The motions between the Somalian, Antarctic, and Australian plates—the three plates believed to meet at the Rodrigues triple junction in the Indian Ocean—are inconsistent with the assumption that all three plates are rigid. The discrepancy is best explained if the Australian plate contains two component plates. Thus, the traditionally defined Indo-Australian plate consists of three component plates and multiple diffuse plate boundaries. The pattern of present deformation indicates that the boundaries between the three component plates are two unconnected zones accommodating divergence and a larger zone, which we interpret as three diffuse convergent plate boundaries and a diffuse triple junction.

Get full access to this article

View all available purchase options and get full access to this article.


Minster J. B., Jordan T. H., J. Geophys. Res. 83, 5331 (1978);
Stein S., Gordon R. G., Earth Planet. Sci. Lett. 69, 401 (1984).
Gordon R. G., Stein S., DeMets C., Argus D. F., Geophys. Res. Lett. 14, 587 (1987).
Royer J.-Y., Chang T., J. Geophys. Res. 96, 11779 (1991).
Stein S., Okal E. A., ibid. 83, 2233 (1978).
E. A. Bergman, J. L. Nabelek, S. C. Solomon, ibid. 89, 2425 (1984).
Stein S., Geophys. J. R. Astron. Soc. 55, 577 (1978);
Wiens D. A., Earth Planet. Sci. Lett. 76, 350 (1986).
Wiens D. A., et al., Geophys. Res. Lett. 12, 429 (1985).
DeMets C., Gordon R. G., Vogt P., Geophys. J. Int. 119, 893 (1994).
DeMets C., Gordon R. G., Argus D. F., J. Geophys. Res. 93, 11877 (1988).
Gordon R. G., DeMets C., ibid. 94, 5560 (1989).
___ and D. F. Argus, Tectonics 9, 409 (1990).
Royer J.-Y., Gordon R. G., DeMets C., Vogt P., Geophys. J. Int. 129, 41 (1997).
Herein, a direction of stretching refers to the orientation of the maximum principal axis of stretching. Similarly, a direction of shortening refers to the orientation of the minimum principal axis of stretching, or equivalently, the maximum principal axis of shortening.
Bergman E. A., Solomon S. C., Phys. Earth Planet. Inter. 40, 1 (1985).
Curray J. R., Moore D. G., Geol. Soc. Am. Bull. 82, 563 (1971);
Eittreim S., Ewing J., J. Geophys. Res. 77, 6413 (1972);
; D. G. Moore, J. R. Curray, R. W. Raitt, F. J. Emmel, Initial Rep. Deep Sea Drill. Proj. 22, 403 (1974);
Neprochnov Y. P., Levchenko O. V., Merklin L. R., Sedov V. V., Tectonophysics 156, 89 (1988);
Bull J. M., ibid. 184, 213 (1990);
Bull J. M., Scrutton R. A., Nature 344, 855 (1990);
; N. Chamot-Rooke, F. Jestin, B. de Voogd, Phèdre Working Group, Geology21, 1043 (1993).
Weissel J. K., Anderson R. N., Geller C. A., Nature 287, 284 (1980).
Geller C. A., Weissel J. K., Anderson R. N., J. Geophys. Res. 88, 1018 (1983).
McAdoo D. C., Sandwell D. T., ibid. 90, 8563 (1985).
D. T. Sandwell and W. H. F. Smith, ibid.102, 10039 (1997).
D. E. Petroy and D. A. Wiens, ibid.94, 12301 (1989).
Stein C. A., Cloetingh S., Wortel R., Geophys. Res. Lett. 16, 823 (1989).
Hebert H., et al., Eos 46, F684 (1996).
Royer J.-Y., Schlich R., J. Geophys. Res. 93, 13524 (1988).
The location of each crossing of the old end of anomaly 5 is identified along a ship-board magnetic profile or aeromagnetic profile, which come from many sources. To avoid the loss of resolution inherent in digitizing analog records, especially from published figures, we used digital data whenever available. The magnetic-anomaly crossings flanking the CIR are taken from (12), except for the southernmost crossing of anomaly 5 on the Indo-Australian side of the MOR, which we now believe lies over sea floor created at the SEIR (Fig. 3A). The magnetic-anomaly crossings flanking the SWIR (Fig. 3B) have been redetermined and incorporate many unavailable to (3). Crossings west of 46°E have been excluded to avoid the portion of the SWIR recording motion between the Antarctic and Nubian plates and the deforming zone assumed to exist between the Nubian and Somalian plates (44). Thus, the new set of crossings are expected to record motion between the Antarctic and Somalian plates. We have added new crossings and reidentified many other crossings of anomaly 5 flanking the SEIR. There is thus little overlap with Royer and Chang's (3) set of magnetic-anomaly crossings flanking the SEIR. As in (3), we exclude data east of ∼140°E to avoid the southeast corner of the Australian plate, which has been hypothesized to be deforming in response to convergence with the Pacific plate (9, 45).
The location of each crossing of a fracture zone is determined from satellite-derived gravity data. Along the CIR, the locations are crossings along individual processed profiles and are identical to those used by (12). Along the other two MOR systems, the crossings are interpreted from a gridded gravity map (19). The center of the fracture zone is assumed to lie at the center of the gravity trough for all fracture zones flanking the SWIR and CIR, which are spreading slowly and presumably resemble the morphology of the better studied fracture zones along the slowly spreading Mid-Atlantic Ridge. The signature of fracture zones flanking the SEIR is more complex. Royer and Sandwell (45) estimated the locations of fracture-zone crossings assuming that the maximum slope in the calculated gravity (that is, roughly midway between an adjacent gravity peak and trough) lies over the center of the fracture zone, as is expected at a fast-spreading MOR such as the East Pacific Rise. The gravity grid from declassified Geosat data (19) indicates that the signature of fracture zones flanking the SEIR varies considerably from fracture zone to fracture zone. Some zones appear highly antisymmetric with respect to the SEIR with a fracture-zone trough on one side of the MOR correlating with a fracture-zone ridge on the other side; this indicates that the maximum slope between fracture-zone ridge and trough on one side of the MOR should be correlated with that on the other. Other fracture zones, however, resemble those along slowly spreading MORs, with a fracture-zone trough on one side of the MOR correlating with a trough on the other side. We selected for analysis three fracture zones south of Australia resembling those on slowly spreading ridges (Fig. 3A). Also needed were sets of fracture-zone crossings along the westernmost SEIR, which lacks clear, straight fracture zones that are continuous between anomaly 5 and the SEIR. For this region, we correlated the midpoint between fracture-zone ridge and trough on one side of the MOR with the midpoint on the other side of the MOR for three fracture zones that clearly offset anomaly 5 by about 30, 50, and 70 km from the westernmost to the easternmost fracture zone, respectively (Fig. 3A). We investigated the self-consistency of the interpretation of fracture-zone crossings and the resulting reconstructions by examining the distance between the nearest magnetic-anomaly crossing to a given fracture zone and the distance of the same crossing when reconstructed across the MOR from what was interpreted as the correlative feature. The interpreted crossings are consistent with our best-fitting rotations. The self-consistency was further investigated by matching up conjugate wandering offsets lying just NW of a fracture zone that intersects the SEIR near 32°S, 77°E (Fig. 3A). The three offsets match up within ∼10 km, consistent with our interpretation.
The best-fitting rotations and their uncertainties were found with the use of the methods of Chang [(3);
Chang T., J. Am. Stat. Assoc. 83, 1178 (1988);
], which use the Hellinger fitting criterion [
Hellinger S. J., J. Geophys. Res. 86, 9312 (1981);
]. These methods allow us to estimate the dispersion of the data or subsets of the data. They produce a statistic r, which is expected to be chi-square distributed with ν degrees of freedom; r is the sum-squared normalized misfit of the magnetic-anomaly and fracture-zone crossings to the best-fitting model. The degrees of freedom ν equals N − 3m − 2n, where N is the number of magnetic-anomaly and fracture-zone crossings, m is the number of independent rotations to be estimated, and n is the number of data segments used. A data segment can consist of three or more fracture-zone crossings that sample the same fracture zone; at least one crossing must come from each side of a MOR. A data segment can also consist of three or more magnetic-anomaly crossings that sample the same MOR segment between fracture zones or a subset of a ridge segment if the segment is long; at least one crossing must come from each side of the present MOR. In the inversion procedure, a great circle is fit to each data segment and requires two adjustable parameters. The precision parameter κ̂ is the ratio of ν to r. If a subset of crossings are input to the rotation parameter-fitting program, each with a nominal error of σnom, then the precision parameter can be used to estimate the standard deviation of the crossings as σnom/κ̂1/2. We typically use σnom = 10 km. We use the estimated standard deviations as an aid in formulating error budgets for the data. To test for the consistency with plate-circuit closure of plate motion data along three plate boundaries meeting at a triple junction, we typically use the following chi-square test. We estimate a statistic Δr, which is the difference in r between a model in which closure is enforced about the three-plate circuit and a model in which closure is unenforced. In the latter case, r = r1 + r2 + r3, where each ri is the r for a best-fitting rotation, as described above. Three is the number of independent rotations estimated (that is, m). Thus, r is expected to be chi-square distributed with three degrees of freedom.
Uncertainties in the data were mainly assigned from preliminary estimates of the standard deviation of subsets of data. Error assignments for crossings along the CIR are identical to those assigned by (12). In brief, all fracture-zone crossings along the CIR were assigned errors of 5 km. Magnetic-anomaly crossings along the CIR were divided into four groups, mainly based on when the magnetic profiles were acquired, with the oldest profiles being assigned the largest uncertainties (1σ error of 5.2 km), the most recently acquired profiles being assigned the smallest error (1σ error of 3.0 km), and the two intermediate groups being assigned errors of 3.2 km and 3.9 km. Magnetic-anomaly crossings along the SWIR were assigned a 1σ error of 4 km, whereas fracture-zone crossings along the SWIR were assigned a 1σ error of 5 km. The dispersion of the data indicates that the uncertainties should be 30 to 40% smaller, but we are reluctant to shrink the assigned errors further. Along the SEIR, fracture-zone crossings were assigned a 1σ error of 6 km. West of ∼78°E along the SEIR, where we reexamined the magnetic-anomaly profiles in detail and where the dispersion is low, we assigned a 1σ error of 4 km to the magnetic-anomaly crossings. The dispersion of these crossings along the western SEIR indicate that the error should be about three times smaller, but we are reluctant to assign smaller errors based on such a small data set. The 102 magnetic-anomaly crossings along the SEIR east of ∼85°E were assigned a 1σ error of 5 km, except for 12 crossings from the Eltanin cruises, which were assigned a larger error of 10 km because these profiles were collected in the 1960s, before the advent of satellite navigation, and have highly dispersed crossings of anomaly 5.
Inasmuch as the precision parameter κ̂ for individual plate pairs exceeds 1 for each plate pair (and exceeds 2 in two of the three cases), it seems more likely that we have overestimated rather than underestimated the errors.
All uncertainties following “±” signs in this paper are 95% confidence limits.
Even if we use the more restrictive F-ratio test for closure, which is equivalent to assuming that all errors assigned to the data are overestimated by a uniform multiplicative constant, a value of F of 2.196 is obtained. This value has a 9% probability of occurring by chance if the plates are rigid and is therefore consistent with closure.
We make a test analogous to that described in (26) when testing if a previously undemonstrated plate boundary intersects a known plate boundary. In this case, Δr is the difference between r for a model in which only one plate separates from a second plate along a known plate boundary and r for a model in which two distinct plates separate along the known plate boundary from a third plate. The latter value of r is given by r1 + r2.
If the fracture-zone crossings are omitted from the western SEIR data, the data are still significantly misfit. The value of Δr is 14.2 with three degrees of freedom, which has a probability of being exceeded by chance of less than 0.003.
The complete covariance matrix for the rotation is Cxx = 2.0735 × 10−6, Cxy = 0.1012 × 10−6, Cxz = −1.5894 × 10−6, Cyy = 5.8913 × 10−6, Cyz = −1.8641 × 10−6, and Czz = 4.3707 × 10−6 sr, where the x, y, and z axes parallel (0°N, 0°E), (0°N, 90°E), and 90°N, respectively.
DeMets C., Gordon R. G., Argus D. F., Stein S., Geophys. J. Int. 101, 425 (1990);
; R. G. Gordon, in Global Earth Physics: A Handbook of Physical Constants, T. J. Ahrens, Ed. (AGU Reference Shelf 1, American Geophysical Union, Washington, DC, 1995), pp. 66–87.
This low rate of spreading likely occurs in the Red Sea (44).
D. F. Argus, thesis, Northwestern University (1990); D. F. Argus and R. G. Gordon, unpublished manuscript.
DeMets et al. (8) recently reexamined the consistency of 0- to 3-Ma plate motion data with closure about the RTJ. If they assumed that all the data uncertainties were overestimated by a uniform multiplicative constant, their data are inconsistent with closure at the 0.03% significance level. This result is not completely convincing, however. If they test for closure using the errors assigned to the data, the nonclosure is insignificant. Moreover, the errors they assigned to azimuths of transform faults along the CIR are realistic but inconsistent with those carried over from earlier work along the other two ridge systems. Therefore, further analysis of the 0- to 3-Ma data is required before the case for or against significant nonclosure can be made convincing. In any event, their results place an upper bound on the integral of possible deformation rate around this circuit of ∼4 to 7 mm/year, which is large enough to allow deformation as large as we find here.
M. J. Tinnon et al. [J. Geophys. Res.100, 24315 (1995)] used earthquake moment tensor data to estimate the pole of rotation between previously defined Indian and Australian plates. Their pole of rotation (10°S, 81°E), although differing insignificantly from what we now interpret as the India-Capricorn pole of rotation, is in even better agreement with the new pole of rotation of India relative to Australia. This is encouraging because many of the earthquakes they analyzed are from what we now interpret as the Capricorn-Australia and India-Australia deforming zones, as well as the triple junction where the three zones of shortening meet.
Molnar P., Nature 335, 131 (1988).
R. L. Fisher, M. Z. Jantsch, R. L. Comer, General Bathymetric Chart of the Oceans (GEBCO) (Canadian Hydrographic Service, Ottawa, 1982), sheet 5.09.
Wilson J. T., Nature 207, 343 (1965);
McKenzie D., Parker R. L., ibid. 216, 1276 (1967);
Morgan W. J., J. Geophys. Res. 73, 1959 (1968);
; X. Le Pichon, ibid., p. 3661.
Plate motions predicted across a narrow plate boundary are very specific, consisting of an explicit prediction of the displacement or velocity across or along any point on the narrow plate boundary. Motions predicted across a wide boundary specify only the integral of the deformation or velocity gradient along a path crossing the wide boundary and connecting a point on one rigid plate to a point on another [J. B. Minster and T. H. Jordan, in Tectonics and Sedimentation Along the California Margin, J. K. Crouch and S. B. Bachman, Eds. (Pacific Section of the Society of Economic Mineralogists and Paleontologists, Los Angeles, CA, 1984), vol. 38, pp. 1–16].
Argus D. F., Gordon R. G., J. Geophys. Res. 101, 13555 (1996).
D. Chu and R. G. Gordon, in preparation.
Royer J.-Y., Sandwell D. T., J. Geophys. Res. 94, 13755 (1989).
Levchenko O. V., Tectonophysics 170, 125 (1989).
This work was supported by CNRS and Plan National de Télédétection Spatiale, the NSF-CNRS U.S.-France cooperative program (NSF grant INT-9314549), and NSF grant OCE-9596284 (R.G.G.). The figures were drafted using the GMT software [P. Wessel and W. H. F. Smith, Eos372, 441 (1991)]. Géosciences Azur contribution number 111.


eLetters is an online forum for ongoing peer review. Submission of eLetters are open to all. eLetters are not edited, proofread, or indexed. Please read our Terms of Service before submitting your own eLetter.

Log In to Submit a Response

No eLetters have been published for this article yet.

Information & Authors


Published In

Volume 277 | Issue 5330
29 August 1997

Submission history

Received: 4 February 1997
Accepted: 13 June 1997
Published in print: 29 August 1997


Request permissions for this article.



Jean-Yves Royer
J.-Y. Royer is with Géosciences Azur, BP 48, 06235, Villefranche sur mer Cedex, France. R. G. Gordon is with the Department of Geology and Geophysics, MS 126, Rice University, Houston, TX 77005, USA.
Richard G. Gordon
J.-Y. Royer is with Géosciences Azur, BP 48, 06235, Villefranche sur mer Cedex, France. R. G. Gordon is with the Department of Geology and Geophysics, MS 126, Rice University, Houston, TX 77005, USA.

Metrics & Citations


Article Usage


Export citation

Select the format you want to export the citation of this publication.

Cited by


    View Options

    Check Access

    Log in to view the full text


    AAAS login provides access to Science for AAAS Members, and access to other journals in the Science family to users who have purchased individual subscriptions.

    Log in via OpenAthens.
    Log in via Shibboleth.
    More options

    Purchase digital access to this article

    Download and print this article for your personal scholarly, research, and educational use.

    Purchase this issue in print

    Buy a single issue of Science for just $15 USD.

    View options

    PDF format

    Download this article as a PDF file

    Download PDF

    Full Text








    Share article link

    Share on social media