of fresh water. This is enough to balance our measured salinity with a mean sea-level change of 135 m. Increas- es in ice-shelf volumes can also balance the salt budget. Given a total volume of 0.7  10 6 km 3 for all the Antarctic ice shelves (50), there would have to be seven times this amount of floating ice at the LGM to balance our data with the sea-level constraints. 47. G. H. Denton, T. J. Hughes, in The Last Great Ice Sheets, G. H. Denton, T. J. Hughes, Eds. (Wiley, New York, 1981), pp. 437. 48. B. P. Boudreau, Diagenetic Models and Their Imple- mentation (Springer-Verlag, Berlin, 1997). 49. P. A. Domenico, F.W. Schwartz, Physical and Chemical Hydrology (Wiley, New York, 1990). 50. D. J. Drewry, Ed., Antarctica: Glaciological and Geophysical Folio, sheets 2–9, Scott Polar Research Institute, University of Cambridge, Cambridge, UK (1983). 51. S.-T. Kim, J. R. O’Neil, Geochem. Cosmochim. Acta 61, 3461 (1997). 52. We thank K. Cuffey for insightful discussions about hydrology and E. Boyle and an anonymous reviewer for helpful comments. E. Boyle is thanked for inspi- ration and continued encouragement. E. Goddard provided assistance with lab work. Supported by NSF grant numbers OCE-0096814 to J.F.A. and OCE-0096909 to D.P.S. 17 July 2002; accepted 29 October 2002 Hybrid Speciation in Experimental Populations of Yeast Duncan Greig, 1,2 Edward J. Louis, 3 Rhona H. Borts, 3 Michael Travisano 2 * Most models of speciation require gradual change and geographic or ecological isolation for new species to arise. Homoploid hybrid speciation occurred readily between Saccharomyces cerevisiae and Saccharomyces paradoxus. Hybrids had high self-fertility (about 82%), low fertility when backcrossed to either parental species (about 7.5%), and vigorous growth under different thermal environ- ments that favored one or the other of the parental species. Extensive karyo- typic changes (tetrasomy) were observed in the hybrids, although genic in- compatibilities accounted for 50% of the variation in self-fertility. Speciation is thought to arise by gradual evolu- tion of genic incompatibilities (1), ecological specialization (2, 3), or chromosomal differenc- es (4 ) that prevent mating or cause inviable or infertile hybrid offspring (5). Rapid species for- mation can potentially occur by hybridization; however, the degree of reproductive isolation between potential new hybrid species and the two parental species is a major limiting factor. Hybrids must be self-fertile and suffi- ciently reproductively isolated to maintain a distinct lineage, but reproductive barriers be- tween parental species must not preclude the initial hybridization. In postzygotically isolated species, where hybrids are typically inviable or sterile (6 ), these conflicting requirements can be achieved by a doubling of chromosome complement in the new species to produce an allotetraploid (7 ). Potentially, these require- ments can also be met by maintaining chromo- some number (homoploid hybrid speciation) (8, 9), but this mechanism is very uncommon in plants and unknown in animals (10). Saccharomyces yeast species are postzy- gotically isolated, because hybrids form readily but are sterile, producing only 1% viable gametes (spores) (11–13). However, popula- tions of yeast can be very large (10 8 ), and viable gametes can be easily obtained. More- over, the ability of Saccharomyces gametes to divide and switch mating type allows for autofertilization (gametophytic selfing) and, po- tentially, for instantaneous homoploid hybrid speciation. We investigated this potential with Saccharomyces cerevisiae and Saccharomyces paradoxus and measured the effects of intrinsic incompatibilities (hybrid sterility and infertility) and extrinsic incompatibilities (relative fitness of hybrids under different environmental con- ditions) (14 ). First, we crossed S. cerevisiae and S. para- doxus and isolated 80 independent viable hap- loid gametes from their F 1 hybrid offspring. After allowing for spontaneous hybrid diploid formation by autofertilization (15), we found that 81.25% were capable of sporulation and that fertility (spore viability) was high (medi- an = 90%; mean = 84.40%, with 95% confi- dence interval of 73.75 to 92.67%) (Fig. 1A) (15). Fertility was slightly reduced from that of the parental species (S. cerevisiae, 99.93%, 99.04 to 99.79%; S. paradoxus, 99.21%, 97.80 to 99.92%) (11), with statistically significant variation among F 2 hybrids (F 61,260 = 15.72, P  0.0001). We tested for reproductive isola- tion of the fertile F 2 hybrids from the parental species (Fig. 1B). The backcross hybrids have fertility that is significantly higher (7.54%, 5.38 to 10.02%) than that of F 1 hybrids (0.03%, 0.00 to 0.18%) (11), but they have fertility that is much lower than that of the F 2 hybrids (F 1,895 = 817.02, P  0.0001). Although rare, hybrid F 2 diploids are both fertile and isolated from their parental species. Crossing F 2 hybrids and assessing fertility of their hybrid offspring demonstrated the existence of multiple different highly fertile F 2 hybrids (15). Ten independent F 2 geno- types, each having 100% fertility, were ran- domly paired and used to generate F 3 hy- brids. All pairs yielded some viable gametes, but the average fertility of F 3 hybrids (10.64%, 0.93 to 28.97%) was much lower than that of their immediate parents; also, there was genetic variation in fertility among the F 3 hybrids caused by interaction between the F 2 parental genomes (F 4,94 = 5.65, P  0.001). Nevertheless, autofertilized F 4 hybrid diploids derived from the viable gametes had particularly high fertility (97.33%, 92.10 to 1 The Galton Laboratory, Department of Biology, Uni- versity College London, Gower Street, London WC1E 6BT, UK. 2 Department of Biology, University of Hous- ton, 4800 Calhoun Road, Houston, TX 77204, USA. 3 Department of Genetics, University of Leicester, Uni- versity Road, Leicester LE1 7RH, UK. *To whom correspondence should be addressed. E- mail: mtrav@uh.edu Fig. 1. Reproductive isolation of sporulation- proficient F 2 hybrids. (A) Hybrids have high fertility when crossed with themselves. (B) Hy- brids have low fertility when crossed with ei- ther parental species (squares, S. cerevisiae; triangles, S. paradoxus). R EPORTS www.sciencemag.org SCIENCE VOL 298 29 NOVEMBER 2002 1773