Comment on “A Diverse Assemblageof Late Cretaceous Dinosaur andBird Feathers from Canadian Amber”Carla J. Dove1* and Lorian C. Straker1,2

McKellar et al. (Reports, 16 September 2011, p. 1619) analyzed Late Cretaceous amber specimensfrom Canada and identified some filaments as dinosaurian protofeathers. We argue that theiranalysis and data do not provide sufficient evidence to conclude that such filaments are feather-likestructures. Further investigation, including destructive sampling, must be carried out for moreconvincing conclusions.

McKellar et al. (1) recently examinedLate Cretaceous amber specimens fromCanada and reported stages I through

Vof feather evolution according to Prum’s (2)model. They concluded that the amber specimenscontained dinosaur feathers because (i) specimensUALVP 52821 [figure 1B in (1)] and UALVP52822 [figure 1, C and D, in (1)] were compa-rable to nonavian dinosaur fossil compressions;and (ii) fibers in specimen TMP 96.9.334 [fig-ure 2, A to C, in (1)] exhibited microstructurecoiling comparable to modern birds. Because theycould not identify the fibers as any organism ofan “end-member” evolutionary-developmentalspectrum, and the fibers occurred concurrentlywith modern feather types, they inferred that thefibers were dinosaurian.

The interpretations of figure 1, B to D, in (1);figure 2, A to C, in (1); and the supporting fig-ures in (3) convince us that adequate analysiswas not conducted on these specimens and thatoverstated conclusions were made on subjectiveobservations. Other figures in (1) (figure 2, D toF, and figure 3) are comparable with the feathermicrostructure in modern birds and cannot beregarded as anything but the ultimate stage offeather evolution.

The filaments described as stage I [UALVP52821, figure 1B in (1)] are an order of magni-tude smaller than the filaments of compressionfossils of Sinosauropteryx prima (4) with whichthey were compared. According to McKellar et al.(1), the largest width of the filaments in UALVP52821 falls within the range of the S. prima’sintegument structures; however, the cited work[(5), p. 1719] reported “smaller ones are consid-erably narrower than 0.1 mm” in diameter. Thisdoes not represent a comparable scale to the

0.027 mm [(3), p. 3] measurements from thespecimen in amber. Further, other researcherswho examined the S. prima impressions reportedmeasurements no thinner than 0.05 mm anddeclared the fibers to be collagen and not pro-tofeathers (6). Although the filament lengthsof UALVP 52821 were not measured in thisstudy, the authors report that these “are consist-ent with the finer filaments found in this spec-imen [S. prima], and fall within the range ofobserved lengths” [(3), p. 5].

The study reports diameter measurements forUALVP 52821 as being within the lower rangeof modern hair measurements (of Australianmammals), but the authors curiously excludedhair as a possible fiber based on diameter andhollowness [(3), pp. 3–4 and figure S1]. Ourinterpretation of figure S4, B to D, and figureS5B in (3) is that the fibers do show internaldivisions and do not appear to be hollow the entirelength of the filament (Fig. 1, A and B). Addi-tionally, comparing the amber fibers to specimensof fossil hair found in Canada (TMP 96.9.998)and France (dated Early Cretaceous) does notexclusively rule out UALVP 52821 as includinghair filaments based on surface texture (cross-hatching) and diameter alone [figure S4, B to E,in (3)]. This analysis is open to subjective inter-pretation based on the published images.

In the spinning disk confocal microscopy(SDCM) and laser scanning confocal micros-copy (LSCM) analysis, McKellar et al. (1) statethat the b-keratin autofluorescence results wereinfluenced by background interference of thetested amber specimens, so we are confused asto why they conclude that UALVP 52821 [(3),p. 7)] contains b-keratin. Although the graphshave similar profiles, the intensity peaks andwavelength excitation values for b-keratin aredifferent [figure S10, B and E in (3)], and wequestion the assumption that the materials aresimilar in nature.

McKellar et al. (1) explicitly state that reflectiveemissions of UALVP 52821 and TMP 96.9.997are inconclusive [(3), pp. S7–S8] because valuescould be from a reflective surface when the spec-

imen pulled away, or from fractures in the am-ber and not from b-keratin matrix. We feel thatthis represents an insufficient conclusion basedon circumstantial evidence.

Another flaw in this particular analysis isthat comparisons were made only to a feather inTMP 96.9.997 and not to the specimen that theyreported as being hair (TMP 96.9.998). Confocalmicroscopy comparing the unknown fiber tothe amber specimen containing hair would bemore appropriate and could possibly rule outa-keratin as a component.

Because the reported stage II morphotypes[figure 1, C and D, in (1) and figure S5 in (3)] didnot undergo the same analytical tests as UALVP52821, we cannot assume that those fibers are thesame [(1), p. 1620] and believe that the authorsshould have conducted similar analyses on thatspecimen instead of basing conclusions on visualobservations.

The interpretations of fibers in TMP 96.9.334[figure 2, A to C in (1)] are not analogous to mod-ern feather microstructures of birds. Featherbarbules (pennula) in modern birds are cylindricaland do not coil in both right and left directionson the same axis (Fig. 2A), nor do the twistsextend distally. Twisting on modern birds onlyoccurs on the flattened, straplike cells of bases[as shown in figure S12E in (3)]. McKellar et al.’sconclusions are invalid because (i) the amberspecimen shows coiling at mid and distal posi-tions of the fiber [figure 2A in (1)]; (ii) no basecell is observable; (iii) internodes are not flat-tened in modern birds; and (iv) figures S6A,S6B, S7A (lower arrow), and S7B (3) do not clear-ly show a rachis structure or a barb ramus, butrather seem to be a concentration of filamentsfocused on a central structure (observed onfigure S7B). We have observed similar coilingon fibers of seed hairs (Fig. 2B), that is, Populustrichocarpa.

Although exploring amber specimens forclues to feather evolution may seem novel, thisstudy lacks evidence and vigor to conclude thatthe fibers in UALVP 52821, UALVP 52822, andTMP 96.9.334 are dinosaurian. The analysis wasnot complete for each specimen, did not conclu-sively rule out hair or specialized plant parts aspossible fibers, makes incorrect comparisons tomodern feather microstructure, and cannot becited as early stages of feather evolution. Becausethe topic of dinosaur feathers has been disputed,we feel that better analysis of the material inquestion, including destructive sampling of theamber specimens, is paramount.

Without concise identification of the variousfilaments depicted, there is no basis for assigningany of them to a particular group of organisms, tosay nothing of dinosaurs.

References and Notes1. R. C. McKellar, B. D. Chatterton, A. P. Wolfe, P. J. Currie,

Science 333, 1619 (2011).2. R. O. Prum, J. Exp. Zool. 285, 291 (1999).3. Supporting online material for (1).

TECHNICALCOMMENT

1Smithsonian Institution, National Museum of Natural History,Division of Birds, MRC 116, Post Office Box 37102, Wash-ington, DC 20013–7012, USA. 2Universidade Federal do Riode Janeiro, Instituto de Biofísica Carlos Chagas Filho, Labo-ratório de Ultraestrutura Celular Hertha Meyer, Ilha do Fundão,Rio de Janeiro, RJ, 21949-900, Brazil.

*To whom correspondence should be addressed. E-mail:dovec@si.edu

17 FEBRUARY 2012 VOL 335 SCIENCE www.sciencemag.org796-b

on

Feb

ruar

y 20

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

4. P.-J. Chen, Z.-M. Dong, S.-N. Zhen, Nature 391, 147(1998).

5. P. J. Currie, P.-J. Chen, Can. J. Earth Sci. 38, 1705(2001).

6. T. Lingham-Soliar, A. Feduccia, X. Wang, Proc. Biol. Sci.274, 1823 (2007).

Acknowledgments: The Feather Identification Lab at theSmithsonian is funded through interagency agreements withthe U.S. Air Force, the U.S. Navy, and the U.S. Federal AviationAdministration. L. C. Straker is a predoctoral fellow fundedthrough the Coordenação de Aperfeiçoamento de Pessoalde Nível Superior/Fulbright Commission, Brazil.

4 November 2011; accepted 5 January 201210.1126/science.1216208

Fig. 1. (A to C) Images from(3) of filaments within UALVP52821 [(A) and (B)] and UALVP52822 (C), with white arrowsadded to point to our inter-pretation of possible internalsepta, cell walls, or other fea-tures that appear as evidenceto dispute hollowness. (D) Pho-tomicrograph of cottonwoodseed (Populus tremuloides) withenlarged inset showing sim-ilar overall characteristics tothe filaments in the amberspecimens (15 mm width andoutside wall thickness is aver-age 42% of total width). Scalebar, 0.1 mm).

www.sciencemag.org SCIENCE VOL 335 17 FEBRUARY 2012 796-b

TECHNICAL COMMENT

on

Feb

ruar

y 20

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

Fig. 2. Similarities of morpholog-ical characteristics of amber speci-men and seed hair filaments. (A)Image from McKellar et al. (1) ofTMP 96.9.334, with white arrowsadded to show the coiled distal fila-ments. Scale bar, 0.2 mm. (B) Cotton-wood seed hairs (Populus trichocarpa)with thick arrows showing similarcoiling, and arrowheads showing semi-flattened areas similar to filamentobserved in figure 2C in (1) (fila-ment width of 9 to 11 mm). Scale bar,0.1 mm.

17 FEBRUARY 2012 VOL 335 SCIENCE www.sciencemag.org796-b

TECHNICAL COMMENT

on

Feb

ruar

y 20

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

Response to Comment on “A DiverseAssemblage of Late CretaceousDinosaurand Bird Feathers from Canadian Amber”Ryan C. McKellar,1* Brian D. E. Chatterton,1 Alexander P. Wolfe,1 Philip J. Currie2

Dove and Straker question our interpretations of plumage from Late Cretaceous Canadian amber.Although we are able to refute concerns regarding both specimen taphonomy and misidentificationas botanical fossils, unequivocal assignment to either birds or dinosaurs remains impossible,as we stated originally. However, reported observations and their further refinement herein areinsufficient to falsify the hypothesized dinosaurian origin for protofeathers.

Dove and Straker (1) raise concerns overwhether our recent Report (2, 3) of plu-mage from Late Cretaceous Canadian

amber contains sufficient evidence to supportconclusions of both avian and dinosaurian ori-gins for inclusions. Specifically, they question twospecimens (UALVP 52821 and UALVP 52822)interpreted as representatives of stages I and IIin Prum’s (4) evolutionary-developmental model,as well as a specimen (TMP 96.9.334) interpretedas representative of stage IV with specializationsfor water uptake. Although some uncertainty isinherent to all interpretations of fragmentary fos-sils, Dove and Straker’s (1) technical concernscan be addressed readily with the specimens.

Concerning TMP 96.9.334, there is a funda-mental error in their interpretation (1) of ourpaper, because we did not suggest that this spec-imen represents dinosaur plumage. Instead, weproposed that this specimen likely belongs tostage IV (3), representing Cretaceous bird plu-mage with structural features linked to its possi-ble use in water uptake (2, 3). Points raised (1)regarding fine anatomical detail are the same fea-tures that suggested to us that placement withinstage IV was more appropriate than stage V.Other criticisms (1) represent different interpre-tations of our discussion and figures of specimentaphonomy (3).

Dove and Straker (1) are correct in assertingthat there is no clear indication of unsegmentedbasal cells within TMP 96.9.334 barbules andthat it is unusual to observe coiling within thesegmented portion of the barbule (pennulum),because these features are unknown in modernbird feathers. As Dove has illustrated and dis-cussed [figure 138 in (5)], some modern (stage V)bird plumage retains traces of segmentation or

“multiple cells” within the curled basal portionof the barbule. The lack of a defined basal platedoes not exclude TMP 96.9.334 from being afeather fragment, whereas the suggestion thatthese filaments are more comparable to Populusseed trichomes (1) is incorrect. Although coiled,trichomes are not segmented (noded) (6, 7) andwould not be expected to consistently straightenapically. Furthermore, Salicaceae, the family that

contains Populus, is unknown before the latestPaleocene (8).

Taphonomic effects that led Dove and Straker(1) to critique TMP 96.9.334 were described inour original work [(3), p. S9], but benefit fromadditional clarification. The TMP 96.9.334 bar-bules all coil in a counter-clockwise direction. Thefive or so barbules (Fig. 1A) that coil clockwisehave been sheared from the main mass, likely asa result of resin flow (3). These dislocated bar-bules do not indicate “coiling at mid and distalpositions” (1), and their nodal expansions clearlyindicate that they are pointing in the wrong di-rection. The central structure in TMP 96.9.334 isalmost certainly a rachis (Fig. 1B): It is solid, nota clump of filaments, and has the appropriate sizeand cross-sectional shape (9).

Concerning UALVP 52821, Dove and Straker(1) chose to emphasize filament size over mor-phology, critiquing the match between dimen-sions in Sinosauropteryx protofeathers and thefilaments in Canadian amber. Overlooking thefact that UALVP 52821 appears to contain justthe distal tips of filaments (3), filament diam-eters are not “an order of magnitude smaller” (1)than those reported for Sinosauropteryx. In fact,the work (10) cited by Dove and Straker (1) con-tains no diameter measurements, because the

TECHNICALCOMMENT

1Department of Earth and Atmospheric Sciences, University ofAlberta, Edmonton, Alberta, T6G 2E3, Canada. 2Department ofBiological Sciences, University of Alberta, Edmonton, Alberta,T6G 2E9, Canada.

*To whom correspondence should be addressed. E-mail:rcm1@ualberta.ca

Fig. 1. Photomicrographs of TMP 96.9.334 and comparison image of a trichome preserved inCanadian amber. (A) Overview of TMP 96.9.334, with dislocated barbules indicated by arrows atbarbule bases. (B) Cross section of rachis where it is truncated at the surface of the amber piece,with rachis surrounded by truncated barbule apices. (C) Stellate trichome (University of Albertaspecimen).

17 FEBRUARY 2012 VOL 335 SCIENCE www.sciencemag.org796-c

on

Feb

ruar

y 20

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

authors had difficulty isolating single filaments.Later work provided generalized estimates thatincluded smaller filaments “considerably narrower

than 0.1 mm” (11) and cautioned that measure-ments may be overestimates due to filament over-lap (11) and taphonomy (11, 12). If the structures

measured by Lingham-Soliar et al. (12) wereindeed collagen fibers (internal structures), theirmeasurements have no bearing on our study ofintegumentary structures. Subsequent work (13)indicated that many of the filaments surroundingSinosauropteryx and Sinornithosaurus were pig-mented and thus cannot be dismissed as collagenfibers. This also bears on the comparison be-tween UALVP 52821 and 52822 with Populustrichomes [figure 1 in (1)]. Trichomes lack pig-mentation (Fig. 1C) and are often branched ormulticellular (6, 7). And again, Salicaceae areunknown from the Cretaceous (8).

Dove and Straker (1) appear to advocatemammalian hair as a better match to UALVP52821 than Sinosauropteryx protofeathers. Al-though they critique our use of diameter andhollowness as criteria for excluding mammalianhair, they overlook our most important criterion,cuticular scales. UALVP 52821 lacks cuticularscales, which was discussed at length (3). Theinternal divisions envisaged by Dove and Straker[figure 1, A to C, in (1)] are taphonomic artifacts(Fig. 2).

Dove and Straker (1) misinterpreted the re-sults of our spinning disk confocal microscopy(SDCM) and laser scanning confocal microscopy(LSCM) analyses (3), which were reported asinconclusive for identifying keratin, because am-ber autofluorescence overwhelms any autofluo-rescence signal of keratin in inclusions (3). Thereis no possibility of distinguishing between a andb-keratin within this amber, and we did not claimto have identified keratin based on these analy-ses (3). Our interpretation of structure withinUALVP 52821 in no way hinges upon LSCMdata; the hollow, cylindrical structure of the fila-ments is plainly visible using light microscopy[figures S3B, S4A, and S11B in (3)]. We analyzedunequivocal feather fragments (TMP 96.9.997)for comparison to UALVP 52821 because the un-pigmented zone of these specimens provided thebest chance of detecting keratin (3). Keratin wasnot identifiable using emission profiles in TMP96.9.997, which rendered the analysis of addi-tional specimens futile. The same problem oc-curred with UALVP 52821.

We understand the recommendation for fur-ther analyses and destructive sampling (1).However, UALVP 52821 and UALVP 52822are currently the only representatives of putativeprotofeathers in amber. The 11 plumage fragments(2, 3) stem from a survey of more than 4000inclusion-bearing pieces of amber. At a minimum,~100,000 amber pieces were collected (14, 15)over a span of decades to obtain these samples.Destructive analysis of such rare specimens is notjustified when it is unlikely to produce irrefutableresults (3). The specimens were not found with di-nosaur skeletal remains, so their biological affinitywill remain open to debate. At present, the evidencesupports a dinosaurian source for the filaments.Dove and Straker (1) have suggested neither anew analytical technique nor amore parsimoniousinterpretation for the specimens involved.

Fig. 2. Additional taphonomic details within photomicrographs of UALVP 52821 and 52822. (A)Diagrammatic explanation of taphonomic and structural features in UALVP 52821, bright-field image[figure S4C in (3)]. Stippling indicates topography, as well as the apparent cross-hatched surface orwrinkles within the outer wall of each filament. In the left filament, the gray area denotes an internalbubble in the vicinity of a large hole in the filament. In the right filament, horizontal arrows indicate cracksin the outer wall that do not extend all the way across the filament, and the filament appears to havecollapsed along its midline, likely as a result of bending or torsion. (B) Taphonomic features in UALVP52821 [figure S4D in (3)]. This is a dark-field image, highlighting surface details. Fractures within theamber appear as bright spots and are present in positions removed from the filaments (horizontal arrows);dark regions correspond to holes within the filaments (inclined arrows). (C and D) “Internal divisions”noted by Dove and Straker in UALVP 52822 (1) represent variations in pigmentation or cracks in the outerwall of the filament (inclined arrows) and do not appear to correspond to segmentation or internaldivisions. (D) Bright spots in a linear arrangement (on the filament indicated with a horizontal arrow)appear to correspond to cross-hatching observed in UALVP 52821.

www.sciencemag.org SCIENCE VOL 335 17 FEBRUARY 2012 796-c

TECHNICAL COMMENT

on

Feb

ruar

y 20

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

References and Notes1. C. J. Dove, L. C. Straker, Science 335, 796 (2012);

www.sciencemag.org/cgi/content/full/335/6070/796-b.2. R. C. McKellar, B. D. E. Chatterton, A. P. Wolfe,

P. J. Currie, Science 333, 1619 (2011).3. Supporting online material for (2).4. R. O. Prum, J. Exp. Zool. B 285, 291 (1999).5. C. J. Dove, Ornith. Mono. 51, 1 (2000).6. J. C. T. Uphof, Handb. Pflanzenanat. 4, 1

(1962).7. E. Werker, Adv. Bot. Res. 31, 1 (2000).

8. J. E. Eckenwalder, in Biology of Populus and ItsImplications for Management and Conservation,R. F. Stettler et al., Eds. (NRC Research Press,Ottawa, 1996), pp. 7–32.

9. A. M. Lucas, P. R. Stettenheim, Avian Anatomy:Integument (U.S. Department of Agriculture, Washington,DC, 1972).

10. P.-J. Chen, Z.-M. Dong, S.-N. Zhen, Nature 391, 147 (1998).11. P. J. Currie, P.-J. Chen, Can. J. Earth Sci. 38, 1705 (2001).12. T. Lingham-Soliar, A. Feduccia, X. Wang, Proc. Biol. Sci.

274, 1823 (2007).

13. F. Zhang et al., Nature 463, 1075 (2010).14. J. F. McAlpine, J. E. H. Martin, Can. Ent. 101, 819

(1969).15. E. M. Pike, Palaios 8, 411 (1993).

Acknowledgments: Funding for this work was provided byNatural Sciences and Engineering Research Council of Canadadiscovery grants issued to B.D.E.C., A.P.W., and P.J.C.

6 December 2011; accepted 5 January 201210.1126/science.1216484

17 FEBRUARY 2012 VOL 335 SCIENCE www.sciencemag.org796-c

TECHNICAL COMMENT

on

Feb

ruar

y 20

, 201

2w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from