ward penetrations of polar assemblages occurred during periods of reduced poleward surface flow, such as the LGM (see, e.g., McIntyre et al., 1976; Kellogg, 1980). These states can be equated with what Broecker and colleagues have termed "conveyor on" and "conveyor off" modes, and with meridional versus zonal orientations of the North Atlantic polar front, respectively (see Figure 6, for example).
Synoptic reconstructions based on studies of a large number of Atlantic deep-sea sediment cores have outlined the general pattern of polar-front movement during the last deglaciation (Ruddiman and McIntyre, 1981). The front began to retreat from its LGM position approximately 13,000 yr BP, then advanced southward once again between about 11,000 and 10,000 yr BP during the well-known Younger Dryas cold interval, just prior to a final retreat to near its present position (see Figure 6). Initial attempts to calculate the rate of polar-front migration during the deglaci
ation suggest that, in the fastest case, it swept between extreme positions in less than 400 years (Bard et al., 1987). This estimate is close to the maximum rate that can be discerned using 14C dates from different sediment cores. The typical precision of 14C ages between 8,000 and 15,000 years BP is ± 80 to ± 160 years at 1s; the compound error associated with dating isochronous events in two different cores is therefore ± 160 to ± 320 at the 80 percent confidence level (1 s), and twice that at the 95 percent confidence level (2s). Although the individual (and therefore compound) errors can be reduced by dating the same stratigraphic level repeatedly, even if there were no other stratigraphic uncertainties such as bioturbation (stirring by bottom-dwelling organisms), approximately 100 such dates would be required to reduce the error from several centuries to several decades (using 2s/n1/2). It is therefore not surprising that such estimates of circulation response time greatly exceed those suggested by ice-core studies.
Rather than fight the errors involved in trying to chart the sweep rate of faunal boundaries between cores, one can alternatively measure the timing and rates of faunal change at a single site that is diagnostic of the overall behavior of the Atlantic circulation system. One such location is the region between the Faeroe Islands and southwestern Norway, which is effectively a gateway to the Nordic seas through which warm surface waters are drawn by deepwater formation and export across the Iceland-Faeroe Ridge and through the Denmark Strait. At this location, the degree and rate of warming or cooling indicated by faunal changes should be indicative of both the degree and the rate of change of poleward surface-water flow through the northern North Atlantic. Because conditions in the Norwegian Sea would be of uniformly polar character in the absence of Atlantic inflow, it is sufficient for the present purpose to measure the proportion of the single polar taxon, Neogloboquadrina pachyderma (sin), with respect to the sum of all planktonic foraminifera. N. pachyderma (sin) constitute more than 95 percent of the fauna at summer sea surface temperatures (SSTs) below 5°C or so (polar and Arctic water masses) and are extremely rare in waters with summer SSTs above about 10°C (Bé and Tolderlund, 1971; see Figure 7). Summer temperatures in this range are associated primarily with the region of mixing between Atlantic and polar waters within the basin today (Johannesson, 1986; Lee and Ramster, 1981). Past changes in N. pachyderma (sin) percentages between values corresponding to the 5°C and 10°C isotherms should record nearly the full range of SSTs associated with past changes in flow of Atlantic waters into the basin.
Measurements of N. pachyderma (sin) percentages were carried out in core Troll 3.1 from an area of high deposition rates in the western North Sea (see Figure 6b for core location). The seismostratigraphy, lithostratigraphy, and dating of the core site are described in Lehman et al. (1991).