LETTERSPUBLISHED ONLINE: 13 SEPTEMBER 2009 | DOI: 10.1038/NGEO622

No asymmetry in geomagnetic reversals recordedby 1.1-billion-year-old Keweenawan basaltsNicholas L. Swanson-Hysell1*, Adam C. Maloof1, Benjamin P. Weiss2 and David A. D. Evans3

Interpreting the past latitude and geography of the continentsfrom palaeomagnetic data relies on the key assumption thatEarth’s geomagnetic field behaves as a geocentric axial dipole.The axial dipolar field model implies that all geomagneticreversals should be symmetric. However, palaeomagneticdata from volcanic rocks produced by the 1.1-billion-year-old Keweenawan Rift system in North America have beeninterpreted to show asymmetric reversals, which had led tothe suggestion that there was a significant non-axial dipolecontribution to the magnetic field during this time1,2. Here wepresent high-resolution palaeomagnetic data that span threegeomagnetic field reversals from a well-described series ofbasalt flows at Mamainse Point, Ontario, in the KeweenawanRift. Our data show that each reversal is symmetric. We thusconclude that the previously documented reversal asymmetryis an artefact of the rapid motion of North America duringthis time. Comparisons of reversed and normal populationsthat were time-averaged over entire polarity intervals, or fromsites not directly on either side of a geomagnetic reversal, havepreviously led to the appearance of reversal asymmetry.

Extrusive and intrusive igneous rocks of the Keweenawanprovince form part of a failed mid-continent rift (MCR) systemthat was active from 1.11 to 1.09Gyr ago. The largest arealexposure of these rocks is in the Lake Superior region, butaeromagnetic and gravity surveys show that the relatively denseand magnetic volcanic rocks of the rift span more than 3,000 km,largely beneath sedimentary cover (Fig. 1)3. Palaeomagnetic datafrom Keweenawan rocks have been compiled to generate anapparent polar wander (APW) path for North America, known asthe Logan Loop4–6. Comparisons of the Logan Loop with APWpaths from other continents have been used to reconstruct thesupercontinent Rodinia7,8.

Palaeomagnetic directions from MCR rocks consistently revealnormal and reversed directions that are not antiparallel (withinclination differences of 20◦–30◦; Fig. 1)4. This pattern hasbeen interpreted as reversal asymmetry, leading to speculationthat significant non-dipole components may have contributedto the surface geomagnetic field1,2. The well-known problem ofKeweenawan reversal asymmetry is often cited as an uncertaintyin palaeogeographic reconstructions7–9, as the introduction ofquadrupole and octupole components to the geomagnetic fieldcan lead to significant discrepancies between true palaeolatitudeand palaeomagnetically derived palaeolatitude10. Much of themovement of the late Mesoproterozoic APW path for Laurentiaoccurs across what has been interpreted as an ‘asymmetric’ reversal.The large magnitude of this apparent motion makes it important tounderstand whether the difference in inclination between normaland reversed directions is an artefact of a significant non-dipole

1Department of Geosciences, Princeton University, Guyot Hall, Washington Road, Princeton, New Jersey 08544, USA, 2Department of Earth, Atmosphericand Planetary Sciences, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA, 3Department ofGeology and Geophysics, Yale University, 210 Whitney Avenue, New Haven, Connecticut 06520, USA. *e-mail: nswanson@princeton.edu.

contribution to the field, or whether it is due to motion ofthe North American continent. This information is critical fordeciphering North American plate motions during the Himalayan-scale Grenville Orogeny and the assembly of cratons to formthe supercontinent Rodinia. A better understanding of the lateMesoproterozoic geomagnetic field is also necessary to evaluateclaims of non-uniformitarian processes in the Proterozoic era, suchas true polar wander and low-latitude glaciation, which rely on theuniformitarian geocentric axial dipole (GAD) hypothesis11,12.

Previous studies have compared the mean of all Keweenawan-aged reversed directions with all normal directions2,4 and/or havegrouped all reversed and normal directions at a single studylocation13,14. To fully evaluate whether the actual reversals areasymmetric, it is essential to obtain high-resolution palaeomagneticdata through stratigraphic sampling that spans reversals in thecontext of detailed volcanostratigraphy.

The basalt flows at Mamainse Point unconformably overliethe Archaean Superior Province on the northeastern shore ofLake Superior (Fig. 1) and represent the most complete recordof extrusive Keweenawan volcanism. Mamainse Point is theonly known locality in the MCR where multiple reversals arerecorded in a succession of extrusive lava flows, as there are threereversals (reversed→ normal→ reversed→ normal) throughoutthe∼4,500m of basalt flows and inter-rift sediments.

Here we report palaeomagnetic data from 73 basalt flows atMamainse Point that span the three reversals at high resolution andshow each reversal to be symmetric (Fig. 2). There is a progressivedecrease in the inclination of the palaeomagnetic data up-stratigraphy from∼70◦ to∼30◦ that we interpret as resulting fromthe equatorwardmovement of Laurentia. As a result of this progres-sive change in palaeolatitude through time, a reversal test that usesmeans of entire polarity zonesmay result in a false negative.

Palaeomagnetic studies at other localities in the MCR haverevealed a general pattern that older rocks carry a reversed magne-tization and younger rocks record a normal magnetization (Fig. 1c;Supplementary Table S3). This trend led early workers to postulatethat there was only one geomagnetic reversal during the 11Myrof Keweenawan volcanism and that this reversal could be usedas a stratigraphic datum for all Keweenawan igneous units5,13–15.When Palmer13 discovered multiple reversals in the sequence ofbasalts at Mamainse Point, this single-reversal paradigm led himto propose the presence of a sequence-repeating fault, despite thefact that, in his words, there is ‘no geological evidence to supportthe fault hypothesis’. In addition to the lack of geological evidence,the fault repetition hypothesis has been discounted by geochemicaldata (Fig. 2b; see Supplementary Information for a more completedescription of previous palaeomagnetic work at Mamainse Pointand other dual-polarity localities in theMCR)16.

NATURE GEOSCIENCE | VOL 2 | OCTOBER 2009 | www.nature.com/naturegeoscience 713© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO622

90°

180°

n Dec. Inc. Reversal test

Negative

Positive

Lower hemisphereUpper hemispherePresent local field

Age

(M

yr)

1085

1090

1095

1100

1105

1110

1115

N

R DCr

OVrNSr

PMrNCr

NSn

PLL

LSTPPB

GCD MCR

KSZ

Grenvil

le

91°

49°

47°

Bayfield-Jacobsville qtz sandstones

Feldspathic sandstone and shale

Conglomerate

Volcanics (dominantly R/N)

Marls, sandstone

Gabbroic intrusions

Thunder Bay

Sault Ste. Marie

Alkaline intrusions

Diabase sills

Diabase dike swarms

Km

MamainsePoint

89° 87° 85°

Duluth

Lake Superior

Fig 2a

0 50

MCRN: 8 290.2 40.8 6.7

MCRR: 8 111.2 –66.7 6.3

KSZN: 3 306.1 55. 1 8.5

KSZR: 5 117.8 –59.8 10.6

95αa b c

Figure 1 | Mid-continent rift geology and palaeomagnetic data. a, Generalized geologic map of the Lake Superior region after Halls and Pesonen4. Theinset map shows the location of the mid-continent rift (MCR) gravity anomaly, the Kapuskasing structural zone (KSZ), the Pikes Peak batholith (PPB) andthe Gila County diabases (GCD). b, Equal-area plot of tilt-corrected polarity zone means and α95 error ellipses from all MCR dual-polarity locales. MCRdata (red and dark blue) show 20◦–40◦ of inclination asymmetry and fail a reversal test, whereas KSZ directions (orange and light blue) pass a reversaltest. c, U/Pb ages with 2σ error bars from locations with both palaeomagnetic and geochronological data (see Supplementary Table S3 and ref. 15). DCr:Duluth Complex reversed, LST: Lake Shore Traps, NCr: Nemegosenda Carbonatite normal, NSn: North Shore Volcanics normal, NSr: North Shore reversed,OVr: Osler Volcanics reversed, PLL: Portage Lake Lavas, PMr: Powder Mill Volcanics.

MP50

MP49MP48

MP121

MP46MP214

MP43

MP107

MP105

MP18

MP101

MP54MP111

MP207MP209

MP211

MP212

Flow boundaryHighly vesiculated,v.fine-grained

Vesiculated,fine-grained

Massive

Reve

rsed

Reve

rsed

Nor

mal

Nor

mal

Lower reversed 1

Lower reversed 2

Lower norm

alU

pper reversedU

pper normal

ReversedNormal

1 (m)

668000

5219000 4,000 ¬10 ¬5 0

20

10

0

5

3,500

3,000

2,500

Greatconglomerate

Basalt clastconglomerate

2,000

1,500

1,000

500

0(m)

Palaeolatitude (°)

90

90°

180°

180°

5218000

5217000

5216000

5215000

5214000

5213000 MamainsePoint

5212000

669000 670000 671000

Lake Superior

MP201MP45

672000 673000

a cb

e

d

0 20 40 60 80

Nd

Upper normal

Upper reversed

Lower normal

Lower reversed 2

Lower reversed 1

1 m

°

35

3331

34 35

35

42

43

49

45/

/

/

/

/

/

85°–70°

70°–60°

60°–50°

50°–40°

40°–30°

30°–20°

Bedding (strike/dip)Roads

Faults

Water

Felsic intrusion

Conglomerate

Conglomerate testAxis labels in meters for UTMzone 16 N (WGS84 datum)

Basement

Basalt

Palaeomagnetic data(absolute inclination by flow)Reversed/normal

45

Figure 2 | Summary of palaeomagnetic data from Mamainse Point. a, Geological map of Mamainse Point30 with a symbol at each flow withpalaeomagnetic data (colour-coded by magnetic inclination with squares for reversed and circles for normal). b, Simplified stratigraphy with εNd data16 andabsolute palaeolatitude for the virtual geomagnetic pole (VGP) of each flow. c, Stratigraphy for basalt flows from the MP107 section with a photo of apahoehoe flow top. d, Equal-area plot of the tilt-corrected Fisher mean and associated α95 ellipse for each flow sampled—colours correspond toindividual sections shown in stratigraphic position in b. e, Equal-area plot of tilt-corrected Fisher means of flow VGPs grouped as indicated by theright-most brackets in b.

714 NATURE GEOSCIENCE | VOL 2 | OCTOBER 2009 | www.nature.com/naturegeoscience

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE GEOSCIENCE DOI: 10.1038/NGEO622 LETTERS

Lower reversed 2

Lower reversed 1

Lower normal

Upper reversed

Upper normal

Mamainse Point palaeopoles

C

J

N

F

NSr 1108+–1

NSn 1097+–2

LST 1087+–2

OVr 1105+–2

Keweenawan reversed

Pre-Keweenawan reversed

Keweenawan normal

Post-rift sediments

Mid-continent rift palaeopoles

a b

PLL 1095+–3

CCr 1108+–1

ABD 1141+–1

PMr 1107+–2

DCr 1107+–1

Figure 3 | The Logan Loop and Mamainse Point poles. a, Equal-area projection of poles for all MCR volcanics for which there are both palaeomagnetic andgeochronological data (see Supplementary Table S3). ABD: Abitibi Dykes, CCr: Coldwell Complex reversed. The yellow palaeopoles are from rift sedimentsdeposited after the cessation of volcanism but do not have rigorous geochronological control. Their relative deposition is such that from oldest to youngest:N→ F→ J→C. N: Nonesuch Shale, F: Freda Sandstone, J: Jacobsville Sandstone, C: Chequamegon Sandstone. b, Palaeopoles from the Mamainse Pointsuccession calculated using the mean of flow VGPs as grouped in Fig. 2 (see Supplementary Table S2 for palaeopole parameters).

Further evidence of multiple reversals during MCR volcanismcomes from coeval alkaline complexes exposed along theKapuskasing structural zone (Fig. 1). At least three of these alkalinecomplexes contain nearly symmetric polarity reversals. Whereassome record a reversed to normal polarity change, others record achange from normal to reversed17. One of these units, the ColdwellComplex, represents at least three intrusive episodes. Althoughthe determination of cross-cutting relationships can be difficultowing to complex cooling histories, palaeomagnetic data from theColdwell Complex are interpreted to reflect a R1→N→R2polarityhistorywith only 5◦–10◦ of inclination asymmetry between reversals(Fig. 1b; Supplementary Table S3)18.

Owing to the progressive decrease in the inclination of themagnetization upwards through the stratigraphy at MamainsePoint (Fig. 2), a reversal test19 on the new data set that comparesall reversed directions with all normal ones fails, consistent with theprevious literature. However, if the three reversals are consideredindividually, each one passes a reversal test. The first reversal occursduring a hiatus in volcanism and the deposition of the basalt clastconglomerate. A reversal test comparing the three reversed flowsbelow the conglomerate in sectionMP214 to the three normal flowsof MP214 above is positive with a ‘B’ classification (γc= 10.0). Thistest shows that the angle between reversed and normal populations(γo) is less than the angle at which the null hypothesis of a commonmean direction would be rejected with 95% confidence (γc) andbecause γc is between 5◦ and 10◦ it is given a ‘B’ classification. Theflows above the basalt clast conglomerate are normally magnetizedand lead up to flows of reversed magnetization without separationby a conglomerate. Three of the flows have anomalous declinationsand could possibly be transitional. A reversal test conductedbetween all of the flows in the lower normal polarity zone (11 flows)and all of the flows in the upper reversed polarity zone (10 flows)passes with a ‘B’ classification (γc = 9.8). The uppermost reversaloccurs within the ‘great conglomerate’. A reversal test between allof the flows of the upper reversed zone and the 13 normal flows ofsectionsMP207 andMP209 above the conglomerate is positive witha ‘B’ classification (γc= 6.1).

A fundamental assumption of many palaeomagnetic studies isthat the surface expression of the geomagnetic field behaves as atime-averaged GAD, where mean field inclination (I ) is a simple

function of latitude (tan(I )=2tan(λ)). Under the GAD hypothesis,palaeomagnetic data that span enough time to average out secularvariation should produce a palaeomagnetic pole that correspondsto Earth’s spin axis. Archaeomagnetic studies show that, over thepast 10,000 years, Earth’s magnetic field is best described by aGAD (ref. 20). Palaeomagnetic data are broadly consistent with aGAD for the past 100–150Myr (the only period for which reliableplate reconstructions can be made independent of palaeomagneticdata) with only a small (3–5%) axial quadrupole component21.Palaeolatitude calculations made assuming a GAD geomagneticfield correspond well with climatic indicators of latitude, such asevaporite basins, for the past 2Gyr (ref. 22).

The interpreted reversal asymmetry in Keweenawan data setsled Pesonen and Nevanlinna1,2 to propose that the ∼1.1Gyrgeomagnetic field contained significant non-dipole components.Specifically, Pesonen and Nevanlinna proposed that in addition tothe main dipole there was a persistent offset dipole. They arguedthat ‘reversal asymmetry’ resulted from the main dipole reversingwhile the offset dipole retained constant polarity. This contributionto the geomagnetic field can alternatively be expressed in terms ofzonal harmonics as a geocentric dipole (g 0

1 ) and an axial quadrupole(g 0

2 ) where the reversal asymmetry results from a reversing dipoleand a non-reversing quadrupole. Reversal asymmetry would alsoarise if there was a significant axial octupole (g 0

3 ), or any otherhigher-order component, that maintained constant polarity duringthe reversal of the dipole.

Another approach for evaluating the GAD hypothesis has beento calculate the elongation parameter for palaeomagnetic datasets from periods of rapid volcanism when the host continentcan be assumed to be effectively stationary23. The resultingelongation/inclination data pairs can then be compared withstatistical models of palaeosecular variation that predict thedistribution of geomagnetic field vectors for aGADand thus a singleelongation value for each inclination. For the Mamainse Point datapresented herein, the applicability of this method is limited becausethe necessary condition of a stationary continent for an intervalin which 100 flows can be sampled cannot be met owing to thesignificant change in palaeolatitude through the section. Althoughit is possible that there is a locality within the rift where high ratesof volcanism could yield flows that effectively meet the stationary

NATURE GEOSCIENCE | VOL 2 | OCTOBER 2009 | www.nature.com/naturegeoscience 715© 2010 Macmillan Publishers Limited. All rights reserved.

LETTERS NATURE GEOSCIENCE DOI: 10.1038/NGEO622

J (10

-4 A

m2 )

InclinationDeclination

Tilt-corrected

Z, N

W

MP53-2.7

c

f

g

h

e

MP212-18.5

L

J (1

0-3

A m

2 )InclinationDeclination

Tilt-corrected

Z, N

W

580

510

510

580

670

8.0

AF (mT) Thermal demag. (°C)

1.00

0.75

0.50

0.25

J/J 0

2.0

140

70 200

300

380

J0 = 1.31e-2 A m2

48

0

580

670

NRM (J0)

AF (0¬10 mT)N2 (77 K)

Thermal (0–650°C)

NRM

AF AF

8.0

AF (mT) Thermal demag. (°C)

1.00

0.75

0.50

0.25

J/J 0

2.0

100

200

300

40

0

J0=2.41e-3 A m250

0

580

670

NRM (J0)AF (0¬10 mT)Thermal (0¬650 °C)

NRM

Haematite

Magnetite

Haematite

Magnetite

580

670

670

500

500580

a

d

b

Figure 4 | Example palaeomagnetic data. a–h, Representative vector-component (Zijderveld) diagrams (a and e), equal-area projections (b and f),least-squares fits (c and g) and magnetic intensity (J/J0) plots (d and h) for Mamainse Point basalts. In the vector-component plots, the haematite fit istraced with a red arrow and the magnetite fit is traced with a blue arrow. These components are summarized in equal-area plots (c and g). In all plots,green-filled squares are natural remanent magnetization (NRM), brown-filled squares are alternating field (AF) demagnetization steps and red-filledsquares reflect thermal demagnetization. In the equal-area plots (b and f), colour-filled/colour-rimmed circles represent lower/upper-hemispheredirections. All data are in tilt-corrected coordinates.

criteria, the fast rates of APW through the interval (Fig. 3) willprobably complicate this analysis inmost localities.

As the three reversals in the succession at Mamainse Point aresymmetric, there is no need to invoke substantial deviations froma GAD such as a non-reversing quadrupole or octupole. Althoughthis result does not preclude non-dipole contributions to thegeomagnetic field during theMesoproterozoic era, it does eliminatethe main rationale that has been used to question the validityof the GAD hypothesis during the time period. This rationale isfurther weakened by the presence of symmetric reversals in 1.4Gyrsedimentary red beds24 and in carbonates of Mesoproterozoicage25. Instead of resulting from reversal asymmetry, the change ininclination observed through the succession is readily explained asa consequence of continental motion.

Palaeomagnetic poles that are paired with U–Pb geochrono-logical data constrain the rate of continental motion dur-ing Keweenawan volcanism (Fig. 3). A comparative analysis ofpalaeopoles suggests that the volcanism at Mamainse Point spansmuch of the period of Keweenawan volcanism given the similarityof the ‘lower reversed 1’ pole to the Coldwell Complex reversed1 pole, which has an age of 1,108± 1Myr and the similarity ofthe ‘upper normal’ pole to the North Shore Volcanics normalpole, which has an age of 1,097± 2Myr. Using the conclusionthat the field was probably a GAD, we can calculate the minimumvelocity that is implied by the motion from the Coldwell Complexreversed 1 pole to the North Shore Volcanics normal pole to be33.6± 3.5 cm yr−1 (see Supplementary Information for error cal-culation methods and Supplementary Table S4 for rates between allKeweenawan palaeomagnetic poles with U/Pb ages). The estimatesfor the rate of motion from 1,108Myr to 1,097Myr range between21.5±7.1 cm yr−1 for Osler Volcanics reversed→North Shore nor-mal, to 33.6±3.5 cm yr−1 for Coldwell Complex reversed→NorthShore normal. The compilation of polar wander rates shows thatduring later Keweenawan volcanism movement slowed (as has

previously been noted by Davis and Green26). The rate from the1,097Myr North Shore normal pole to the 1,087Myr Lake ShoreTrap pole is 12.2±6.4 cm yr−1. These rates for the motion of NorthAmerica during the first 10million years of Keweenawan volcanismare faster than the modelled speed limit of 20 cm yr−1 for platetectonic motions imposed by bending stresses associated with thesubduction of oceanic lithosphere27. The only plate to reach thistheoretical limit during the past 150Myr was India as it travellednorth following the break-up of Gondwanaland. Either 20 cm yr−1is not a firm speed limit for plate motion at 1Gyr, or some of themotion observed could be a result of true polar wander that drovethe North American continent and the mid-continent rift, towardsthe equator at rates slightly faster than those of plate tectonics28—apossibility that could be testedwith data fromother continents.

The period of Keweenawan volcanism was a time of significantequatorward motion of the North American continent. As a resultof this progressive change in palaeolatitude, previous reversaltests that used means from combined results in each polarityproduced false negatives. Detailed stratigraphic context combinedwith high-resolution data are essential for understanding both thebehaviour of the geomagnetic field and APW paths in the lateMesoproterozoic era.

MethodsThe extrusive basalt flows at Mamainse Point are remarkably preserved and rangefrom less than 1m to more than 20m thick. Individual flows usually follow thisgeneral stratigraphy from base to top: (1) massive (relatively coarse-grained) withscattered elongate-upwards pipe vesicles (see Supplementary Fig. S12), (2) massivewith occasionally apparent flow banding, (3) massive with some vesicles,(4) highly vesiculated and fine-grained and (5) flow tops often with pahoehoetextures and slight ferruginization (Fig. 2c; see Supplementary Figs S9–S11 forfurther photos of the basalts). Individual flows and sedimentary hiatuses canbe delineated, allowing us to place our palaeomagnetic sampling in a robuststratigraphic context. Although there are numerous conglomerates interbeddedwith the flows particularly in the upper portion of the stratigraphy (Fig. 2a),there are two especially thick conglomerates. The lower of these is the ‘basalt

716 NATURE GEOSCIENCE | VOL 2 | OCTOBER 2009 | www.nature.com/naturegeoscience

© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE GEOSCIENCE DOI: 10.1038/NGEO622 LETTERSclast conglomerate’, which is contained within our section MP214 where theconglomerate is 39m thick. The upper conglomerate, or ‘great conglomerate’, is∼310m thick and shows a progressive increase in the percentage of clasts derivedfrom Superior Province basement relative to basalt clasts (see clast counts inSupplementary Fig. S8).

Multiple samples (usually 5–6) were collected and subsequently analysedfrom 73 flows spanning the stratigraphy from the lower contact with the SuperiorProvince metamorphic rocks to above the third reversal in the succession (seeSupplementary Information for a description of the detailed demagnetizationroutines used). The thermal demagnetization behaviour and hysteresis loopsof the basalt samples (Fig. 4 and Supplementary Fig. S1,S2 and S4–S7) showthat the magnetization is carried by both haematite and magnetite. Whereas themagnetite exists as a primary igneous mineral in the basalt, the haematite probablyformed as a result of deuteritic oxidation shortly after the eruption of each flow.Our detailed thermal demagnetization protocol allows for comparison of thehaematite and magnetite components. Our data demonstrate that the characteristicmagnetization components carried by the two mineralogies (unblocked between500 and 575 ◦C for magnetite and 650–675 ◦C for haematite) are consistently ofa very similar direction (Fig. 4). This result suggests that the haematite formedduring initial cooling of the basalts or during pedogenesis that occurred soon afterflow emplacement and recorded the same snapshot of the geomagnetic field asthe magnetite component. In some instances, the high intensity of the haematitecomponent prevented isolation of the magnetite component. For 16 flows, it wastherefore necessary to use the haematite component as the characteristic remanentmagnetization (ChRM) and for an extra 12 flows magnetite fits were used for somesamples and haematite for others, as indicated in Supplementary Table S1. When amagnetite component was confidently isolated, as indicated by a ‘shoulder’ in plotsof J/J0, a least-squares fit through the demagnetization steps in its unblocking rangewas used as the ChRM (see Supplementary Table S1 for the mean direction of eachflow). A number of flows have an extra magnetization component that unblocksbetween 590 and 650 ◦C that is of opposite polarity to the components thatunblock between 500 and 575 ◦C and between 650 and 675 ◦C (see SupplementaryFig. S2) and is further discussed in Supplementary Information. Fisher meanswere calculated for each flow from the ChRM directions isolated through principalcomponent analysis29 of samples within that flow (Supplementary Table S1).

Clasts were sampled and analysed from the basalt clast conglomerate(site MP45; Supplementary Fig. S11) and from basalt clasts within the greatconglomerate (site MP201; Supplementary Fig. S11) allowing for two conglomeratetests (see Supplementary Fig. S3). Least-squares fits to both the magnetite andhaematite components for each test are consistent with selection from a randompopulation (see Supplementary Information for details of the test results).These conglomerate test results further confirm that both the magnetite andhaematite components observed in the flows were acquired before the erosion ofthe flows and the deposition of these conglomerates that were deposited withinthe larger stack of Keweenawan basalts. Therefore, we can be confident in theuse of these directions as a record of the local magnetic field at the time theflows erupted and cooled.

The terms ‘normal’ and ‘reversed’ when applied to palaeomagnetic directionsfrom the MCR are not intended to specify the absolute orientation of thegeomagnetic field at the time of volcanism, as there are unresolved parts of theAPW path for Laurentia since the Mesoproterozoic era. Rather, the designationdistinguishes between directions with negative inclinations (reversed) anddirections with positive inclinations (normal), a convention used in all previouspalaeomagnetic studies of Keweenawan volcanics.

Received 22 April 2009; accepted 5 August 2009;published online 13 September 2009; corrected online13 April 2010

References1. Pesonen, L. & Nevanlinna, H. Late Precambrian Keweenawan asymmetric

reversals. Nature 294, 436–439 (1981).2. Nevanlinna, H. & Pesonen, L. Late Precambrian Keweenawan asymmetric

polarities as analyzed by axial offset dipole geomagnetic models. J. Geophys. Res.88, 645–658 (1983).

3. Van Schmus, W. R. & Hinze, W. J. The midcontinent rift system. Annu. Rev.Earth Planet. Sci. 13, 345 (1985).

4. Halls, H. & Pesonen, L. Paleomagnetism of Keweenawan rocks. Geol. Soc.Am. Mem. 156, 173–201 (1982).

5. Dubois, P. Paleomagnetism and correlation of Keweenawan rocks. Bull. Geol.Surv. Can. 71, 1–75 (1962).

6. Robertson, W. & Fahrig, W. The great Logan Loop—the polar wandering pathfrom Canadian shield rocks during the Neohelikian era. Can. J. Earth Sci. 8,1355–1372 (1971).

7. Weil, A., Van der Voo, R., Mac Niocaill, C. & Meert, J. The Proterozoicsupercontinent Rodinia: Paleomagnetically derived reconstructions for 1100 to800Ma. Earth Planet. Sci. Lett. 154, 13–24 (1998).

8. Buchan, K. et al. Rodinia: The evidence from integrated palaeomagnetism andU–Pb geochronology. Precambr. Res. 110, 9–32 (2001).

9. Piper, J. The Neoproterozoic supercontinent Palaeopangea. Gondwana Res. 12,202–227 (2007).

10. Kent, D. & Smethurst, M. Shallow bias of paleomagnetic inclinations in thePaleozoic and Precambrian. Earth Planet. Sci. Lett. 160, 391–402 (1998).

11. Evans, D. Stratigraphic, geochronological, and paleomagnetic constraints uponthe Neoproterozoic climatic paradox. Am. J. Sci. 300, 347–433 (2000).

12. Maloof, A. et al. Combined paleomagnetic, isotopic and stratigraphic evidencefor true polar wander from the Neoproterozoic Akademikerbreen Group,Svalbard, Norway. GSA Bull. 118, 1099–1124 (2006).

13. Palmer, H. Paleomagnetism and correlation of some Middle Keweenawanrocks, Lake Superior. Can. J. Earth Sci. 7, 1410–1436 (1970).

14. Robertson, W. Pole positions from theMamainse Point Lavas and their bearingon a Keweenawan pole path. Can. J. Earth Sci. 10, 1541–1555 (1973).

15. Hanson, R. et al. Coeval large-scale magmatism in the Kalahari and LaurentianCratons during Rodinia assembly. Science 304, 1126–1129 (2004).

16. Shirey, S., Klewin, K., Berg, J. & Carlson, R. Temporal changes in thesources of flood basalts: Isotopic and trace element evidence for the1100Ma old Keweenawan Mamainse Point Formation, Ontario, Canada.Geochim. Cosmochim. Acta 58, 4475–4490 (1994).

17. Symons, D. et al. Synopsis of paleomagnetic studies in the Kapuskasingstructural zone. Can. J. Earth Sci. 31, 1206–1217 (1994).

18. Lewchuk, M. & Symons, D. Paleomagnetism of the Late Precambrian ColdwellComplex. Tectonophysics 184, 73–86 (1990).

19. McFadden, P. & McElhinny, M. Classification of the reversal test inpalaeomagnetism. Geophys. J. Int. 103, 725–729 (1990).

20. McElhinny, M., McFadden, P. & Merrill, R. The time-averaged paleomagneticfield 0–5Ma. J. Geophys. Res. 101, 25007–25027 (1996).

21. Besse, J. & Courtillot, V. Apparent and true polar wander and the geometry ofthe geomagnetic field over the last 200Myr. J. Geophys. Res. 107, 2300 (2002).

22. Evans, D. Proterozoic low orbital obliquity and axial-dipolar geomagnetic fieldfrom evaporite palaeolatitudes. Nature 444, 51–55 (2006).

23. Tauxe, L. & Kent, D. in Timescales of the Paleomagnetic Field Vol. 145 ofGeophysical Monograph (eds Channell, J., Kent, D., Lowrie, W. & Meert, J.)101–116 (American Geophysical Union, 2004).

24. Elston, D. P., Enkin, R. J., Baker, J. & Kisilevsky, D. K. Tightening the belt:Paleomagnetic-stratigraphic constraints on deposition, correlation, anddeformation of the middle Proterozoic (ca. 1.4 Ga) Belt-Purcell supergroup,United States and Canada. GSA Bull. 114, 619–638 (2002).

25. Gallet, Y., Pavlov, V., Semikhatov, M. & Petrov, P. Late Mesoproterozoicmagnetostratigraphic results from Siberia: Paleogeographic implications andmagnetic field behavior. J. Geophys. Res. 105, 16481–16499 (2000).

26. Davis, D. & Green, J. Geochronology of the North American Midcontinentrift in western Lake Superior and implications for its geodynamic evolution.Can. J. Earth Sci. 34, 476–488 (1997).

27. Conrad, C. & Hager, B. Mantle convection with strong subduction zones.Geophys. J. Int. 144, 271–288 (2001).

28. Evans, D. True polar wander and supercontinents. Tectonophysics 362,303–320 (2003).

29. Kirschvink, J. The least-squares line and plane and the analysis of paleomagneticdata. Geophys. J. R. Astron. Soc. 62, 699–718 (1980).

30. Giblin, P. E. Preliminary Geological Maps 553–555 (Ontario Departmentof Mines, 1969).

AcknowledgementsWe thank D. Jones, C. Rose and N. Eichelberger for field assistance; C. Bayne of BayNiche Conservancy for land access; D. Kent and M. Jackson for assistance with hysteresisexperiments; J. Kasbohm for help drafting the geological map; J. Kirschvink for supportof pilot analyses at the Caltech Paleomagnetics Laboratory; and R. Mitchell and T.Kilian for support in the Yale palaeomagnetism facility, which was funded by NSF andthe David and Lucile Packard Foundation. This work has benefited from discussionswith D. Davis, J. Feinberg, H. Halls, D. Kent, B. Kopp, T. Raub and L. Tauxe. D. Biceencouraged N.L.S.-H. to begin study of Keweenawan volcanics when N.L.S.-H. wasan undergraduate at Carleton College. This work was financially supported througha Sigma Xi Grant-In-Aid awarded to N.L.S.-H., A.C.M.’s Agouron Post-doctoralfellowship and Princeton University. B.P.W. acknowledges support from the NSFGeophysics Program.

Author contributionsFieldwork was conducted by N.L.S.-H. (2 field seasons), A.C.M. (2 field seasons) andB.P.W. (1 field season) following initial project planning by A.C.M. and subsequentproject planning by A.C.M. and N.L.S.-H. Samples were analysed by N.L.S.-H. in thelaboratories of D.A.D.E. and B.P.W. N.L.S.-H. analysed the data, wrote the manuscriptand drafted Figs 2–3 with input from A.C.M., B.P.W. and D.A.D.E. Figure 1 was draftedby A.C.M. and N.L.S.-H.

Additional informationSupplementary information accompanies this paper on www.nature.com/naturegeoscience.Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions. Correspondence and requests for materials should beaddressed to N.L.S.-H.

NATURE GEOSCIENCE | VOL 2 | OCTOBER 2009 | www.nature.com/naturegeoscience 717© 2010 Macmillan Publishers Limited. All rights reserved.

In the version of this Letter originally published, Supplementary Tables S2, S3 and S4 contained several errors. Consequently, the estimates of the rate of motion on page 716 should have read: “The estimates for the rate of motion from 1,108 Myr to 1,097 Myr range between 21.5 ± 7.1 cm yr−1 for Osler Volcanics reversed → North Shore normal, to 33.6 ± 3.5 cm yr−1 for Coldwell Complex reversed → North Shore normal.” Furthermore, the position of the Duluth Complex reversed pole in Fig. 3a and the positions of the lower normal and upper normal palaeopoles in Fig. 3b were slightly incorrect and should have been as shown here. These errors have been corrected in the HTML and PDF versions of the text.

No asymmetry in geomagnetic reversals recorded by 1.1-billion-year-old Keweenawan basalts

Nicholas L. Swanson-Hysell, Adam C. Maloof, Benjamin P. Weiss and David A. D. Evans

Nature Geoscience 2, 713–717 (2009); published online: 13 September 2009; corrected after print: 13 April 2010.

corrigeNdum

Lower reversed 2

Lower reversed 1

Lower normal

Upper reversed

Upper normal

Mamainse Point palaeopoles

C

JF

ba

ABD 1141±1

DCr 1107±1

PMr 1107±2

CCr 1108±1NSr 1108±1OVr 1105±2

PLL 1095±3LST 1087±2

Keweenawan reversed

Pre-Keweenawan reversed

Keweenawan normal

Post-rift sediments

Mid-continent rift palaeopoles

NSn 1097±2

N

© 2010 Macmillan Publishers Limited. All rights reserved.