Paleomagnetic data from Detroit Seamount of the Emperor seamount chain (northwestern Pacific Ocean) provide a direct way to estimate the latitudinal position of the Hawaiian hot spot in the Late Cretaceous (75–81 Ma), providing important constraints on geodynamic models for Hawaiian plume motion. Crystalline basement of Detroit Seamount was sampled at four drill sites by Ocean Drilling Program (ODP) Legs 145 (Sites 883 and 884) and 197 (Sites 1203 and 1204). Here we report detailed thermal and alternating field (AF) demagnetization data collected from basalt lava flows recovered at ODP Site 883. Most samples showed two nearly antiparallel components of natural remanent magnetization (NRM). AF demagnetization was often insufficient to separate the magnetization components; therefore thermal demagnetization data were used to define their directions. Both magnetization components are chemical remanent magnetizations (CRMs) carried by single-domain to pseudo-single-domain titanomaghemite produced by low-temperature oxidation of titanomagnetite. Rock magnetic experiments suggest that the normal polarity component is a CRM which pseudomorphs the primary thermal remanent magnetization (TRM) of parental titanomagnetite. In contrast, the reversed polarity component is a CRM self-reversed with respect to a primary TRM [Doubrovine and Tarduno, 2004]. Inclination data analysis of characteristic remanence (normal polarity component) results in the five distinct inclination groups and an average inclination of 57.7°−13.8°+12.3° (95% confidence limits quoted). The angular dispersion of paleomagnetic directions suggests that the Site 883 basalts adequately average paleosecular variation and that the inclination average represents the time-averaged geomagnetic field. Combining all available paleomagnetic data from the four sites on Detroit Seamount, a revised paleolatitude of 34.4°−5.8°+3.2° was calculated. This estimate is significantly north of the present latitude of the Hawaiian hot spot (∼19°N), suggesting that the hot spot moved southward from Late Cretaceous to Paleogene times (81–47 Ma) at rates of ∼33–67 mm/yr. If the inclination results from ODP Site 883 represent the best estimate among the data available from Detroit Seamount, rates at the higher range of values are favored. Because such rates are comparable to the velocities of lithospheric plates, the new data from ODP Site 883 present a strikingly different view on hot spot stability.
 The Hawaiian-Emperor chain of volcanic islands and seamounts has long been considered to be the best example of a hot spot track, marking motion of the Pacific plate over a stationary mantle plume [Wilson, 1963; Morgan, 1971]. The prominent bend separating the westward-trending Hawaiian chain from the northward-trending Emperor Seamounts has usually been interpreted as resulting from a change in plate motion relative to a fixed hot spot [e.g., Duncan and Clague, 1985; Clague and Dalrymple, 1989]. However, global plate circuit reconstructions [Molnar and Atwater, 1973; Molnar and Stock, 1987; DiVenere and Kent, 1999] have failed to reproduce the change in direction of Pacific plate motion which could be associated with the formation of the bend at ∼43 Ma (or the recently revisited age of the bend at 47 Ma by [Sharp and Clague, 2002]). And there is a general lack of major tectonic events on the Pacific basin margins at this time [Norton, 1995]. However, lingering uncertainties in the plate circuits have resulted in a long-standing debate on hot spot fixity.
 Paleomagnetic data from the Pacific seamounts [Tarduno and Gee, 1995] have independently suggested large relative motion between the Pacific hot spots and the hot spots in the Atlantic and Indian oceans. Paleolatitudes available to date from four seamounts in the Emperor chain (Koko, 49 Ma, Nintoku, 56 Ma [Tarduno et al., 2003], Suiko, 65 Ma [Kono, 1980] and Detroit, 75–81 Ma [Tarduno and Cottrell, 1997; Tarduno et al., 2003]) suggest large average rates of Hawaiian hot spot southward motion from Late Cretaceous to Paleogene times inconsistent with the fixed hot spot prediction. In this paper we present new paleomagnetic data from basalt cores recovered at ODP Site 883 (Detroit Seamount) and compare them with previously published data from sites on Detroit Seamount. A revised paleolatitude of Hawaiian hot spot in Late Cretaceous incorporating all available data from Detroit Seamount is presented and its implications for the hot spot motion are discussed.
2. ODP Site 883 Basalts
 ODP Site 883 (51°12′N, 167°46′E) was drilled at the summit region of Detroit Seamount of the northern Emperor Seamounts during ODP Leg 145 (Figure 1). Two drill holes penetrated 37.8 m (Hole 883E) and 26.7 m (Hole 883F) into the basaltic basement and a sequence of fractured, highly altered aphyric to moderately plagioclase-/plagioclase-olivine-phyric basalt lava flows was recovered [Rea et al., 1993]. Twenty-seven lava flow units were identified in cores from Hole 883E and sixteen flow units were recognized in Hole 883F on the basis of the presence of quenched glassy margins. The summit region of Detroit Seamount was recently revisited by ODP Leg 197, which drilled two holes penetrating 60.0 m and 138.5 m into the basement ∼0.5 km south-southeast from Site 883 (ODP Site 1204, Holes A and B) [Tarduno et al., 2002]. The basalt units from Site 1204 were interpreted as pahoehoe lava flows consisting of multiple lobes, which originated from subaeral vents and were emplaced in a nearshore environment. The units underwent low-temperature alteration under submarine conditions. On the basis of the detailed observations at Site 1204, it is highly likely that the 27 lithologic units described from Site 883 basalts represent separate lobes within a few lava flows rather than true eruption units.
 The age of Site 883 basalts is constrained to the Late Cretaceous (Maastrichtian–Campanian) by biostratigraphy of sediments directly above and filling the fractures within the basaltic basement [Beaufort and Ólafsson, 1995]. Nannofossil assemblages of Campanian age (∼73–76 Ma) were reported from sediments overlying basalts at Site 1204 [Tarduno et al., 2002]. Nannofossils of this age were also observed in a sedimentary interbed within the last core from Hole 1204B, suggesting that the entire basement section drilled at Sites 883 and 1204 is of this age. An 40Ar/39Ar age of 75.8 ± 0.6 Ma was obtained from whole-rock basalt samples and feldspar separates from the nearby ODP Site 1203 (50°57′N, 167°44′E) [Tarduno et al., 2003]. This radiometric age is consistent with the age derived from the nannofossil assemblage at Site 883.
3. Demagnetization Data
 One hundred eight samples were collected as 2.5 cm minicores drilled perpendicular to the split surface of azimuthally unoriented cores recovered at ODP Site 883 for thermal and alternating field (AF) demagnetization experiments (66 samples from Hole 883E and 42 samples from Hole 883F). The sampling strategy was to take minicores at irregular intervals so that the entire basalt section was covered and at least three samples (where available) were collected from each lava flow unit.
 All paleomagnetic measurements were done in the paleomagnetic laboratory at the University of Rochester. Samples were thermally demagnetized at 25°C steps within a 50–625°C range using an AST TD-48 thermal demagnetization device. In addition, twenty one subsamples taken from minicores prior to thermal demagnetization were demagnetized by stepwise alternating field treatment in 5–10 mT increments to a 100–190 mT maximum applied field. Measurements of magnetization remaining after each demagnetization step were made using a 2G DC Superconducting (SQUID) Rock Magnetometer. Demagnetization data were then plotted as orthogonal vector projections and analyzed using principal component analysis [Kirschvink, 1980].
 Upon thermal demagnetization, approximately 70% of the 108 samples showed two distinct, nearly antiparallel components of natural remanent magnetization (NRM): a reversed polarity component removed at low unblocking temperatures (∼300–400°C), and a normal polarity component isolated at higher temperatures (up to 585°C) (Figures 2a–2c). Several samples showed a smaller apparent angular difference between the two components. The remaining samples yielded only a single NRM component of normal polarity after the removal of a small, spurious magnetization at unblocking temperatures ≤150–200°C (Figure 2d).
 A reversed polarity component was removed at ∼15–20 mT during AF demagnetization, and a normal polarity component was isolated at higher applied fields (typically 20–150 mT). When compared on a sample basis (Figure 3), the low-coercivity reversed polarity segments of the AF demagnetization plots were usually shallower than those defined at low unblocking temperatures in thermal demagnetization data. Directions of normal polarity component isolated at high temperatures and high applied fields were sometimes consistent (Figures 3a and 3b). However, there were notable exceptions: for samples with a strong or dominant reversed polarity NRM component (i.e., those showing a reversed NRM before demagnetization) we usually observed smoothly curved AF demagnetization paths. The nominal inclinations of the normal polarity components for such samples are steeper than those determined from thermal demagnetization (Figures 3c and 3d). We also note that when the reversed polarity component is weak, it can be almost or completely masked by a dominant component of normal polarity (Figures 3a and 3b). This behavior indicates a significant overlap of coercivity spectra for the magnetic carriers of the two magnetization components, resulting in unresolved NRM components and the potential for the erroneous assignment of directions of characteristic remanent magnetization.
 We note that this magnetic behavior was missed in a prior analysis of basalt cores from ODP Site 883 [Sager, 2002] that relied on AF demagnetization. The two-component nature of the magnetization, and overlapping coercivities clearly preclude the use of AF demagnetization data to constrain paleomagnetic directions or paleosecular variation for the basalt of Site 883. Therefore below we will limit discussion of the magnetization components to those derived from thermal demagnetization data where the components can be properly identified and isolated.
4. Magnetization Components
4.1. Correlation of Component Directions
 The reversed polarity, low unblocking temperature component would normally be interpreted as a thermoviscous or chemical magnetic overprint acquired during a reversed polarity interval. The normal polarity component (mean inclination I = 57.7°−13.8°+12.3°, n = 5 inclination units, see section 5.1) is slightly steeper than the reversed polarity component (I = −53.1°−9.5°+10.2°, n = 5, using the same inclination units as defined by the normal polarity component). This small difference, however, is not significant at the 95% confidence level for more than 50% of the samples and might reflect a minor normal polarity viscous overprint acquired during the last 0.7 Myr, which has not been resolved in demagnetization data. Moreover, the directional changes of the low unblocking temperature component through the Site 883 rock section follow closely those of high unblocking temperature component: the steeper the normal polarity component, the steeper the reversed component and vice versa (Figure 4). Such correlation is not what one could expect if the NRM components were independent (e.g., primary magnetization and a later overprint). This, together with nearly antipodal directions of the two components, suggests a partial self-reversal of magnetization [Doubrovine and Tarduno, 2002].
 A detailed rock magnetic study of the Site 883 (and Site 1204) basalts has confirmed the suggested partial self-reversal [Doubrovine and Tarduno, 2004]. In section 4.2, we briefly summarize the salient points of this work as they lay the foundation for further geophysical interpretations of the NRM components.
4.2. NRM Carriers
4.2.1. Magnetic Mineralogy
 Examinations of polished thin sections using reflected light and scanning electron microscopy reveal that the magnetic mineralogy of ODP Site 883 basalts is represented by fine grained (<1 μm to several tens of μm), usually skeletal grains of titanomagnetite (Fe3−xTixO4, 0 ≤ x ≤ 1 is the ulvöspinel content) altered by various degrees to titanomaghemite:
where R = 8/[8 + z(1 + x)], [ ] represents a lattice vacancy, and z is the oxidation parameter [O'Reilly, 1984]. No hematite, hemoilmenite or other magnetic phases were observed. This mineralogy is similar to that described from the basalt recovered at ODP Site 1204 [Tarduno et al., 2002].
4.2.2. Rock Magnetic and Compositional Data
 Magnetic hysteresis loops measured using a Princeton Measurements Corporation Alternating Gradient Force Magnetometer at the University of Rochester show that the bulk magnetic properties of Site 883 basalts straddle the single-domain to pseudo-single-domain boundary [Day et al., 1977; Dunlop, 2002]: 0.3 ≤ Mrs/Ms ≤ 0.7; 1.3 ≤ Hcr/Hc ≤ 1.8.
 High-field (Ms(T)) thermomagnetic data (provided courtesy of Professor Friedrich Heller at the Institute for Geophysics, ETH Zürich) and magnetic susceptibility versus temperature (κ(T)) curves measured with an AGICO KLY-4S Kappabridge equipped with a furnace (at the University of Rochester) showed Curie temperatures that range between 275 and 430°C. Inversion of titanomaghemite [O'Reilly and Readman, 1971; Özdemir, 1987] was observed above ∼350°C. Through energy-dispersive X-ray spectroscopy (using a Leo 982 scanning electron microscope equipped with an Edax Phoenix EDS system at the University of Rochester) we observed x values from 0.65 to 0.86. X-ray powder diffraction studies of magnetic separates (using a Philips Multipurpose Diffractometer at the University of Rochester) revealed a 8.344–8.354 Å range of lattice parameter for the samples which showed the two NRM components and a 8.358–8.362 Å range for the samples with a single component magnetization. This suggests higher oxidation states of samples carrying the two NRM components [Readman and O'Reilly, 1972; O'Reilly, 1984], although the difference is comparable with the uncertainty of lattice parameter determinations (typically 0.008 Å at the 95% level). These data, together with the observed Curie points constrain the oxidation states of titanomaghemite to high z values (z ≥ 0.8, Figure 5) [Readman and O'Reilly, 1972; Nishitani and Kono, 1982].
4.2.3. Ionic Reordering and N-Type Thermomagnetic Behavior of Titanomaghemite
 Ionic reordering is a process of migration of cation vacancies created on tetrahedral crystallographic sites (A) into the octahedral positions (B) at high oxidation states. This is accomplished by the diffusion of iron cations from B to A sites, reducing the magnetization of the B sublattice and (slightly) increasing the magnetization of the A sublattice. During low-temperature oxidation, a titanomagnetite with a dominant B sublattice magnetic moment can thus be transformed to a titanomaghemite with a dominant A sublattice moment, producing a self-reversal of spontaneous magnetization (Ms). Remanent magnetization, if controlled by a contribution from single-domain grains, will also self-reverse (Figure 6).
Schult [1968, 1971] pointed out that the transformation from B- to A-sublattice dominated magnetization should be accompanied by a Q → P → L→ N → Q′ transition of Ms(T) curve types [Néel, 1948] and that the titanomaghemites which undergo a self-reversal by ionic reordering should exhibit either N-type curves with compensation points (Tk) above room temperature (Figure 5) or Q′-type Ms(T) curves.
 To test for N-type behavior, we measured the temperature dependences of hysteresis parameters at the Institute for Rock Magnetism (University of Minnesota) in a −263–347°C range using a Princeton Measurements Corporation Vibrating Sample Magnetometer (VSM) equipped with a cryostat (measurements from −263 to 127°C) and a furnace (measurements from 97 to 347°C). Combining the data, we observed that the samples carrying one NRM component have broad N-type minima of Ms with compensation temperatures distributed below 0°C and P-type Ms(T) curves. Samples carrying two NRM components showed even broader N-type minima with compensation points distributed above and below room temperature.
 The presence of magnetic carriers with N-type thermomagnetic behavior and compensation points above room temperature is further supported by self-reversals of partial thermoremanent magnetization (pTRM) given in a laboratory field of 40 μT, which were observed on samples carrying two components of magnetization at temperatures between 250 and 350°C. The temperatures at which self-reversed pTRM is blocked are below those characterizing the initial stage of titanomaghemite inversion and correspond to the unblocking temperatures of the reversed polarity NRM component.
4.3. Interpretation of Magnetization Components
 The compositions of titanomaghemite from the ODP Site 883 basalt (0.65 ≤ x ≤ 0.86, z ≥ 0.8; section 4.2.2) and observations of the N-type thermomagnetic behavior (section 4.2.3) suggest that ionic reordering occurred in the titanomaghemite during low-temperature oxidation, resulting in a partial self-reversal of magnetization [O'Reilly and Banerjee, 1966]. Therefore we interpret the magnetization components as follows.
 Both magnetization components of Site 883 basalts are carried by single- to pseudo-single-domain grains of titanomaghemite and can be classified as chemical remanent magnetizations (CRMs) produced by low-temperature oxidation [Dunlop and Özdemir, 1997]. The normal polarity CRM is carried by titanomaghemites of P- and N-type with compensation points below room temperature; these magnetic grains carry a CRM which inherits the direction of the primary thermoremanence (TRM) [e.g., Özdemir and Dunlop, 1985; Brown and O'Reilly, 1988] (Figure 6). Titanomaghemites of N-type with compensation points above room temperature have undergone a transformation from a magnetic moment dominated by B sublattice to an A sublattice dominating magnetization due to ionic reordering; these grains carry a self-reversed CRM (reversed polarity component), in a direction opposite to the primary TRM (Figure 6).
 The lower unblocking temperatures of the reversed polarity component are probably related to smaller grain sizes (which would facilitate higher oxidation degrees required for the creation of N-type behavior with high compensation temperatures) and/or to higher Ti contents of titanomaghemite grains carrying self-reversed CRM. Unblocking temperatures of the normal polarity component overlap with the temperatures of titanomaghemite inversion. However, in the absence of an ambient field (i.e., in our AST TD-48 thermal demagnetization device) the magnetization of the newly created inversion product is entirely controlled by the preexisting NRM [e.g., Marshall and Cox, 1971]. The inversion product picks up the magnetization of the normal polarity component and no change in the direction of remanence occurs.
 Therefore, although it is unlikely that an original TRM is present, we can consider the normal polarity NRM component as a true characteristic remanent magnetization (ChRM) representing the direction of primary magnetization. This interpretation is consistent with data from Site 1204, where the normal polarity component was interpreted as the primary magnetization acquired during chron 32/33N because of a single magnetization component seen in the least altered/reduced samples from a 46-m-thick diabase unit intercalated with altered pillow basalts [Tarduno et al., 2003].
5. Analysis of ChRM Inclination Data
5.1. Inclination Groups
 Inclination-only analysis of azimuthally unoriented ChRM directions [McFadden and Reid, 1982] was used to calculate average inclination for the Site 883 and estimate angular dispersion. One of the concerns in obtaining an averaged paleomagnetic field direction from a basalt section is the uncertain times between the emplacement of lava flows. In the case of rapid eruptions, several adjacent lava flows can record the same “snapshot” of the geomagnetic field, leading to a biased average if equal statistical weight is assigned to each flow unit. Another potential problem for the Site 883 basalts is the uncertainty as to whether the 27 lithologic units identified in the section are all separate flows or lava lobes comprising a smaller number of flows (see section 2).
 To address these problems we first calculated the flow-mean inclinations for each lithologic unit (Table 1). Then, we checked whether the inclination-only averages from adjacent units are distinguishable at the 95% confidence level using a Z statistic [Kono, 1980]. If adjacent means did not differ at 95% confidence level, the data were combined and tested against an average from the next adjacent flow. Repeating this procedure, we define 5 inclination groups for Hole 883E and 3 inclination groups for Hole 883F (Figure 7, Table 2).
Note: n, number of samples; I, flow-mean inclinations; ΔI95, 95% confidence interval; k, best estimate of precision parameter [McFadden and Reid, 1982]. Inclinations in italics are nominal values for flow units with n ≤ 2.
Table 2. Inclination Groups
Calculated using combined inclination groups A–C and Hole 883E groups D and E.
 The hole-mean inclinations calculated using the inclination groups are nearly identical: I = 57.7°−13.9°+12.4° (n = 5, k = 47) for Hole 883E and I = 57.5°−20.7°+20.2° (n = 3, k = 102) for Hole 883F (95% confidence limits are quoted). Taking into account a 3.9 m offset in basement depths between Holes 883E and 883F [Rea et al., 1993], inclination groups from the two drill holes show a remarkable correlation (Figure 8). Average inclinations of the 3 groups from Hole 883F are indistinguishable at the 95% confidence level from those of the 3 upper groups from Hole 883E (groups A–C, Table 2). Data from the two holes can thus be combined. Site-mean inclination average calculated using the 3 combined inclination groups (A–C) and the 2 groups from the bottom of Hole 883E (D and E), which were not sampled in a shorter section from Hole 883F, is I = 57.7°−13.8°+12.3° (n = 5, k = 48) (Table 2). This value is exactly the same as the inclination average from the Hole 883E data alone, and its precision does not show any noticeable improvement when Hole 883F data are incorporated.
 The inclination mean of Sager  (61.5°−6.4°+10.6°), obtained from the same basalt cores we have studied, is within error of our new estimate. However, we note that the flow-mean inclinations and the subdivision of the Site 883 sequence into inclination units differ in the two studies. We attribute these differences to the reliance on AF demagnetization (83% of the data) in the Sager  study. The subset of thermal demagnetization data of Sager  is too limited in the number of samples analyzed (22 samples) and the demagnetization steps applied to allow a meaningful analysis of inclination units which could be directly compared with our results. Because of the problems associated with isolation of magnetization components in AF demagnetization data (section 3), detailed thermal demagnetization data (i.e., our 5-inclination group model) are required to accurately represent the structure of ChRM inclination in the Site 883 basalt sequence.
5.2. Paleolatitude and Angular Dispersion
 Our model with 5 inclination groups suggests a paleolatitude of 38.3°−12.6°+15.6° for Site 883. To check whether the site-mean inclination represents the time-averaged geomagnetic field, and hence whether the Site 883 basalt sequence adequately samples secular variation, the angular dispersion of the paleomagnetic directions was calculated and compared to global data from lava flows [McFadden et al., 1991].
 The best estimate of the precision parameter [McFadden and Reid, 1982] for the 5-inclination group model (k = 48) was transferred into pole space using the formulation of Cox :
where λ is the paleolatitude. This approach assumes that the parent virtual geomagnetic poles (VGPs) form a Fisherian distribution [Fisher, 1953]. The estimated precision parameter for the poles (K = 32) was next used to calculate the polar angular dispersion (S = 14.7°) and the directional angular dispersion (s = 11.7°) using [Cox, 1970]
where S is expressed in degrees (°) and
 The angular dispersion of virtual geomagnetic poles is available from global data on paleosecular variation of lavas (PSVL). We use a secular variation model where the geomagnetic field is composed of independent dipole (antisymmetric) and quadrupole (symmetric) families (model G of McFadden et al. ). The model is only correct as an approximation, but it is nevertheless attractive because it relates secular variation to results from dynamo theory. The scatter of VGPs predicted by model G can be expressed as [McFadden et al., 1991]
where Ss is the contribution from the secondary (or quadrupole) family and Sp is the contribution from the primary (or dipole) family. For a time window relevant to our study (45–80 Ma), Ss = 9.7° ± 1.5° and Sp/λ = 0.34 ± 0.03 [McFadden et al., 1991]. These values yield a predicted S = 16.2° ± 5.7° for the paleolatitude of Site 883. Our estimate of VGP dispersion (S = 14.7°) based on the Site 883 data is within error of the expected value, suggesting the basalt sequence from Site 883 adequately samples secular variation. The 5-inclination group average thus seems to record the time-averaged Late Cretaceous geomagnetic field.
6.1. Comparison With Other Data Sets From Detroit Seamount
 Site 1204 is only ∼0.5 km away from Site 883 (Figure 1). In thermal demagnetization data from basalts recovered from that site, 5 independent inclination groups were identified on the basis of the lithofacies succession, resulting in an average inclination of 60.1°−8.3°+6.8°. This value is indistinguishable from our Site 883 average, however the polar angular dispersion (S = 3.1°) is much lower than the value predicted from the global PSVL data [McFadden et al., 1991].
 It seems counterintuitive that a shorter sequence from Site 883 (∼38 m) spans a time interval sufficient to average paleosecular variation, while the ∼140-m-thick section from Site 1204 shows much lower variation between the paleomagnetic inclination groups. In the Site 1204 section, the upper part of basaltic sequence (Unit 2 from Hole 1204A, 821–863 mbsf, and Units 1–2a, from Hole 1204B, 814–872 mbsf, which roughly correspond to the Site 883 sequence, 819–850 mbsf) is described as consisting of thick compound pahoehoe lavas composed of multiple sheet lobes. The lower part (Units 2b, 2c and 3 from Hole 1204B) was not sampled at Site 883. This is obvious because Hole 1204B Unit 2b (872–918 mbsf) is a large (46-m-thick) pahoehoe inflated sheet lobe [e.g., Self et al., 1998; Thordarson and Self, 1998], which shows a diabase texture quite different from what was observed in the Site 883 section.
 Large fractures within the basalt from Site 883 are filled with a calcareous sedimentary rock consisting of calcite, clays, plagioclase and quartz. This sedimentary fill is also present in gaps between the adjacent lava flows, suggesting that significant time intervals may have elapsed between the events of lava accumulation. Similar sediments, filling the cooling cracks of lava and cementing the decimeter-tick intervals of lapilli breccia intercalated with basalt lobes, were observed in Hole 1204B, Unit 1. However, because of the poor recovery (<40%), no attempt was made to subdivide the basalt into subunits. Therefore we speculate that Unit 1 from Hole 1204B may represent several compound pahoehoe flows rather than a single eruption unit. This may also be true for other thick pahoehoe lavas from Site 1204 (Hole 1204A, Unit 2; Hole 1204B, Units 2a and 2c). If correct, this would imply that the paleomagnetic inclinations calculated from the thick pahoehoe lava units from Site 1204 may have already averaged paleosecular variation to a certain extent, resulting in a fairly accurate inclination average but an artificially low estimate of angular dispersion. Alternatively, we might consider that the basalt sequence from Site 883 consists of a single compound pahoehoe flow emplaced on a timescale of years to decades [Thordarson and Self, 1998]. However, this alternative contrasts with our observation of distinct paleomagnetic inclination groups (Figures 7 and 8).
6.1.2. Site 884
 Site 884 was drilled on the lower eastern flank of Detroit Seamount during ODP Leg 145 (Figure 1). The basement was penetrated in one drill hole (Hole 884E: 51°27′N, 168°20′E) where a 87-m-thick section of pillow lavas and massive flows was sampled. The age of Site 884 basalt, constrained by 40Ar/39Ar radiometric data, is 81 ± 1 Ma [Keller et al., 1995]. Thermal demagnetization data from the 13 lava flow units defined in Site 884 cores showed a single NRM component of reversed polarity after removal of a small viscous overprint. The reversed polarity ChRM was interpreted as primary magnetization acquired during chron 33R. Inclination data analysis resulted in the 10 inclination groups (distinct at the 95% confidence level), which led to an average of I = −55.7°−6.2°+7.7° and a paleolatitude λ = 36.2°−7.2°+6.9°. The estimated polar angular dispersion (S = 15°) is indistinguishable from that predicted from global PSVL data [McFadden et al., 1991] for 45–80 Ma and 80–110 Ma age intervals, suggesting that the basalts from Site 884 average secular variation. The absolute value of Site 884 inclination average is not distinct at 95% confidence level from our estimate from Site 883. The precision parameters of two data sets are very similar; a smaller confidence interval for the Site 884 inclination results seem to result solely from the larger number of inclination groups. Thus the new data from Site 883 confirm the results of Tarduno and Cottrell  and Cottrell and Tarduno .
6.1.3. Site 1203
 Site 1203 (50°57′N, 167°44.4′E) is located in an area of flat basement covered by 462 m of sediments on the summit of Detroit Seamount, ∼30 km south from Sites 883 and 1204 (Figure 1). A 453-m-thick sequence of 18 compound lava flows intercalated with the 14 sedimentary interbeds was penetrated at this site, making it the longest basement section drilled at Detroit Seamount. The age of Site 1203 basalts is constrained to 75–76 Ma by both the biostratigraphy of sediments within the basement sequence and 40Ar/39Ar dating [Tarduno et al., 2003]. In thermal demagnetization data, most basalt and sediment samples showed a simple univectorial decay of magnetization of normal polarity after removal of a small, spurious overprint at low unblocking temperatures. Two antipodal NRM components, similar to those seen in Site 883 and Site 1204 data, were observed in the two uppermost pahoehoe units (Units 1 and 3) which appear more weathered that the deeper basalts.
 Sixteen inclination groups were identified for the Site 1203 basalt samples, suggesting an average value of I = 48.6°−10.6°+7.0°. Results from AF demagnetization data (where Units 1 and 3 were excluded due to the 2-component behavior) yielded similar value (I = 50.0°−10.6°+7.3°, n = 14). The mean value derived from thermal demagnetization of the Site 1203 basalts is some 9° shallower than the mean Site 883 value, although the uncertainty limits are relatively large for both estimates. The inclination average from the 14 sedimentary units within the Site 1203 basalt sequence (thermal demagnetization: I = 53.2°−11.4°+5.0°; AF demagnetization: I = 54.7°−6.4°+3.1°) is ∼4–5° steeper than the basalt mean and closer to the Site 883 value.
 It is not clear why the inclination from the Site 1203 basalts is shallower than those from the contemporary (75–76 Ma, Sites 883 and 1204) and older (81 Ma, Site 884) basalts, which show a remarkable consistency (average inclination values are within a range of ∼4°). The 8–11° difference can not be attributed to a borehole tilt, which, in all cases, appears to be small (≤2°) [Tarduno and Cottrell, 1997; Tarduno et al., 2002]. Neither can it result from paleosecular variation of geomagnetic field (angular dispersion of Site 1203 basalt data, S = 18.4°, is slightly higher than but within the uncertainty limits of a value expected from global PSVL data). Tarduno et al.  noted that the mean inclination from the Site 1203 sediments should be a minimum estimate because of potential inclination shallowing. A lower value derived from the basalt data suggests that the lavas available from Site 1203 may have fortuitously underrepresented high inclination values of the Late Cretaceous geomagnetic field.
6.2. Paleolatitude of Detroit Seamount
 Five thermal demagnetization data sets are now available from Detroit Seamount: the basalt data from Sites 883 [this study], 884 [Tarduno and Cottrell, 1997], 1203, 1204 [Tarduno et al., 2003] and data from the Site 1203 sediments [Tarduno et al., 2003]. In this section we will combine the data to calculate the average inclination for Detroit Seamount and estimate the paleolatitude of Hawaiian hot spot at ∼75–81 Ma.
 Following Tarduno et al. , two approaches can be considered. First, we assume the Site 1203 basalts underrepresented the high inclination values and thus calculate a mean using inclination averages from the basalts of Sites 883, 884 and 1204 (Hole B) and from the sediments of Site 1203. This approach results in an inclination average of I = 56.8°−6.2°+6.0° (n = 4) and a paleolatitude of λ = 37.4°−6.1°+6.8° (paleolatitude model C). The second, and perhaps more conservative approach, is to use the individual basalt inclination groups from the four sites (5 from Site 883, 10 from Site 884, 16 from Site 1203 and 5 from Site 1204) and to average group-mean values assigning equal statistical weight to each group. This analysis results in an average I = 53.9°−6.4°+3.2° (n = 36) and a paleolatitude λ = 34.4°−5.8°+3.2° (paleolatitude model D). Comparison with the original paleolatitude A and B models of Tarduno et al.  (which were calculated using the same method as C and D, respectively, but without the data from Site 883; model A: I = 56.5°−12.6°+12.4°, n = 3, λ = 37.1°−11.4°+15.3°; model B: I = 52.9°−6.9°+3.7°, n = 31, λ = 33.5°−6.1°+3.7°) shows that the incorporation of Site 883 data does not strongly affect the average inclination and paleolatitude in each model. However, the incorporation of the Site 883 data considerably reduces the uncertainty limits in model A.
 Paleolatitude estimates derived from either model are significantly north of the present latitude of Hawaiian hot spot (∼19°N) and suggest rapid average rates of southward hot spot motion between 81 and 47 Ma (assuming the 47 Ma bend of Hawaiian-Emperor chain was formed due to cessation of rapid hot spot migration) of 66.4−22.0+24.5 mm/yr (Model C) or 55.6−23.1+11.6 mm/yr (Model D). Both values are consistent with previously published estimates [Tarduno and Cottrell, 1997; Tarduno et al., 2003] as well as with independent rates estimated using global plate circuits [Raymond et al., 2000].
 Basalts from ODP Site 883 carry two nearly antiparallel component of magnetization. Both component are CRMs carried by highly oxidized, single-domain to pseudo-single-domain titanomaghemite. A series of rock-magnetic experiments suggests that the normal polarity component of NRM is a CRM that inherited its direction from a primary TRM while the reversed polarity component is a CRM self-reversed with respect to the parental TRM.
 Inclination data analysis of the normal polarity characteristic remanent magnetization from the Site 883 basalt results in 5 statistically distinct inclination groups and average inclination of 57.7°−13.8°+12.3°. The angular dispersion of paleomagnetic direction suggests that the basalt sequence averages paleosecular variation.
 The high sample coverage of the Site 883 basalt section may have allowed us to obtain a more detailed representation of secular variation of the geomagnetic field direction than that observed from a nearby Site 1204. In contrast, undersampling (both through core gaps and paleomagnetic) at Site 1204 may have led to internal averaging of secular variation. As a result, the Site 1204 inclination data could represent an accurate inclination mean, but an underestimate of paleosecular variation.
 Incorporation of the new paleomagnetic data from Site 883 in previous paleolatitude models for Detroit Seamount, based on data from Sites 884, 1203 and 1204 [Tarduno et al., 2003], results in an inclination average of 53.9°−6.4°+3.2° and a revised paleolatitude of 34.4°−5.8°+3.2°. This value suggests a ∼33–67 mm/yr rate of southward motion of Hawaiian hot spot from Late Cretaceous (75–81 Ma) to Middle Eocene (∼47 Ma) times, supporting the earlier estimates of hot spot migration rate [Tarduno and Cottrell, 1997; Tarduno et al., 2003].
precision parameter for a Fisherian distribution of paleomagnetic directions.
precision parameter for a Fisherian distribution of virtual geomagnetic poles.
coercivity of remanence.
saturation remanent magnetization.
number of samples or inclination groups.
angular dispersion of paleomagnetic directions.
angular dispersion of virtual geomagnetic poles.
contribution of primary geodynamo family to the overall scatter of virtual geomagnetic poles.
contribution of secondary geodynamo family to the overall scatter of virtual geomagnetic poles.
compensation temperature for titanomaghemite with N-type thermomagnetic behavior.
ülvospinel content of titanomagnetite.
oxidation parameter of titanomaghemite.
 This research used samples provided by the ODP. The ODP is supported by the U.S. National Science Foundation and participating countries under management of the Joint Oceanographic Institutions. We thank Alexei Smirnov, Rory Cottrell, Subir Banerjee and Kenneth Hoffman for discussions, and Allyson O'Kane for help in paleomagnetic measurements. This work was supported by the U.S. National Science Foundation.