As part of a larger project investigating paleosecular variation in South America, 36 paleomagnetic and geochronology sites were sampled on Meseta del Lago Buenos Aires (47°S, 289°E), 200 km east of the active plate boundary in southern Patagonia. Basaltic lavas sampled range from late Miocene to late Pleistocene yet most of the individual lavas are younger than 3.3 Ma. Thirty-eight isochron ages determined via 40Ar/39Ar incremental-heating experiments frame a detailed stratigraphy. The isochron ages range from 66.9 ± 4.1 ka for scoria cones and a youthful lava flow in the Rio Pinturas valley to 10.12 ± 1.35 Ma for thick plateau-forming lavas exposed along the SE edge of the Meseta. The last 3.3 myr is characterized by seven episodes of volcanism at ca. 3.2–3.0 Ma, 2.4 Ma, 1.7 Ma, 1.35 Ma, 1.0 Ma, 750 ka, 430–330 ka, and <110 ka. The bulk of lavas forming the surface of the Meseta erupted in the last 1.7 my. All sites have stable magnetization, and after step demagnetization using either thermal or alternating field techniques, yield characteristic directions held by magnetite and/or titanomagnetite. Ten sites have distinct transitional directions (defined by pole latitudes <55°), and associated ages indicate possible connections to known reversals within the Matuyama Chron, including the onset of the Jaramillo subchron (1.016 ± 0.01 Ma), the Cobb Mountain subchron (1.25 ± 0.03 Ma), the Ontong-Java 1 event (1.37 ± 0.03 Ma), and the termination of the Olduvai subchron (1.72 ± 0.02 Ma). Remaining sites are divided into normal (14 sites) and reversed polarity (12 sites). At 95% confidence mean normal and reversed directions overlap. The mean direction for 26 sites is I = −63.0°, D = 3.4°, α95 = 5.4°, which is indistinguishable from the expected geocentric axial dipole. Paleosecular variation, measured by the dispersion of virtual geomagnetic poles about the rotation axis, is higher than expected at 20.0°. Model G, using data for the past 5 Myr, predicts a dispersion of 17°. This discrepancy may be due to true dispersion of the field in southern South America or it may be an artifact of inadequate sampling.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The magnetic field of the Earth is known to vary over time periods ranging from milliseconds to millions of years. In an attempt to fully understand aspects of the time-averaged field (TAF), considerable effort has gone into studying the paleosecular variation of the field over millions of years. Research in this area falls into several categories including: detailed studies of magnetic field behavior as presented by paleomagnetic field directions at one location on the Earth's surface, modeling experiments concerned with the global representation of the magnetic field, and theoretical studies of field behavior including expected normal variations.
 Paleosecular variation studies on lava flows (PSVL) use the instantaneous nature of magnetization in lava flows to sample the Earth's field at discrete positions over long time periods. From the early 1960s onward models have been developed and refined to investigate the variation of PSV over latitude [Merrill et al., 1996]. Recently, a number of studies have investigated TAF by applying global data sets of PSVL to field models [Johnson and Constable, 1996, 1997, 1998; McElhinny and McFadden, 1997]. McElhinny and McFadden  suggest that many older individual studies are inadequate for current needs, although remeasurement of some of those earlier studies have verified the original results [Brown, 2002]. In any event, it is necessary to produce not only accurate results but also to increase geographic coverage to constrain field behavior globally. This is especially germane to the Southern Hemisphere, where fewer studies of high quality exist.
 In this paper we present paleomagnetic and geochronologic results from Meseta del Lago Buenos Aires, Patagonia, a Miocene - Pleistocene volcanic field at 47°S in the southern part of South America. This study provides important new data on the behavior of the magnetic field over the last few million years from the Southern Hemisphere for comparison to other studies in the Southern Hemisphere, to results from similar latitudes in the Northern Hemisphere, and to provide well-dated and determined magnetic data for incorporation into global models of the field. All radioisotopic ages reported or discussed here are relative to standard values of Renne et al.  and the ages of geomagnetic chrons and their boundaries [e.g., Cande and Kent, 1995; Berggren et al., 1995] have been adjusted accordingly.
2. Geologic Setting
 Meseta del Lago Buenos Aires (MLBA) is a basaltic plateau that has a total area of ∼5000 km2 and is located in the northwestern corner of the Santa Cruz Province, Argentina, between 46.25°–47.25°S and 288.25°–289.50°E (Figures 1 and 2). The MLBA is a quasi-planar plateau with an ambient elevation of ca. 1400 masl, that is dotted with more than 100 scoria cone vents, many of which attain peak elevations of over 1600 mal [Baker et al., 1981; Mercer and Sutter, 1982; Singer et al., 2004] (Figure 2). The Monte Zeballos volcanic complex near the southwestern corner of MLBA rises to 2700 masl (Figure 2). The MLBA comprises Cretaceous and Tertiary sediments capped by 10–30 m thick, columnar-jointed, laterally extensive tholeiitic lavas that we will show below were erupted between ca. 11 and 7 Ma, and which in turn are perforated and overlain by voluminous Pliocene to Quaternary alkaline lavas and intrusions (Figure 2) [Baker et al., 1981; Gorring et al., 2003]. These rocks rest unconformably over Jurassic rhyolites and upper Proterozoic to Devonian low-grade metamorphic basement [Hervé et al., 1981; Pankhurst et al., 1998].
 MLBA lies in the cool rain shadow of the Andes, thus the climate is arid, rates of weathering and erosion are relatively low, and remarkably fresh exposures of numerous lava flows are preserved. Sites for paleomagnetic study were sampled in 1998 and 2001 using a gasoline powered drill, with 6 to 10 samples drilled over several square meters of exposed lava at each site. Sample orientations were collected using sun and magnetic compasses and locations established by GPS and plotting on regional topographic maps. Samples for 40Ar/39Ar dating were either drilled or chiseled from outcrops at the same locations as where paleomagnetic cores were taken. To the north and east of MLBA access was via national highway 40, whereas the rugged interior, the southwest margin of the plateau, and the steep western side are accessible only via four-wheel drive vehicle and traverses on foot (Figure 2).
 A total of 36 sites were drilled for paleomagnetic studies; all sites but two have associated geochronologic samples available for study. The two undated sites are from flows that are in direct stratigraphic contact with dated flows. Sites are concentrated in the plateau-capping flows, individual volcanic cones, and flows that spilled over the plateau into the broad valley to the east, although a small number of sites were collected in the older plateau-forming flows.
4. Paleomagnetic Methods
 Magnetic data was processed in the Paleomagnetism Laboratory at the University of Massachusetts. Samples were measured on a 2G cryogenic magnetometer, model 755R, with alternating field and thermal demagnetization done using a Molspin AF demagnetizer and an ASC thermal demagnetizer, respectively. Susceptibility measurements were done using a Sapphire susceptibility meter.
 Paleomagnetic samples were subjected to detailed demagnetization using both alternating field (AF) and thermal techniques. AF demagnetization was performed in steps from 0 to 100 mT, with 10 to 12 steps per sample. Thermal demagnetization, done at 10 temperature steps from room temperature to 600°C, was used for a minimum of two samples from each flow. Characteristic directions were determined for both demagnetization methods using line-fitting techniques [Kirschvink, 1980] using at least 4 points for each calculation and obtaining maximum angles of deviation of less than 5°.
5. The 40Ar/39Ar Methods
 At the University of Wisconsin-Madison incremental heating experiments used a metal resistance furnace to degas ca. 25 mg aliquots of holocrystalline groundmass in 5 to 11 steps between 600° and 1375°C. Complete degassing of the feldspar separate from sample PAT-56 required 1675°C. Procedures for preparing and irradiating samples, extracting gas from samples and fluence monitors, correcting for blanks, mass spectrometry, and estimating analytical uncertainties are detailed in Singer et al.  and Singer and Brown . Results obtained using identical procedures on several of the lava flows including PAT-2, PAT-3, PAT-5, and PAT-6, are fully reported in Singer et al. , whereas Ackert et al.  summarize the multiple experiments on groundmass from samples of sites PAT-49 and PAT-29 on the Rio Pinturas basalt flow.
 The 40Ar/39Ar method requires that the age of a sample be calculated relative to a mineral standard which has been previously dated, usually by conventional K-Ar techniques. The monitor mineral used here was sanidine from the Alder Creek rhyolite (ACs). Turrin et al.  measured the age of ACs at 1.186 Ma relative to 27.84 Ma sanidine from the Fish Canyon tuff. In light of recent intercalibration of these two 40Ar/39Ar standards relative to 98.79 ± 0.96 Ma GA-1550 biotite, we have adopted an age of 1.194 Ma for ACs [Renne et al., 1998]; however, a consensus regarding the age of the GA-1550 standard has not yet been reached [Lanphere and Dalrymple, 2000]. We note that adopting an age of 1.186 Ma for the ACs standard, equivalent to assuming an age for Fish Canyon tuff sanidine of 27.84 Ma that is compatible with the timescales of Cande and Kent  and Berggren et al.  would shift the ages reported here about 1% younger, but would not alter any of our conclusions. Conversion of our ages to make them consistent with other values for standard ages, e.g., Villeneuve et al. , can be done simply using the equation in Dalrymple et al. . Samples were irradiated for one hour adjacent to ACs monitors in evacuated quartz vials at the Oregon State University Triga reactor in the Cadmium-Lined In-Core Irradiation Tube (CLICIT). Corrections for undesirable nucleogenic reactions on 40K and 40Ca are: [40Ar/39Ar]K = 0.00086; [36Ar/37Ar]Ca = 0.000264; [39Ar/37Ar]Ca = 0.000673 [Wijbrans et al., 1995].
 Inverse-variance weighted mean plateau ages and uncertainties are calculated according to Taylor . Precision estimates for monitors, based on 5–6 measurements each suggest that the uncertainty in J, the neutron fluence parameter, was typically 0.6% (±2σ); this uncertainty was propagated into the final plateau and isochron age for each analysis, but contributes <0.1% to the total uncertainty in these age estimates. Ages were calculated using the decay constants of Steiger and Jäger  and are reported with ±2σ analytical uncertainties.
 Criteria used to determine ages were: (1) age plateaus are defined by at least three contiguous steps all concordant in age at the 95% confidence level and comprising >50% of the 39Ar released, (2) a well-defined isochron exists for the plateau points as defined by the F-variate statistic SUMS/(N − 2) [York, 1969], and (3) the isochron ages [York, 1969] are preferred over the weighted mean plateau ages because they combine estimates of analytical precision plus internal disturbance of the sample (scatter about the isochron) and they make no assumption about the trapped argon component. Where multiple subsamples were measured from a lava, the isochron combining all the plateau steps gives our best estimate of the age of the flow [Singer et al., 2002, 2004].
6. Paleomagnetic Results
 Natural remanent magnetization (NRM) and demagnetization behavior was measured on a total of 274 samples from 36 sites on or near Meseta del Lago Buenos Aires. Both alternating field and thermal demagnetization techniques were used on samples from the same site with consistent agreement between the two methods (Figure 3). AF demagnetization on normal flows removed only minor overprints at low levels (Figure 3a) while reversed flows showed larger overprints, but ones that were removed by fields of 20 to 40 mT (Figure 3c). Thermal demagnetization on most samples showed similar responses with small overprints removed from the normal flows by 300°C (Figure 3b) and large overprints removed from the reversed flows at levels of 400 to 500°C (Figure 3d). Samples from sites having transitional directions also showed small but varying amounts of overprinting and then straight-line decay to the origin with both AF and thermal techniques (Figure 3e and 3f).
 Susceptibility measurements made on each core and averaged for each site are shown in the histogram in Figure 4a. Values range from 0.55 × 10−2 to 9.51 × 10−2 SI, with a mean of 2.23 × 10−2 SI. NRM values before demagnetization range from 1.08 A/m to 8.00A/m (Figure 4b), with a mean of 4.03 A/m. Sites show no indication of high magnetizations (>20 A/m) associated with lightning strikes [Cox, 1961; Tauxe et al., 2003] nor do any show exceedingly low magnetizations (<1 A/m) as might be associated with times of low field strength such as excursions or reversals.
 Inclinations and declinations for all the flows, along with associated statistics and virtual geomagnetic poles (VGP) are given in Table 1 and plotted in Figure 5a. Most flows show excellent within site statistics, precision parameter (k) and 95% circles of confidence (α95) of Fisher . A histogram of α95 (Figure 4c) indicates that only 6 flows have α95 values greater than 5.5°, with a mean value of all flows of 4.9° and a range of 2.6° to 13.9°.
Table 1. Summary of Site Mean Paleomagnetic Data on Lavas From Meseta Lago Buenos Airesa
Age (Ma) ± 2sigma
N, number of samples used in site calculations; Polarity, direction of the field; N, normal; R, reversed; T, transitional, selection described in text; a-95; and K, radius of 95% cone of confidence about the mean direction and precision parameter [Fisher, 1953].
1.016 ± 0.010
0.109 ± 0.006
0.760 ± 0.014
1.074 ± 0.028
0.984 ± 0.035
1.290 ± 0.030
0.070 ± 0.024
0.743 ± 0.012
1.490 ± 0.050
7.710 ± 0.070
0.754 ± 0.114
1.143 ± 0.064
0.327 ± 0.016
1.720 ± 0.020
0.341 ± 0.033
1.730 ± 0.020
0.336 ± 0.012
0.433 ± 0.012
7.860 ± 0.160
0.752 ± 0.030
1.360 ± 0.050
1.341 ± 0.103
1.370 ± 0.030
1.390 ± 0.040
1.340 ± 0.030
1.315 ± 0.029
0.067 ± 0.004
10.120 ± 1.350
3.030 ± 0.030
2.422 ± 0.035
3.180 ± 0.040
1.709 ± 0.026
3.290 ± 0.220
1.250 ± 0.031
 In recent years strict criteria have been suggested for determining sites suitable to represent the time averaged field [Johnson et al., 1998; Tauxe et al., 2000]. One important criterion is that α95 values must be <5° (rounded to the nearest integer) For this data set that would exclude the 6 sites with α95 > 5.5°. This criterion is considerably stricter than earlier constraints on α95, when 20° was often used as a cut-off [McElhinny and McFadden, 1997]. Another criterion usually applied is to omit sites with low latitude poles, the often-called “transitional” directions. A common latitude cut-off angle is 45°, although some authors have used smaller values [Johnson and Constable, 1996], larger values [Constable and Johnson, 1999] or argued for a variable cut-off size depending on the angular dispersion [Vandamme, 1994]. For this latitude the Vandamme  method would require a cut-off angle of around 35° [McElhinny and McFadden, 1997], noting that in this case the angle is referring to the co-latitude and not the actual pole latitude.
 When the two criteria are applied to our data set from Meseta del Lago Buenos Aires 10 sites are removed from the population, the six sites with α95 above 5.5°, which all happen to also have VGP latitudes <55°, and four additional sites, all with small α95 values but pole latitudes <55°. Details of these shallow latitude sites, collectively called transitional, are given in Table 3. A total of 26 sites fulfill the criteria to represent the time averaged field, with 14 sites having normal polarity and 12 sites having reversed polarity, and are referred to as Selection I. The directions of this subset are plotted in Figure 5b along with the mean directions for both normal and reversed polarity sites.
 The normal and reversed means, with the reversed direction inverted through the origin, overlap, as can be seen in Figure 6. The geocentric axial dipole (GAD) field for this latitude lies directly between the normal and inverted reversed directions. Also shown is the direction of the IGRF2000 field for this location, which is considerably shallower than the GAD.
 Recently Tauxe et al.  have suggested that the precision parameter, k, is a better determinant of data quality, and that at least 5 samples per site are needed to accurately determine k. If these criteria are applied to the Meseta del Lago Buenos paleomagnetic sites, six sites are eliminated (five with k values <100 and one with N < 5). This data set, referred to as Selection II, eliminates four of the same transitional flows as in Selection I and two additional flows not rejected previously.
 Virtual geomagnetic pole positions for all the sites from MLBA are displayed in the Southern Hemisphere (Figure 7). Statistics for the mean poles from the total population and various subsets are listed in Table 4.
7. The 40Ar/39Ar Results
 Twenty four of the 38 lavas in Table 2 yielded remarkably concordant age spectra with >95% of the gas released defining the plateau ages (Figure 8). In contrast, only seven of these lavas yielded age spectra where <75% of the gas defined the age plateau, yet even these mildly discordant experiments yielded plateau and isochron ages that met the criteria outlined above (Table 2). None of the 38 isochrons yielded a 40Ar/36Ari value significantly different than 295.5 indicating that excess argon is not present in detectable levels that might bias the results (Figure 8). Moreover the age spectra and isochrons do not suggest that inherited or xenocrystic argon is of concern. Accordingly, we interpret the isochron ages from the 38 samples as the best estimate of time elapsed since these lavas erupted, cooled through their Curie temperatures, and acquired their thermoremanent magnetization direction.
Table 2. Summary of 40Ar/39Ar Incremental Heating Experiments on Lavas From Meseta del Lago Buenos Airesa
 The basaltic lavas range in age from an 11.03 ±0.14 Ma colonnade at the base of the Meseta-capping flow sequence south of Estancia Telken (Figure 2) to the 66 ± 3 ka Rio Pinturas pahoehoe flow that erupted from vents defined by two small scoria cones near the head of the colorful Rio Pinturas, one km from older eroded lavas that define the eastern escarpment of MLBA (Figure 2). Statistical analysis of the distribution of 40Ar/39Ar ages, including three lavas previously dated by Ton-That et al.  (Figure 2), reveals that basaltic volcanism was episodic rather than continuous over the last 11 million years (Figure 9). The reconnaissance nature of this study prevents extrapolating erupted volumes solely from the 40Ar/39Ar measurements, however, based on our field observations and limited mapping, this approach does yield a first-order impression of how MLBA formed. The ages define at least ten periods of peak volcanism, including multiple flows at ca. 11–10 Ma, 7.3–7.8 Ma, 3.2–3.0 Ma, 2.4 Ma, 1.7 Ma, 1.35 Ma, 1.0 Ma, 750 ka, 430–330 ka, and 109–66 ka (Figure 9). In addition, a single >10 m thick basalt flow on the northwest corner of MLBA, previously K-Ar dated by Mercer and Sutter , was 40Ar/39Ar dated by Ton-That et al.  at 5.12 ± 0.008 Ma (Figure 9). The most prominent peaks occurred during the latter part of the Matuyama reversed chron 1.70, 1.35 and 1.0 Ma and in the Brunhes chron 750 ka, 400 ka, and <110 ka (Figure 9). Both the Late Miocene and Late Pliocene-Pleistocene volcanic episodes in this region are consistent with diachronous subduction of segments of the Chile Rise spreading center, opening of a slab window beneath the South American plate, and melting the underlying asthenospheric mantle [Gorring et al., 2003 and references therein] (Figure 1).
 Paleomagnetic directions were determined from 34 of the 37 dated lava flows; normal paleodirections were determined from 16 of the lavas, whereas 12 others gave reversed directions, and 7 yielded transitional directions (Table 1). Together with the 40Ar/39Ar ages, these data indicate that 11 of the normally magnetized lavas erupted during the Brunhes chron, 17 erupted during the Matuyama chron, and 6 erupted during the Gauss or earlier chrons. However, not all the lavas temporally confined to the Matuyama chron are reversely magnetized. PAT 5, PAT 6, and PAT 1 preserve normal or transitional directions and ages corresponding to the Jaramillo normal subchron 1.07 to 0.99 Ma [e.g., Singer et al., 1999; Singer and Brown, 2002]. Sites PAT 27 and PAT 57, that may correspond to eroded remnants of the single large Casa Piedra Flow (Figure 2), are ca. 1.25 to 1.29 Ma and transitionally magnetized, suggesting the possibility that this flow erupted during the onset of the Cobb Mountain event [e.g., Singer et al., 1999]. Moreover, sites PAT 36 and PAT 38 are transitionally magnetized, suggesting that their ages of ca. 1.72 Ma may correspond to the termination of the Olduvai normal subchron. Ten other lavas erupted during the Matuyama chron are reversely magnetized.
 The 36 paleomagnetic sites and the 38 new 40Ar/39Ar dates from Meseta del Lago Buenos Aires provide important information from this here-to-for poorly studied region. Although the presence of late Tertiary basalts in Patagonia has been known for some time [Mercer, 1969; Baker et al., 1981], this is the first detailed study of the geochronology and paleomagnetism of such rocks since the pioneering work of Fleck et al. . It adds to a growing body of petrologic and geochemical data that are used to refine models of slab window magmatism under southern South America [Gorring and Kay, 2001; Gorring et al., 1997, 2003]. The paleomagnetic data allows us to investigate the time averaged field deep within the Southern Hemisphere as well as look at well-dated transitional results.
8.1. Transitional Directions
 Among the 36 paleomagnetic sites from MLBA are ten sites that have low latitude poles. As listed in Table 3, these sites have pole latitudes that range from 7.2° to 54.6°, putting them in the classification of transitional directions. That over 20% of the total sites in this study yield transitional directions may at first seem excessive. Early work by Doell and Cox  looking at over 1000 lava flows from around the globe with ages less than 10 Ma, found only about 2% of directions fell into a transitional (low latitude pole) classification. Recent paleomagnetic and geochronologic results [e.g., Singer et al., 2002] indicate that transitional directions, especially those related to excursions, are perhaps more common in the magnetic record than previously thought. Constable and Johnson  looking at statistical models for paleosecular variation compared to lava data over the past 5 Ma find that many observables are anisotropic. This includes the percentage of VGP latitudes <45°, which for the Southern Hemisphere at the latitude of MLBA is 5% or 10% depending on the model used. Thus it seems possible that periods of transitional field behavior could be larger in the Southern Hemisphere, but still not as large as seen in this study. The question remains as to whether these sites are far from the expected directions due to other circumstances- such as lightning strikes, undetected local tectonics (block rotations, small-scale tilting, non-horizontal emplacement), or subsequent alteration of magnetic mineralogy, or whether they represent true field positions.
Table 3. Paleomagnetic Directions and Ages of Sites with Transitional Directionsa
Age, isochron age, in Ma, with ±2σ; DIRECTION, inclination and declination in degrees; POLE, latitude and longitude of virtual geomagnetic pole, in degrees. Asterisks indicates lava directly overlies PAT 27.
0.341 ± 0.033
0.984 ± 0.035
End of Jaramillo
1.016 ± 0.010
1.250 ± 0.031
1.290 ± 0.030
1.315 ± 0.029
1.720 ± 0.020
End of Olduvai? or
1.730 ± 0.020
7.860 ± 0.160
Onset of C4n.2n
 Geochronologic data helps to investigate the validity of these transitional directions in that age determinations from all the sites are closely related to known chron or subchron boundaries. As indicated in Table 3, the ten transitional directions fall into several groups. Two flows, PAT 6 at 0.984 ± 0.035 Ma and PAT 1, with an age of 1.016 ± 0.010 Ma, correspond to the end and the upper part of the Jaramillo subchron, currently defined as existing from 1.069 ± 0.006 Ma to 1.000 ± 0.005 Ma (Singer et al. ; age of chron boundaries here and below were adjusted to reflect intercalibrated values of neutron fluence standards in Renne et al. , that have been adopted in this paper). There are four sites (PAT 27, 28, 48 and 57) with ages between 1.25 and 1.32 Ma that have poles with low southern latitudes and longitudes near 300° (Table 3; Figure 7). These ages are slightly older than those currently associated with the Cobb Mountain Subchron of 1.194 ± 0.006 Ma [Turrin et al., 1994; Renne et al., 1998; Singer et al., 2002], but may represent a short pre-Cobb Mountain feature observed in recent ODP cores from the Bermuda Rise [Yang et al., 2001]. Two flows on the Meseta, PAT 36 and 38, have ages of 1.72 and 1.73 Ma respectively, as well as near-equatorial poles. These ages are slightly younger than the accepted age of the Olduvai Subchron (1.95 to 1.77 Ma; adjusted from Cande and Kent ), but given the 2σ uncertainties, they are only 20 kyr older than the weighted mean of two K-Ar ages of 1.62 ± 0.06 Ma from basalt flows thought to record the elusive Gilsa event [Udagawa et al., 1999]. Finally, the oldest transitional direction, PAT 41, dated at 7.860 ± 0.16 Ma, is coincident with that of the termination of a normal chron, C4n.2n, at 7.76 Ma (adjusted from Cande and Kent ). Only the youngest flow, PAT 37 at 0.341 ± 0.033 Ma is not readily correlated with a known polarity event or excursion.
 In addition to the consistent agreement of the geochronology with known periods of field transition, the magnetic data supports the contention that true field behavior is being measured. Although none of these sites exhibit low NRM intensity one might associate with transitional field positions, they also do not have high intensities that might be associated with lightning strikes (Figure 4b). Average NRM for these 10 flows is 4.68 A/m, only slightly greater than the average for the entire set of 4.03 A/m. These sites do have a larger average α95 than the total group (6.3°Compared to 4.9°), but with 8 of the 10 flows having α95 ≤ 6.6° (Figure 4c). The data from these flows hardly classify as “scattered” but rather are only slightly more dispersed than the acceptable flows. The fact that there are several flows with similar ages and similar directions (Table 3), but not co-located, rules out local tilting or small block rotations as a cause of aberrant directions. The transitional site with the greatest scatter (PAT 38 with α95 = 13.9°) has an identical age to the transitional flow with the least scatter (PAT 36, with α95 = 2.6°). It appears that these sites, with some caution, can be considered to represent field behavior at the time of their emplacement.
 One of several ways to view paleosecular variation at any one location is to investigate the dispersion of directions or poles. It has long been common practice to calculate the dispersion of the VGPs about the rotation axis (Sp) and then compare this value to predicted models [e.g., Doell and Cox, 1972]. Table 4 reports the angular dispersion of the VGPs from Meseta del Lago Buenos Aires for the entire collection as well as the two subsets discussed above, after correcting for within-site dispersion using the method of McElhinny and McFadden . Both the total population and Selection II yield large values for dispersion of the field (Sf), while the subset of flows with pole latitudes > 55° gives a dispersion of 20.0° with 95% error limits of 16.8° and 24.8°. This value is plotted on Figure 10 for comparison to the fit of paleosecular variation data from lava flows for the last 5 million years to Model G of McFadden et al. . The data from MLBA Selection I lies appreciably above Model G, even with 95% error limits, indicating this data is more dispersed than predicted from Model G.
Table 4. Paleosecular Variation Statistics for Meseta del Lago Buenos Aires Lava Flowsa
Number, number of sites in group; Direction, mean inclination and declination of group; K, precision parameter; α95 and A95, 95% confidence circle about mean direction and mean pole; Pole, mean latitude and longitude of group pole; SF, dispersion of poles with respect to geographic pole corrected for within-site dispersion; Su and Sl, upper and lower 95% limits on SF from Cox . Selection I and II described in text.
 To consider whether the dispersion viewed here is a valid measure of the TAF in South America requires other data sets use for comparison. Current work on lavas from the Pali Aike field in southern Patagonia (52°) yields a dispersion value closer to Model G [Mejia et al., 2001]. Work on lavas from Tatara-San Pedro volcanic complex in central Chile (36°S) has produced a large number of measurements on lava flows younger than the Brunhes/Matuyama reversal (Brown, unpublished data, 2002). Initial results indicate rather large dispersion (18.2°) but additional flow statistics must be incorporated to obtain final results. Other data from the Southern Hemisphere includes results from the Crozet Archipelago [Camps et al., 2001], also at 47°S. Results of 45 lava flows, also using a 45° latitude cut-off, give angular dispersion of 18.6° with errors of 14.6° and 22.4°, showing good overlap with the MLBA results. Results from Easter Island [Brown, 2002] at 27°S lie right on Model G, with a dispersion of 14.1° and error limits of 12.4° and 16.2°.
 At this point it is difficult to know if the higher than expected dispersion in southern South America is due to actual field behavior in this part of the world, or whether it is an artifact of inadequate or biased sampling. Tauxe et al.  make a case for the TAF only being accurately determined from data sets of several hundred flows. Working in the American southwest, they combine studies over an area of 5° latitude and nearly 20° longitude. If this is the case, areas such as southern South America will need many more studies to provide the necessary data to define the TAF.
 To augment and improve observations relevant to understanding the time averaged character of the Earth's magnetic field, we have measured paleomagnetic directions and determined 40Ar/39Ar ages from three dozen lava flows or shallow intrusions at Meseta del Lago Buenos Aires, 47°S in Argentine Patagonia. Our results lead to the following conclusions:
 1. Thirty-six lava flows give stable paleomagnetic directions. After omitting sites with low latitude poles (<55°), the remaining 26 flows have a mean direction of inclination = −63.0°, declination = 3.4°, α95 = 5.4°.
 2. Thirty-eight new 40Ar/39Ar incremental heating experiments on all but two of the paleomagnetic sites gave ages that range from 11.030 ± 0.140 Ma to 0.067 ± 0.004 Ma, with the bulk of the ages being younger than 3.3 Ma.
 3. Geochronology indicates that the volcanism was episodic with peak periods of eruptive activity at 3.20–3.00 Ma, 2.40 Ma, 1.70 Ma. 1.35 Ma, 1.00 Ma, 750 ka, 430–330 ka and <110 ka.
 4. Transitional flows, as determined by shallow VGP latitudes, make up over 20% of the total population. The ages determined from the transitional flows indicate they were emplaced within or near short subchrons or at reversal boundaries.
 5. Secular variation of Selection I lava flows gives an angular dispersion of VGPs of 20.0°, larger than dispersion predicted by Model G. It is not known if this larger dispersion is in response to regional variations in the magnetic field or is an artifact of inadequate sampling.
 Special thanks to Coco and Petty Nauta at Estancia Telken for providing us with a home away from home and invaluable assistance with logistics and sampling. Field assistance from Lyn Gualtieri, Jason Gowers, Walter Jaslaneck, Fidel Costa, and Robert Ackert is gratefully acknowledged. Reviews by Catherine Constable, an anonymous reviewer and the associate editor were most helpful. Supported by NSF grant EAR 9805424 to LB and NSF grant EAR 9909309 to BS.