Nature Macmillan Publishers Ltd 1998

8There is a long-standing historical contro-versy over the question of US PresidentThomas Jeffersons paternity of the childrenof Sally Hemings, one of his slaves14. Tothrow some scientific light on the dispute,we have compared Y-chromosomal DNAhaplotypes from male-line descendants ofField Jefferson, a paternal uncle of ThomasJefferson, with those of male-line descen-dants of Thomas Woodson, Sally Hemingsputative first son, and of Eston Hemings Jef-ferson, her last son. The molecular findingsfail to support the belief that Thomas Jeffer-son was Thomas Woodsons father, but pro-vide evidence that he was the biologicalfather of Eston Hemings Jefferson.

In 1802, President Thomas Jefferson wasaccused of having fathered a child, Tom, bySally Hemings5. Tom was said to have beenborn in 1790, soon after Jefferson and SallyHemings returned from France where hehad been minister. Present-day members ofthe AfricanAmerican Woodson familybelieve that Thomas Jefferson was the fatherof Thomas Woodson, whose name comesfrom his later owner6. No known docu-ments support this view.

Sally Hemings had at least four morechildren. Her last son, Eston (born in 1808),is said to have borne a striking resemblanceto Thomas Jefferson, and entered white soci-ety in Madison, Wisconsin, as Eston Hem-ings Jefferson. Although Estons descendantsbelieve that Thomas Jefferson was Estonsfather, most Jefferson scholars give morecredence to the oral tradition of the descen-dants of Martha Jefferson Randolph, thepresidents daughter. They believe that SallyHemings later children, including Eston,were fathered by either Samuel or PeterCarr, sons of Jeffersons sister, which wouldexplain their resemblance to the president.

Because most of the Y chromosome ispassed unchanged from father to son, apartfrom occasional mutations, DNA analysis ofthe Y chromosome can reveal whether ornot individuals are likely to be male-line rel-atives. We therefore analysed DNA from theY chromosomes of: five male-line descen-dants of two sons of the presidents paternaluncle, Field Jefferson; five male-line descen-dants of two sons of Thomas Woodson; onemale-line descendant of Eston Hemings Jefferson; and three male-line descendantsof three sons of John Carr, grandfather of Samuel and Peter Carr (Fig. 1a). No Y-chromosome data were available frommale-line descendants of President ThomasJefferson because he had no surviving sons.

Seven bi-allelic markers (refs 712),eleven microsatellites (ref. 13) and the mini-satellite MSY1 (ref. 14) were analysed (Fig.1b). Four of the five descendants of Field

Jefferson shared the same haplotype at allloci, and the fifth differed by only a singleunit at one microsatellite locus, probably amutation. This haplotype is rare in the pop-ulation, where the average frequency of amicrosatellite haplotype is about 1.5 percent. Indeed, it has never been observed out-side the Jefferson family, and it has not beenfound in 670 European men (more than1,200 worldwide) typed with the microsatel-lites or 308 European men (690 worldwide)typed with MSY1.

Four of the five male-line descendants ofThomas Woodson shared a haplotype (withone MSY1 variant) that was not similar tothe Y chromosome of Field Jefferson but wascharacteristic of Europeans. The fifth Wood-son descendant had an entirely differenthaplotype, most often seen in sub-SaharanAfricans, which indicates illegitimacy in theline after individual W42. In contrast, thedescendant of Eston Hemings Jefferson did

have the Field Jefferson haplotype. The hap-lotypes of two of the descendants of JohnCarr were identical; the third differed byone step at one microsatellite locus and byone step in the MSY1 code. The Carr haplo-types differed markedly from those of thedescendants of Field Jefferson.

The simplest and most probable expla-nations for our molecular findings are thatThomas Jefferson, rather than one of theCarr brothers, was the father of Eston Hem-ings Jefferson, and that Thomas Woodsonwas not Thomas Jeffersons son. The fre-quency of the Jefferson haplotype is lessthan 0.1 per cent, a result that is at least 100times more likely if the president was thefather of Eston Hemings Jefferson than ifsomeone unrelated was the father.

We cannot completely rule out otherexplanations of our findings based on ille-gitimacy in various lines of descent. Forexample, a male-line descendant of Field

NATURE | VOL 396 | 5 NOVEMBER 1998 | www.nature.com 27

Jefferson fathered slaves last childscientific correspondence

Ancestrya b Haplotypes

Bi-allelicmarkers

Microsatellite STRs MinisatelliteMSY1

0000001 15.12.4.11.3.9.11.10.15.13.7 (3) 5.(1)14.(3)32.(4)16

0000001 15.12.4.11.3.9.11.10.15.13.7 (3) 5.(1)14.(3)32.(4)16

0000001 15.12.4.11.3.9.11.10.15.13.7 (3) 5.(1)14.(3)32.(4)16

0000001 15.12.4.11.3.9.11.10.15.13.7 (3) 5.(1)14.(3)32.(4)16

0000001 15.12.4.11.3.9.11.10.16.13.7 (3) 5.(1)14.(3)32.(4)16

0000001 15.12.4.11.3.9.11.10.15.13.7 (3) 5.(1)14.(3)32.(4)16

0000011 14.12.5.12.3.10.11.10.13.13.7 (1)17.(3)36.(4)21

0000011 14.12.5.11.3.10.11.10.13.13.7 (1)17.(3)37.(4)21

0000011 14.12.5.12.3.10.11.10.13.13.7 (1)17.(3)36.(4)21

0000011 14.12.5.11.3.10.11.13.13.13.7 (1)16.(3)27.(4)21

0000011 14.12.5.11.3.10.11.13.13.13.7 (1)16.(3)27.(4)21

0000011 14.12.5.11.3.10.11.13.13.13.7 (1)16.(3)27.(4)21

1110001 17.12.6.11.3.11.8.10.11.14.6 (0?)1.(3a)3.(1a)11.(3a)30.(4a)14.(4)2

0000011 14.12.5.11.3.10.11.13.13.13.7 (1)16.(3).28.(4)20

Peter President Thomas Jefferson

ThomasJefferson

II

Field

J5 J12 J20

J13 J21 J31 J37 J43 J47

J29 J35 J41

J30 J36 J42

J6

J14 J23 J33 J39

J45 J49

J46 J50

SallyHemings

Eston H10 H15 H17 H21

Thomas C6 C11

C8 C13 C21 C26

C7 C12 C20 C24 C28 C30 C31

C19 C23 C27

C29JohnCarr

Dabney

Overton

W40 W55

W41 W56

W57 W69

W58 W70

W8 W27

Lewis

James

W9 W28 W42

W12 W30 W46 W61

ThomasWoodson

FFiigguurree 11 Male-line ancestry and haplotypes of participants. a, Ancestry. Numbers correspond to referencenumbers and names in more detailed genealogical charts for each family. b, Haplotypes. Entries in bold high-light deviations from the usual patterns for the group of descendants. BBii--aalllleelliicc mmaarrkkeerrss.. Order of loci: YAP-SRYm8299-sY81-LLY22g-Tat-92R7-SRYm1532. 0, ancestral state; 1, derived state. MMiiccrroossaatteelllliittee sshhoorrtt ttaannddeemmrreeppeeaattss ((SSTTRRss)).. Order of loci: 19-388-389A-389B-389C-389D-390-391-392-393-dxys156y. The number of repeatsat each locus is shown. MMiinniissaatteelllliittee MMSSYY11.. Each number in brackets represents the sequence type of therepeat unit; the number after it is the number of units with this sequence type. For example, J41 has 5 units ofsequence type 3, 14 units of sequence type 1, 32 units of sequence type 3, and 16 units of sequence type 4.

Nature Macmillan Publishers Ltd 1998

8

fall into two classes, according to whetherthe vacuoles in medullary keratin scatterlight coherently (with a relationshipbetween the phases of waves scattered bydifferent surfaces) or incoherently. In theRayleigh and Mie scattering models4,6, thecolours are produced by incoherent scatter-ing of smaller visible wavelengths by anarray of spatially independent air vacuoles inthe medullary keratin. Alternatively, in theconstructive interference model, the coloursare produced by interactions between lightwaves scattered coherently by the keratinairmatrix13,5. Out-of-phase waves will destruc-tively interfere and cancel out, whereas in-phase waves will constructively reinforceone another and be reflected coherently. Thephase relationships of scattered waves aredetermined by the size and spatial distribu-tion of the scatterers1,2,7. Periodic spatialrelationships between scatterers will pro-duce consistent path-length differencesamong scattered waves and reinforce a lim-ited set of wavelengths.

To investigate how the structural colourof feather barbs is produced, we apply anelectromagnetic theory of coherent lightscattering7 that was developed to explainthe transparency of the human cornea. In a

quasi-random array of scatterers, coherentscattering is predicted only for light waveswith a wavelength of twice that of the largestcomponents of the Fourier transform of thespatial variation in refractive index7. Weperformed a discrete 2D Fourier analysis ofdigitized transmission electron micrographsof cross-sections of the medullary keratinfrom the blue feather barbs of C. maynanato examine whether it has periodic spatialvariation in refractive index. The spongymedullary layer of C. maynana barbs con-sists of a matrix of keratin bars with air vac-uoles of strikingly uniform diameter andspacing2. The matrix has no planes of sym-metry and is randomly orientated to thesurface of the feather. The 2D Fourier powerspectrum of the medullary keratin matrixshows a nearly circular ring around the ori-gin 0.0059 nm21 in diameter (Fig. 1a). Thisring indicates that the tissue has a highlyordered, nanoscale spatial variation inrefractive index that is nearly uniform in alldirections.

Reflection spectra of the blue featherbarbs display a peak between 500 and 520nanometres (Fig. 1b). We used the 2DFourier power spectrum to predict thereflectance spectrum of these barbs. The

scientific correspondence

28 NATURE | VOL 396 | 5 NOVEMBER 1998 | www.nature.com

Jefferson could possibly have illegitimatelyfathered an ancestor of the presumed male-line descendant of Eston. But in the absenceof historical evidence to support such possi-bilities, we consider them to be unlikely.Eugene A. Foster*, M. A. Jobling, P. G. Taylor, P. Donnelly, P. de Knijff,Rene Mieremet, T. Zerjal, C. Tyler-Smith*6 Gildersleeve Wood, Charlottesville, Virginia 22903, USAe-mail: eafoster@aol.comDepartment of Genetics, University of Leicester, Adrian Building, University Road, Leicester LE1 7RH, UKDepartment of Statistics,University of Oxford, South Parks Road, Oxford OX1 3TG, UKMGC Department of Human Genetics, Leiden University, PO Box 9503, 2300 RA Leiden, The Netherlands Department of Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU, UK

1. Peterson, M. D. The Jefferson Image in the American Mind

181187 (Oxford Univ. Press, New York, 1960).

2. Malone, D. Jefferson the President: First Term, 18011805

Appendix II, 494498 (Little Brown, Boston, MA, 1970).

3. Brodie, F. M. Thomas Jefferson: An Intimate History (Norton,

New York, 1974).

4. Ellis, J. J. American Sphinx: The Character of Thomas Jefferson

(Knopf, New York, 1997).

5. Callender, J. T. Richmond Recorder 1 September 1802. [Cited in:

Gordon-Reed, A. Thomas Jefferson and Sally Hemings: An American

Controversy (Univ. Press of Virginia, Charlottesville, 1997)].

6. Woodson, M. S. The Woodson Source Book 2nd edn

(Washington, 1984).

7. Hammer, M. F. Mol. Biol. Evol. 11, 749761 (1994).8. Whitfield, L. S., Sulston, J. E. & Goodfellow, P. N. Nature 378,

379380 (1995).

9. Seielstad, M. T. et al. Hum. Mol. Genet. 3, 21592161 (1994).10. Zerjal, T. et al. Am. J. Hum. Genet. 60, 11741183 (1997).11. Mathias, N., Bayes, M. & Tyler-Smith, C. Hum. Mol. Genet. 3,

115123 (1994).

12. Kwok, C. et al. J. Med. Genet. 33, 465468 (1996).13. Kayser, M. et al. Int. J. Legal Med. 110, 125133 (1997).14. Jobling, M. A., Bouzekri, N. & Taylor, P. G. Hum. Mol. Genet. 7,

643653 (1998).

Coherent light scatteringby blue feather barbs

The structural colours of avian featherbarbs are created by the scattering of lightfrom the spongy matrix of keratin and airin the medullary layer of the barbs15. How-ever, the precise physical mechanism for theproduction of these colours is still contro-versial1,3,4,6. Here we use a two-dimensional(2D) Fourier analysis of the spatial varia-tion in refractive index of the blue featherbarbs of the plum-throated cotinga (Cotingamaynana, Cotingidae) to show that thecolour is produced by constructive interfer-ence between light waves scattered coher-ently by the nanostructured keratinairmatrix of the barbs.

Mechanisms proposed to explain the pro-duction of structural colours of feather barbs

FFiigguurree 11 Two-dimensional Fourier power spectrum and observed and predicted reflectance spectra of a bluefeather barb of Cotinga maynana. a, 2D Fourier power spectrum of the spongy medullary keratin matrix of ablue feather barb. b, Observed reflectance spectrum of a blue feather barb (black squares, right axis) and pre-dicted reflectance spectrum (bars, left axis) based on the 2D Fourier power spectrum (a). The discrete Fouriertransform describes how a signal or image is composed of different periodic component waves10. A 2D Fourieranalysis was conducted using the 2D Fast Fourier Transform algorithm in MATLAB11 on a digitized 500-pixel2

image of a transmission electron micrograph (magnification, 3 29,000)2. The image was processed in MATLABusing variance rescaling with block processing and median filtering11. The 2D Fourier power spectrum (a)resolves the spatial variation in optical density in the tissue into its periodic components in any direction in theplane from all points. The size and direction of any component spatial frequency is given by the length anddirection of a vector from the origin to that point. The squared amplitude of the component is given by thecolour (scale bar on the right). The ring indicates a highly ordered, nanoscale fluctuation in optical densitymoving in any direction in the plane of the tissue at a spatial frequency of about 0.0059 nm21. Reflectance ofblue feather barbs (b, squares, right axis) was measured using a Zeiss microspectrophotometer at 30 wave-lengths between 400 and 700 nm12. The predicted reflectance spectrum (b, bars, left axis) expresses the per-centage of the total power distributed among the components of the 2D Fourier power spectrum over alldirections. The predicted reflection spectrum was calculated using the radial average of one quadrant of the2D Fourier power spectrum with half-pixel width intervals, normalizing the volume under the rotated radial aver-age function to one, calculating the volume under radial sections of the radial average Fourier power function,multiplying these volumes by twice the mean refractive index (1.283), estimated from the micrographs usingthe refractive indices of keratin (1.54) and air (1.00)2, and plotting these values as wavelengths in nanometres.

1.125

0.750

0.375

0

0.375

0.750

1.125Spa

tial f

requ

ency

(nm

1

10

2 )

Spatial frequency (nm1 102) Wavelength (nm)1.125 0.750 0.375 0 0.375 0.750 1.125

5

4.5

4

3.5

3

2.5

2

1.5

1

0.5

0

107

16

14

12

10

8

6

4

2

0

35

30

25

20

15

10

5

% T

otal

Fou

rier

pow

er

% R

efle

ctan

ce

ba

300 350 400 450 500 550 600 650 700