Having considered the paleomagnetic data for each site, it is natural to consider the implications of the combined data set. Sites 999 and 1001 both indicate similar amounts of northward motion for the Caribbean plate over the past 80 m.y. Similarly, Site 998, just to the north of the Caribbean plate, has moved northward at roughly the same rate.
The data from Sites 999 and 1001 can be combined given that the Euler pole that describes the motion of the Caribbean plate relative to the spin axis is not located near the sites. A nearby Euler pole could produce vastly different amounts, and even directions, of latitudinal motion for these sites. Two lines of evidence support a distant Euler pole. First, the similar northward motions estimated at Sites 999 and 1001 are consistent with motions about a distant Euler pole. Second, the long continuous nature of the Swan Island and Oriente transform faults, which separate the North American and Caribbean plates along the Cayman Trough, combined with the motion of the North American plate as recorded by its apparent polar wander path, require that the Caribbean plate has rotated little relative to the spin axis over the past 50 m.y.
Taken together, the results from Sites 999 and 1001 give a clear indication of the change of paleolatitude of the Caribbean plate since the Late Cretaceous. In order to illustrate this point, we plot all the paleolatitudes relative to the current position of Site 999 (Fig. 14). To place the Site 1001 data in this reference frame, we take the change in paleolatitude predicted by each paleolatitude datum (i.e., the current latitude of Site 1001 minus the paleolatitude estimate) and subtract this from the current latitude of Site 999. For example, the 65- Ma paleolatitude estimate from Site 1001 is 4.7°, which translates to a paleolatitude of 1.7° (= 12.74° - [15.76°- 4.7°]) at Site 999. Figure 14 can also be viewed as a plot of northward motion vs. age. Only the 9-Ma mean paleolatitude from Site 999 falls off this plot. Because it was based on only three sample means, far less than the other mean paleolatitudes, and has a large uncertainty, we ignore it in the following discussion, though its inclusion would not change any of the conclusions.
The combined paleolatitudes shown in Figure 14 indicate that Site 999, and hence the southern portion of the Caribbean plate, was near or at the equator in the Late Cretaceous, and has since migrated progressively northward. Data from Sites 999 and 1001 agree well with each other and with a model in which the northward migration occurs at a constant rate. The best-fit rate of the northward motion is 18 ±4 km/m.y., a rate that is consistent with all 13 of the paleolatitude estimates from both sites. Within the uncertainties, however, the data do not preclude changes in the rate of northward motion, such as the example indicated by the thin solid line in Figure 14.
Of the 18 mean paleolatitudes estimated in this study, 17 are from sedimentary units. Even though many sediments have been shown to be accurate recorders of the geomagnetic field direction (e.g., Opdyke and Henry, 1969; Harrison, 1966; Opdyke, 1972; Kent, 1973), the accuracy of other sedimentary units has been questioned. Several past studies have shown that the inclinations for some sedimentary units are systematically shallower than the known or true inclinations of the paleomagnetic field (e.g., Kent and Spariosu, 1982; Celaya and Clement, 1988; Tarduno, 1990; Gordon, 1990). This systematic shallowing is referred to as the "inclination error." So far, the size of the inclination error has been difficult to resolve as it may depend on a variety of variables, such as the degree of compaction, porosity, water content, sediment composition, degree of bioturbation, and others (e.g., Arason and Levi, 1990a, 1990b; Celaya and Clement, 1988; Tan and Kodama, 1998). What is known, is that the inclination error can be significant, with the error in some cases possibly exceeding 20° (Celaya and Clement, 1988; Tarduno, 1990; Gordon, 1990). These studies and theoretical studies (e.g., Arason and Levi, 1990a, 1990b) have also shown that the error is largest in mid-latitudes and decays at low and high latitudes. The main questions for this study are do the sediments that we use contain an inclination error, and if so, how large is the error?
Though there is no direct evidence that our sedimentary results are biased by the inclination error, there are several observations that suggest that the inclination error is likely small, if present at all. First, as would be expected given the rate of plate motions, the results from young sediments, particularly those younger than 10 Ma, give paleolatitudes consistent with the current latitude. Some of these sediments from Site 999 are from depths >200 mbsf, so a good deal of compaction has taken place. In the study by Celaya and Clement (1988), of the six sites they studied, the inclinations from three sites were biased. At those three sites, the inclination shallowed with depth in the upper 200 m. In our study, the apparent absence of shallowing in the young sediments from the upper 200 m indicates that these Leg 165 sediments record the paleofield direction accurately. Second, the good agreement between the mean paleolatitudes (and mean inclinations) obtained from the basalts and from the overlying sediments, indicates that little if any inclination shallowing has occurred. Within the 95% confidence interval for the basalt estimates, however, large inclination errors could be permitted, but northward motion is required by the data even when pushed to its outer limits. Third, Tan and Kodama (1998) show that the inclination error increases with clay content, with the largest errors occurring in sediments with >40% clay content. For the Nacimiento and Ladd Formations, the two examples used in Tan and Kodama (1998) with <20% clay content, the inclination error was 7°-11°. The clay content for Leg 165 cores is typically <40% (Fig. 15A; pp. 173 and 322 of Sigurdsson, Leckie, Acton, et al., 1997), and is <20% for samples below 400 mbsf from Site 1001 (those samples used in the 74-Ma paleolatitude estimate). Furthermore, the examples used in Tan and Kodama (1998) come from mid-latitude sites, so the size of the inclination error would be larger than expected for low-latitude Caribbean sites. Hence, even if the results of Tan and Kodama were directly applicable to the Caribbean sites, the inclination error would very likely be smaller than ~8°. Fourth, Ceyala and Clement (1988) suggest that the inclination error is present in sediments where the carbonate content is >80%, with clays making up at least part of the remaining composition. They observed no inclination error downcore when the carbonate content was <80%. Though these observations conflict with Tan and Kodama's (1998) findings, it is interesting to note that only the sediments used for the 74-Ma paleolatitude estimate fall within the composition range that Celaya and Clement suggest is biased (Fig. 15A). Finally, consider the paleolatitude estimates from sediments that indicate northward motion, which are all those older than 10 Ma. When the change in paleolatitude from these is plotted against carbonate percentage, water content, or porosity, no obvious trends are apparent (Fig. 15). For example, as water content decreases downhole owing to compaction, Celaya and Clement (1988) suggest that the inclination error should increase. As shown in Figure 15, the water content is actually higher for sediments that indicate the greatest northward translation of the Caribbean plate. This occurs because the overburden at Site 1001 is roughly half that for similar age sediments from Site 999. Even if we consider the data from Site 999 separately from Site 1001, there is poor correlation between compaction-related water loss (or porosity) and northward translation. This is because there is little change in water content or porosity below 600 mbsf at Site 999, though the paleolatitudes continue to decrease with depth.
We conclude that there is little if any supporting evidence that would argue for an inclination error that could explain the observed systematic decrease of paleolatitude with age illustrated in Figure 14. Unfortunately, we cannot prove that an inclination error does not exist. Furthermore, the uncertainties in the paleomagnetic data are large enough to accommodate an inclination error as large as 10° (corresponding to ~5° of paleolatitude) without requiring a complex motion history for the Caribbean plate. Assuming an inclination error of 10° for sediments older than 10 Ma reduces the average rate of northward motion to 8 km/m.y. Larger inclination errors produce a large misfit with the paleolatitude estimated from the basalts and would require that the plate first moved northward and then southward to arrive at its current position. The preferred motion history is, therefore, that shown in Figure 14 in which the Caribbean plate migrates northward at 18 km/m.y. However, the uncertainty interval should likely be expanded to account for the possibility of an inclination error. Assuming that 10° of inclination error represents an upper bound for the 95% confidence interval, then the average rate of northward motion and it 95% confidence region is 18 (+4 or -10) km/m.y.
Three fundamentally different types of models have been proposed for the formation and evolution of the Caribbean plate: (1) the "Pacific" model has the Caribbean plate originating in the Mesozoic from a piece of one of the Pacific basin plates, most likely the Farallon plate (e.g., Malfait and Dinkelman, 1972; Pindell and Dewey, 1982; Duncan and Hargraves, 1984; Pindell et al., 1988; Burke, 1988; Pindell, 1994); (2) recent versions of the "intra-American" model have the Caribbean plate originating between North and South America, though well west of its current position (e.g., Klitgord and Schouten, 1986; Meschede and Frisch, 1998); and (3) "Fixist" models indicate the Caribbean plate formed and has maintained roughly its current position (e.g., Morris et al., 1990). This third type of model is incompatible with the paleomagnetic data from this study, as well as being incompatible with the opening history of the Atlantic Ocean (Pindell, 1994), and will not be discussed further.
The most recent versions of the Pacific and intra-American models, as presented by Pindell (Pindell et al., 1988; Pindell and Barrett, 1990; Pindell, 1994) and Meschede and Frisch (1998), respectively, predict very similar motion histories for the Caribbean plate since Campanian time. The two models merge in the Cenozoic, and only relatively minor differences occur in plate boundary geometries back to the early Campanian (~82 Ma). At that time, both models have the Caribbean plate separated from the Pacific basin plates by a subduction zone. The motion predicted by these models only differs significantly for early times, where the Pacific model assumes the Caribbean plate was part of the northeastward migrating Farallon plate.
For the time spanned by our data, the recent versions of these two models differ insignificantly in the motion they predict. Thus, our data cannot be used to support one model over the other. Other observations, such as ages and paleolatitudes of terranes bounding the Caribbean, have been used to support one model or the other (Pindell, 1994; Montgomery et al., 1994; Meschede and Frisch, 1998), though definitive evidence may come only from drilling into the oldest crust of the interior of the Caribbean plate.
The new paleomagnetic data, instead, place paleolatitude constraints on these models. As an example, we have placed a paleolatitude and paleolongitude grid over the reconstructions of Pindell et al. (1988) in Figure 16. The paleolatitudes are selected such that they are consistent with the Caribbean paleomagnetic data, as well as with North American paleomagnetic data and with the reconstructions of North America relative to the hot spots as given by Müller et al. (1993). The paleolongitudes are from the hot-spot reconstructions alone. To illustrate that the grid violates none of the Caribbean paleomagnetic data, we show the position of Sites 999 and 1001 in each panel of Figure 16 and then plot the reconstructed paleolatitudes of Site 999 in Figure 14.
The combined paleomagnetic and hot-spot data suggest the rate of northward motion for the Caribbean plate was slower in the Late Cretaceous to the Oligocene than subsequently. This seems reasonable when the motion of the Caribbean plate is considered relative to North America. For example, the 30-Ma paleomagnetic pole for North America (Diehl et al., 1988) can be used to estimate the average rate of northward motion for a point on the North American plate just north of the Caribbean plate. This pole gives a northward motion rate of 21 km/m.y. for North America averaged over the past 30 m.y. The rate of northward motion of the Caribbean plate would be slightly faster than this because the Caribbean plate was moving east-northeast relative to North America during this time. Similarly, paleomagnetic data from North America (Diehl et al., 1983; Gordon, 1984; Acton and Gordon, 1990) indicate that the southern portion of the North American plate moved southward from the Late Cretaceous to the Eocene. The reconstructions of Pindell et al. (1988) illustrate that the Caribbean plate had a larger northward component of motion relative to North America in the Late Cretaceous to Eocene (~80-40 Ma) than in subsequent times. Thus, over this interval, the net motion of the Caribbean plate relative to the spin axis was still northward, but at a rate slower than that since 30 Ma (Fig. 14, Fig. 16).
The paleolongitudes illustrate that the North American plate had a large westward component of motion, which averaged ~50 km/m.y. from 80 to 45 Ma. Over this same time interval, the Caribbean plate also moved westward relative to the hot spots, though at a slower rate of ~34 km/m.y. Since 45 Ma, the North American plate has continued moving westward at an average rate of 24 km/m.y, whereas the Caribbean plate has moved negligibly.
In one version of the Pacific model, the lithosphere that became the Caribbean plate was thickened as it passed over an incipient Galapagos hot spot in the Late Cretaceous (Duncan and Hargraves, 1984; Burke, 1988). At this time, the Galapagos hot spot would have been in its plume-head stage. Lithosphere above the plume head, which would have been part of the Farallon plate, was thickened (8-20 km thick) and became part of the Caribbean igneous province. Afterward, the Farallon plate and this oceanic plateau migrated northward and eastward away from the equatorial Galapagos hot spot. In this version of the Pacific model, this overly thickened piece of lithosphere became the core of the Caribbean plate.
The paleomagnetic data (Fig. 14) indicate that the southern Caribbean plate (Site 999) was directly over or near the equator at ~75-80 Ma. Assuming the Galapagos hot spot has maintained its position at ~0.3°S, then the paleolatitude data are consistent with the Caribbean plate passing over the Galapagos hot spot in the Late Cretaceous.
The paleolongitudes, however, indicate that it is unlikely that the Caribbean plate could have been as far west as the Galapagos hot spot (Meschede, 1998; Meschede and Frisch, 1998). For Site 999 to have been at the longitude of Galapagos at 100 Ma and then to move to that position at 80 Ma shown in Figure 16, would require that the Farallon plate move eastward relative to the hot spots at ~130 km/m.y. This rate of eastward motion is roughly two to three times faster than what has been estimated for the motion of the Farallon plate relative to the hot spots during Late Cretaceous time. For example, the average rate of eastward motion of the Farallon plate relative to the hot spots is 64 km/m.y. based on the 74-100 Ma stage pole from Duncan and Hargraves (1984) and is 41 km/m.y. based on the 85- to 100-Ma stage pole from Engebretson et al. (1985).