JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO. B2, PAGES 2785-2814, FEBRUARY 10, 1996 Geomagnetic polarity bias paRems through the Phanerozoic Thomas J. Algeo Department of Geology, University of Cincinnati, Cincinnati, Ohio Abstract. Phanerozoic geomagnetic polarity biaspattems havebeen reconstructed using polaritydatafrom 278 stratigraphic formations of Cambrian-Jurassic age combined with data from an established geomagnetic polarity timescale for the Cretaceous-Recent. In addition to the well-known Cretaceous Normal PolaritySuperchron andKiamanReversed Polarity Superchron, other first-order polarity features arerecognized: (1) a Middle Cambrian-Middle Ordovician Burskan Reversed Polarity BiasInterval, (2) a Late Ordovician-Late Silurian Nepan Normal Polarity BiasInterval, (3) an EarlyJurassic Normal Polarity BiasInterval, and, possibly, (4) a Middle Jurassic Normal Polarity Bias Interval.A combination of strong polarity biasand low reversal rates during the Ordovician may indicate the existence of a "dual-polarity superchron" containing a single majorpolarity transition, the Middle Ordovician PolarityShift. Reconstruction of an accurate Phanerozoic polarity trend permits application of a "polarity bias test" to evaluate the primary character of magnetic remanences. A polarity biastestof British Siluro-Devonian remanences reveals thatgroup "A" remanences (00-20 ø palcopole latitude; 30-100% normal polarity) exhibit polarity concordance with coeval non-British remanences, whereas group "B" remanences (250-50 ø palcopole latitude; 0-20%normal polarity) arestrongly discordant, suggesting that the latter are largelyof secondary origin. Analysis of groups of magnetic remanences also permits estimation of (1) characteristic timescales for formation polarity data and(2) evaluation of sources of age-dependent polarity-ratio variance. For the Cambrian-Jurassic polaritydataset,formations exhibita meancharacteristic timescale of 1.0- 1.5 m.y., and circa 50% of polarity-ratio variance is attributable to palcomagnetic sampling of a binomial variable (i.e., geomagnetic field polarity) and50% to other factors (i.e., stochastic and systematic depositional andsampling biases, incorrect ageestimates for stratigraphic formations andcharacteristic remanences, complex magnetizations, andlow epochal reversal frequencies). Introduction Long-term variations in geomagnetic field behaviorare thought to reflect changes in conditions within the core andat the core-mantle boundary associated with generation of the Earth's magnetic dynamo [e.g., Merrill andMcElhinny, 1983; Jacobs, 1994].Such variations have been correlated with mantle plume activity [Loper and McCartney, 1986; Larson, 1991a: Larson and Olson,1991; Loper, 1992], globalclimate change [Larson, 1991b], eustatic elevations [Gaffin, 1987; Marzocchi etal., 1992], extinction events [Courtillot, 1990], and true polar wander [Courtillot and Besse, 1987]. Most studies of long-term geomagnetic variation have focused on reversal frequency [e.g., McElhinny, 1971; Lowrie and Kent, 1983; Mazaud and Laj, 1991; Marzocchi and Mulargia, 1992; Johnson et al., 1995] or statistical differences between the normal and reversed polarity states [e.g., Merrill et al., 1979:McFadden and Merrill, 1984; McFadden et al., 1987] rather than onpolarity bias. However, the lack of research on secular variation and controls on geomagnetic field polarity isunfortunate because (1) polarity bias patterns are subject to less uncertainty than reversal rates, which are strongly influenced by inclusion of short (circa <40 ka), poorly known magnetic events and excursions [e.g., LaBrecque et al., 1977; McFadden and Merrill, 1984; McFadden et al., 1987], and (2) suchpaRems may provide important insights regarding geodynamo operation. The lastcomprehensive study of Phanerozoic polarity bias paRems wasthatof Irving andPullaiah [1976],and the advent of improved magnetic cleaning methods and publication of substantial amounts of paleomagnetic datain the intervening period warrant a reexamination of such paRems using an updated polarity data set. Therefore the goals of this study were (1) to assemble a polarity data set for the Cambrian-Jurassic derived largely from paleomagnetic studies published since 1975, (2) to evaluate sources of age-dependent variance among formation polarity ratios, and (3)toconstruct a polarity trend for the Phanerozoic. Important features of the present study include thefollowing: (1) application of a setof rigorous criteria for acceptance of characteristic magnetic remanences as primary (i.e., approximately syndepositional), (2) determination of characteristic timescales associated with formation polarity ratios and polarity trends, and (3) use of a binomial probability model to analyze paRems of variance in polarity data and toidentify and selectively removeunrepresentative polarity ratios. These procedures produced a data set that is smaller (n = 278) butinternally more consistent thanthat of Irving and Pullaiah[1976],allowing construction of a reliable polarity trend for the Phanerozoic. Copyright 1996 by the American Geophysical Union. Papernumber 95JB02814. 0148-0227/96/95JB-02814505.00 Definitions "Polarity ratio,"as used in thisstudy, is the ratioof normal polarity to totalpolarity (i.e.,normal plusreversed) relative to 2785