DISCUSSION

The MAT is a powerful method of exploring multivariate faunal data and estimating paleoenvironmental conditions. A robust modern data base is of first-order importance in the successful interpretation of paleontological data using the MAT. This study has enabled us to carefully assess the Pacific portion of the global coretop database described by Prell (1985). There are a number of taxonomic problems in this database. For example, differences in taxonomy between workers are evident in the original data (i.e., the faunal slides). The database itself switches between taxonomic counting categories in different regions because of different ideas or original uses for the data. In our implementation of the MAT, data had to be reorganized into a set of categories recognizable in both Pliocene and Holocene deposits. This effectively eliminated most of the minor taxonomic problems.

The distribution of modern samples with respect to the modern environment is also a primary factor for the MAT. Slightly more than 50% of all Pacific data is within ±10° of the equator, and the distribution of the remaining data is heavily skewed toward the southern Hemisphere and extreme western North Pacific (see Fig. 3). It is possible that more analogous samples might have been found in other oceans had we chosen to compare our Pliocene data to the entire modern coretop database.

Dissolution is another first-order problem for MAT analysis. Figure 8 shows the relationship between dissolution and SCD for samples from DSDP Site 445. As indicated above, SCD values increase as dissolution preferentially removes the more fragile elements of the assemblage. In general, solution-susceptible taxa represent the warm end members of planktonic foraminifer assemblages. Therefore, increased dissolution results in cooler SST estimates. This is especially true in tropical assemblages consisting of greater percentages of fragile tests. In low latitudes, minor dissolution can remove large percentages of the assemblage. Conversely, in polar or subpolar regions, planktonic foraminiferal assemblages exhibit much lower diversity and the individual taxa are less prone to dissolution. The robust and dissolution-resistant taxa will not show a discernible change in SCD values unless pervasive dissolution removes large proportions of the assemblage. Even then, unless the dissolution preferentially removes particular taxa the resultant sample may contain proportions analogous to modern samples.

We suggest a sliding scale for acceptable SCD values should be employed, in part, dependent upon latitude. In high-latitude regions anything more than a minor increase in SCD value should be reason to suspect the analysis and question the integrity of the original sample. In lower latitude regions where minor dissolution can have a greater effect a larger SCD cutoff should be employed.

The assumption of stability of ecological tolerances over long time scales requires further discussion. Together with related grouping schemes utilized herein, this assumption of stability (basic to most paleontological studies) is subject to criticism. The question is whether in the course of 3 million years of evolution, planktonic foraminifers (or other organisms) adapt in a manner that significantly changes their environmental preferences from those of their ancestors. Since it is nearly impossible to answer this question with certainty, regardless of timescale, the grouping of ancestors with modern descendants and/or fossil taxa with modern taxa having similar morphologies or biogeographic distributions is inherently open to criticism. Nevertheless, various geochemical and isotopic techniques can be employed to determine whether a particular taxon occupies the same position with respect to stable isotopes (delta¹⁸O, delta¹³C) over time or whether an ancestor and descendent occupy the same ecological space. Interestingly, these tests also rely on an equally challenging and untestable assumption: that organismic vital effects remain constant over time. In truth, it is often not possible to ascertain whether the assumptions (basic to almost all paleontological research) that must be made to apply MAT to Pliocene age sequences can ever be unequivocally proven correct. Any attempt to infer environmental states from geochemical observations is flawed from the outset by the assumption that a biological system is known in sufficient detail today and that the system did not interact differently with the physical environment in the past.

Irrespective of these considerations, we hold some confidence that the assumptions which, after careful experimentation,we have made are valid, not because the results we obtainedare those we anticipated, but because the same results were obtained with different types of organisms and different quantitative methods. For example, the Pliocene of the northeastern Atlantic Ocean has been intensively studied over the past 15 years. Willard (1994) suggests that during the middle Pliocene areas of Iceland now covered by tundra vegetation were covered by deciduous forest, suggesting temperatures 3° to 5° degrees warmer than today. Ostracods were used by Cronin (1991) to show shallow sea-bottom temperatures (the same zone sampled by planktonic foraminifers except closer to the shoreline) off Iceland were warmer than today by about 5°C (both winter and summer) during the warm intervals of the middle Pliocene. Dowsett and Poore (1990) and Dowsett et al. (1992) estimated sea surface temperatures to be 3.6°C and 7.9°C (winter and summer) warmer than today during mid Pliocene "interglacials" in the same region. These three fossil groups (plants, animals and protists) all give more-or-less the same results using three independent techniques. While we agree that assumptions about environmental preferences are inherent in all three studies, it is highly unlikely that these investigations would all suggest the same general answer unless they were sensing on a strong and consistent signal. Therefore, we offer no answers for the critics of paleontologically-based environmental reconstruction, but also hold that through careful integrated analysis using different biological systems one can gain improved confidence in results.

Dowsett et al. (1994) reconstructed winter and summer SST fields of the North Pacific from a limited number of data points that showed a step-like increase in SST across the Pacific rather than a smooth gradient, such as that which typifies the modern Pacific. Our results from Site 769, combined with work done at Sites 573 (Hays et al. 1989), 586 (Jenkins 1992a, b), 846 (Shackleton et al. 1995) and 806 (Andersson 1997) confirm that low-latitude Pacific SST did not increase during the middle Pliocene when higher latitude sites in the Pacific and Atlantic were experiencing warming. In fact, there is some evidence for minor cooling that, if it can be confirmed through more sensitive analyses and better stratigraphic correlation, would support the hypothesis of increased oceanic meridional heat flow as the mechanism for middle Pliocene warming. The warming estimated at Site 445 is in keeping with warming estimates from Sites 580, 880, and 883 (Barron 1992, 1995) and 887 (Dowsett and Ishman 1995), and suggests the existence in the middle Pliocene of a strong western boundary current like the modern Kuroshio Current. The results of this study are an important refinement of the original PRISM northern hemisphere reconstruction (Dowsett et al. 1994) and have been incorporated into the first PRISM global reconstruction (Dowsett, Barron, and Poore 1996).