at Site 999 and below 352 mbsf at Site 1001.
An alternative way of estimating rebound is from comparison of wet bulk density determined in the laboratory and the downhole density log. The wet bulk density should in principle be lower than the logging density by an amount corresponding to the rebound. This strategy proved unsuccessful: the wet bulk density is in every data point larger or equal to the logging density, and no consistent depthwise pattern is seen (Fig. 5, Fig. 6). Rather, these data indicate that the downhole density is artificially low, which may occur in a case of excessive hole roughness, whereby the density logging tool obtains a signal from not only the formation but also part of the water-filled hole. Indeed, the caliper logs for Holes 999B and 1001A indicate severe hole roughness (Sigurdsson, Leckie, Acton, et al., 1997).
The determination of preconsolidation from the constant rate of strain curves is difficult because the samples were not left to creep at any stress level, and because the applied strain rate of 10-6 s-1 is high when compared to natural conditions (~10-15 s-1). The measured stresses thus become too high relative to the strain (Ruddy et al., 1989). To estimate the preconsolidation for the present samples, Cassagrande's construction, as presented by Jacobsen (1992), was applied: the tangent of the normal consolidated part of the compaction curve crosses the stress axis in the log stress-strain plot in the point k´ (Fig. 2). The preconsolidation stress, pc´, then becomes: pc´ 
2.5 k´. For simplicity and because of the expected overestimation of stress, the maximal experienced burial stress is estimated by making the modified preconsolidation equal to k´.
Using this estimate, the modified preconsolidation comes close to in situ stress for the samples from Site 999
(Fig. 7). Because no major hiatus is reported in the strata deposited during the period 0-20 Ma (corresponding to the depth interval 0-500 mbsf) at Site 999 (Sigurdsson, Leckie, Acton, et al., 1997), and because the acoustic velocity
(Fig. 1) and petrographic description
(Table 1) indicate that only the deepest loaded sample is from the zone of initial cementation, the k´ probably represents the maximal experienced stress reasonably well and can be used for the interpretation of the burial history of Site 1001.
The modified preconsolidation data for Site 1001 indicate that the three uppermost samples have been subjected to effective burial stresses not significantly larger than the present (Fig. 7; Table 2). The three deepest samples are overconsolidated. Only the topmost of these is shown in Figure 7 because the two deepest samples (from 395 and 459 mbsf) did not yield during the compaction experiments (Fig. 4), and the preconsolidation stress is above the maximally applied 40 MPa. These two samples are both cemented; therefore, chemical processes have probably controlled their present high-yield strength rather than previous deep burial (Pl. 2, Fig. 5, Fig. 6). The sample from 304 mbsf is not cemented, as indicated by a high specific surface and from thin sections (Table 1, Table 3; Pl. 2, Fig. 4), and it is possible to estimate a modified preconsolidation stress of 9 MPa (Fig. 7; Table 2). By comparison with the present burial stress curve, this suggests a previous burial to ~550 mbsf (250 m deeper). This would also imply that the two deeper samples have been buried to ~650 and 700 mbsf, respectively. By comparison between the acoustic velocity curve for Sites 999 and 1001, this conclusion seems reasonable because a downward shift of 250 m of the curve for Site 1001 brings it to match the curve for Site 999.
The hiatus where the material was removed by erosion should thus be located between the depth of the normally consolidated sample from 206 mbsf and the depth of the overconsolidated sample from 304 mbsf. The paleontological data indicate a minor hiatus at 304 mbsf (Sigurdsson, Leckie, Acton, et al., 1997) (i.e., just above the sample). Hence, it cannot be ruled out that the high preconsolidation may be caused by seafloor induration, although no cementation of microfossils was observed in thin section and no cementation is indicated by the high specific surface (Table 3). The major hiatus of Site 1001 is at 166 mbsf (Sigurdsson, Leckie, Acton, et al., 1997), above the sample collected for loading testing at 206 mbsf, where no significant overconsolidation was found.
The apparent contradiction may be tentatively resolved as follows. The in situ stress curve for Site 1001 is less steep than that defined by Site 999, which may be assumed to be normal because it matches the reference curve for Site 807 down to ~500 mbsf, the depth of the onset of cementation at Site 999
(Fig. 7). If we assume that the shape of the stress curve for Site 1001 reflects an earlier deeper burial, we obtain a match of the curves by shifting the part of the curve below the hiatus at 166 mbsf down by 250 m. This would imply that 250 m + 166 m 400 m of section is missing above the major hiatus. The modified preconsolidation for a sample from 206 mbsf should by this interpretation be ~4 MPa, not grossly different from the measured 3 MPa, and the sample from 304 mbsf would have been buried to ~550 mbsf, giving rise to a preconsolidation of 6 MPa, not unreasonably low when compared to the measured 9 MPa. The expected preconsolidation thus becomes lower when it is compared to the burial curve for Site 999 rather than the curve for Site 1001.
No macroscopic stylolites were found in the argillaceous calcareous sediments at Sites 999 and 1001, whereas wispy lamination was commonly observed (Table 5; Pl. 1, Fig. 6; Pl. 3), which follows observations by Bathurst (1987). At Site 999, wispy laminations were noted below 572 mbsf (Sigurdsson, Leckie, Acton, et al., 1997), within the depth range where they are found in the Ontong Java Plateau. At Site 1001, wispy lamination was observed below 305 mbsf (Sigurdsson, Leckie, Acton, et al., 1997). An interpretation that involves a previous burial depth of 550 mbsf for material collected at 304 mbsf thus brings the presence of wispy laminations at Site 1001 in line with respect to burial depth with the observations at Sites 999 and 807, so that all intervals with wispy lamination have been buried to at least 500 mbsf.
The present depths of the first appearance of wispy lamination at Sites 999 and 1001 correspond to depths from which cementation is observed; therefore, a direct link might be expected: material dissolved at the microstylolites precipitates in the matrix as cement. (Micro)stylolites are indeed frequently interpreted to be the source of cement during burial diagenesis (e.g., Choquette and James, 1990; Maliva and Dickson, 1992; Borre and Fabricius, 1998). Dissolution at wispy laminations and stylolitic seams could, from these data alone, be concluded to be associated with the onset of cementation during burial diagenesis, but the observations at the Ontong Java Plateau indicate that this is not the case. At the Ontong Java Plateau, dissolution seams are observed from below a depth of 490 mbsf and stylolites below a depth of 830 mbsf (Lind, 1993b), whereas cementation is observed below 1100 mbsf (Borre and Fabricius, 1998). These observations indicate that the formation of wispy lamination (and probably also macroscopic stylolites) is governed by burial depth, whereas the onset of cementation is also governed by other factors, such as temperature and chemical composition of the pore water (Øxnevad and Meshri, 1997).
For the upper 572 mbsf of Site 999, the sedimentation rate is nearly constant (Sigurdsson, Leckie, Acton, et al., 1997) and the modified preconsolidation stress curve follows the burial stress (Fig. 7). It could, therefore, at first glance seem surprising that the compaction curves in the stress-porosity plot (Fig. 3) do not form a common trend. To understand the cause for this scatter, all compaction curves for deeper samples were shifted along the porosity axis to match the curve for the shallowest sample, and the porosity shift measured (Fig. 8). A similar procedure was followed for the four upper samples of Site 1001, omitting the two deeper tests that did not reach the yield point (Fig. 9). The samples that are porosity shifted the least (Fig. 8, Fig. 9) contain the largest content of porous microfossils (Table 1; Pl. 1, Fig. 4; Pl. 2, Fig. 3, Fig. 4). The shift is nonproblematic for three of the tests on samples from Site 1001, and for four of the tests representing Site 999. For each of these tests, the trend lines follow the same curve of normal consolidation subsequent to yielding. The shift in compaction curves along the porosity axis indicates a common ideal initial porosity of 72%, as represented by the initial porosity of the sample from 11 mbsf at Site 999, and thus represents a slightly higher seafloor porosity. The sample from 114 mbsf at Site 999 contains only a little intraparticle porosity (Table 1; Pl. 1, Fig. 2); thus, the porosity shift of 5% for this sample suggests that the ideal initial seafloor porosity is composed of 5% intraparticle porosity and 67% matrix porosity. The compaction curves for the sample from 323 mbsf at Site 999 and the sample from 152 mbsf at Site 1001 (Pl. 1, Fig. 3; Pl. 2, Fig. 2) break the overall pattern. The normal-consolidated part of the curve for these samples shows a larger porosity decrease for a given increase in stress than the other samples.
I chose the compaction curves for the fine-grained carbonate matrix of Site 807 (Lind, 1993a) as a reference for the interpretation of the compaction curves obtained from Sites 999 and 1001 because of the relative uniformity of the calcareous ooze and chalk of the Ontong Java Plateau. The relatively pure carbonate lithology at Site 807 is reflected in a relatively low and constant specific surface of the material (Table 4; Fig. 10). The common trend line for Site 1001 falls relatively close to the trend in matrix porosity of calcareous ooze, as defined for the tests for Site 807, whereas the trend line for Site 999 is shifted to higher porosities (Fig. 8, Fig. 9). This difference is an artifact of the shifting procedure, and the trend lines may be matched by additional shifting. This indicates that the microfossils play only an insignificant role in the compaction pattern and that, despite the generally high clay content of the Caribbean samples, the compaction curves follow the pattern from Site 807 in the Ontong Java Plateau.
Regarding the composition of the samples, the two deepest limestone samples from Site 1001 and the chalk from 206 mbsf at Site 1001 have specific surfaces comparable to the samples from Site 807, and the three samples are characterized by a relatively high carbonate content as reflected in insoluble residues between 8% and 12% (Table 3). The remaining samples have high insoluble residues and corresponding high specific surfaces. The dominance of clay in the insoluble residue of the samples is thus demonstrated by the positive correlation between insoluble residue and specific surface (Fig. 11). The samples that follow the compaction trend of Site 807 have a three-modal pattern in grain-size distribution; thus, all data are represented by three classes of sizes (Fig. 12). The similarity in compaction trends thus indicates that the compaction is controlled by the fine-grained calcite, whereas the larger microfossils are passive, and the fine-grained clay fraction is dispersed in the pore fluid. The latter effect may be possible because the water-bearing clay is dispersed in the pores between the carbonate particles, rather than concentrated in layers.
Even if dispersed, the clay is in some cases the dominating constituent: the two samples with distinct compaction patterns (from 323 mbsf at Site 999 and from 152 mbsf at Site 1001) are characterized by a small proportion of larger grains, by their mudstone texture (Table 1; Pl. 1, Fig. 3: Pl. 2, Fig. 2), and by their large insoluble residue and specific surface (Table 3). The large porosity reduction for these samples for a given stress increase is thus probably a consequence of the near bimodal packing of fine carbonate particles and the slightly more dominating clay (Fig. 12, Sections 165-999A-35X-5, and 165-1001B-2R-2), so that the compaction of these samples is controlled by the fine-grained clay fraction. The logic behind this explanation is that clay should compact easier than carbonate ooze, resulting in a larger porosity reduction for a given increase in uniaxial stress.
The above interpretation is the same as the critical porosity model of Nur et al. (1995). In newly deposited sediment near the seafloor, the grain-size fraction that supports the grain structure contains its maximal (critical) porosity. Higher porosities would imply that the entire material is in suspension. Finer grained material in suspension in the pores does not contribute to the mechanical stability of the frame. The studied material from the Caribbean Sites 999 and 1001 fit this pattern, probably because of the dispersed nature of the fine-grained particles. This may also explain the unexpected results of Lind (1997): relatively low porosities corresponding to a given traveltime and an overall linear relationship between acoustic traveltime and porosity was found for Leg 165 samples (Sigurdsson, Leckie, Acton, et al., 1997) in the entire porosity interval from <10% to >70% irrespective of carbonate content. This indicates a practically constant governing texture reflected in a critical porosity of 73% for all samples (Fig. 13). Similarly, for the purer chalk facies sediments of the Ontong Java Plateau, a practically constant critical porosity of 68% was indicated by the near linear acoustic traveltime porosity trend (Lind, 1997).
