SPREAD-WING POSTURES AND THE WATER REPELLENCY OF FEATHERS: A TEST OF RIJKE'S HYPOTHESIS

A. M. ELOWSON Departmentof Zoology,University of Wisconsin,Madison, Wisconsin 53706 USA

ABSTRACT.--Rijke(1967, 1968) proposed that the water repellency of feathers and the presenceor absenceof spread-wingpostures in water birds could be explainedby a structural mechanismfirst describedfor textiles.The textile model predictsthat the tendencyof water dropletsto bead up on grid-like surfacesis a mathematicalfunction in which the primary independent variable is an index, (r + d)/r, where r is the radius of and d is one-half the distancebetween cylinders in the grid. Largerindices indicate more water-repellent surfaces. Rijke found larger indices in the feathers of ducks than in the feathers of cormorantsand anhingas;hence, he concludedthat the latter birds mustspread-wing to dry their wettable feathers. However, there were mathematical inconsistencies, undefined variables and con- cepts,and inadequatedata in Rijke'spapers. Despite these flaws, Rijke's hypothesis has been frequentlycited in the ornithologicalliterature. I report here my evaluationof the applica- bility of the textile model to feathersand my test of Rijke's prediction that speciesthat assumespread-wing postures when wet have smaller (r + d)/r values for their ramus and barbulestructure than speciesthat do not. I usedscanning electron microscopy to measure the featherstructure of 14 speciesof water birds in 6 differentcategories (breast, back, and four regionsof a remex). I found that the textile-featheranalogy is not realistic,because feather structureis considerablymore complexand variable than the geometricmodel that is fundamentalto the textileequations. My (r + d)/r valuesshow considerable overlap among three behaviorally distinct groups of water birds: those that predictably,occasionally, or never assumespread-wing postures.Statistically, the (r + d)/r values of the rami in some feather categoriesof the group of speciesthat showsspread-wing behavior were smaller than thoseof the other two groupsof birds (which did not differ). Index valuesof the barbule structure,which constitutesmost of the feathersurface, however, do not differ significantly amongthe three groupsof birds. I alsomeasured the shapeof water droplets(by contact angles) on the breast and remex feathersof a Mallard (Anasplatyrhynchos) and a Reed Cor- morant (Phalacrocoraxafricanus) and comparedthese values between the two speciesas well as with thosemathematically predicted by the textile model. In general,the observedwater dropletshave a shapemore like that predictedby the (r + d)/r values of barbulesthan of rami. Droplets on the feathersof the Reed Cormorantwere more bead-shapedthan those on Mallard feathers, although the reverse should be true if the textile model holds for feathers.I concludethat Rijke's hypothesisis invalid for two reasons:the textile model cannotbe appliedreliably to feathers,and it doesnot accountfor the spread-wingbehavioral differencesamong water birds. Received12 April 1983,accepted 12 October1983.

RIJKE(1967, 1968)published two nearly iden- however, reviewed Rijke's papers (1967, 1968; tical papers in which he proposed that a math- hereinafter reference to Rijke refers to these ematical model derived to explain the water years unless other datesare given) and pointed repellency of textiles could also resolve three out several problems, most notably, several ornithological issues:(1) the water repellency mathematical inconsistencies, an undefined of feathers;(2) the absenceof spread-wingpos- variable, and an unsupported assumption. De- tures in most speciesof water birds; and (3) the spite thesedrawbacks, Rijke's conclusionshave function of this behavior in cormorants and an- been generally accepted in the ornithological hingas. Rijke's model provided a plausible an- literature (see below). swer to the paradox that Townsend (in Bent Hailman recommendedthat I look into Rijke's 1922: 241) pointed out: if spread-wing postures paperswhen I was preparing a note on my field are necessaryto dry the wings, why is the be- observation of an Osprey (Pandionhaliaetus) havior not shown by water birds other than performing spread-wing behavior (Elowson- cormorantsand anhingas?Hailman (1969a, b), Haley 1982). In doing so, I found several more 371 The Auk 101: 371-383. April 1984 372 A.M. ELOW$ON [Auk,Vol. 101 inconsistencies(detailed below) that prompted me to reevaluate the hypothesis with respect to issues(1) and(2) above. I present the results of that study in this paper. I have not consid- ered here the function of spread-wing pos- tures,which is a complexquestion that has been investigated elsewhere (Bernstein and Maxson 1982, Hennemann 1982, Winkler 1983, and the references therein). Fig. 1. Contactangles of water dropletsresting Rijke basedhis hypothesison a resemblance on (a) wettableand (b) water-repellentsurfaces. of feather structure to textile structure. The tex- tile model (Cassie and Baxter 1944, Baxter and Cassie1945) statesthat the water repellency of by a bird at any given time). Despitethe fact porous surfaces increases with the size of an that Rijke did not evaluatethe wettability of index that is calculated from the dimensions of the feather coat as a whole, several authors have elements in the surface structure (i.e. the citedhim asproviding a mechanismthat makes threads). Cassie and Baxter (1944) even sug- this suggestionplausible (Clark 1969,Kennedy gestedthe applicability of their model to feath- 1971,Kahl 1971,George and Casler1972, Sieg- ers, which may have motivated Rijke to test the fried et al. 1975,Mahoney 1981,Bernstein and idea. Using light microscopy,he measured the Maxson1982, Hennemann 1982).Finally, Ken- feather structure of seven speciesand found nedy (1969) has cited Rijke'swork as having larger indices in ducks than in cormorants. establisheda drying function for the spread- Therefore, he concluded that cormorants, un- wing posturesof cormorantsand anhingas. like ducks,have wettable feathers that require Neither Rijke's hypothesisnor my test of it drying by spread-wing postures. is comprehensiblewithout someexplanation of Rijke's work has been noted in the literature the textilemodel itself. In what follows,I pre- in two somewhat different contexts: feather sent that model followed by Rijke's data and structure determines its water repellency and finally the predictions that should hold if the the feathers of cormorantsand anhingas are hypothesis were true. wettable. Someauthors have cited the original The textilemodel and feathers.--Thedistinc- textile papers (Cassie and Baxter 1944, Baxter tion between "water-repellent" and "water- and Cassie1945) as having establishedthe first proof" surfacesneeds clarification, as authors concept (Thompson 1953; Kennedy 1970a, b, (Rowen and Gagliardi 1947, Crisp 1963) have 1972; Rutschke 1976). Rutschke (1960) took mea- pointed out multiple uses of the latter term. surements from mallard feathers and con- Truly waterproof materials,such as a yellow curred with the estimates Cassie and Baxter rain slicker, are almost impermeable to water. (1944) gave.Many more authors,however, have Water-repellent surfaces, by contrast, cause cited Rijke's conclusionswith respectto feather water to bead up and roll off under brief ex- structure and water repellency (Clark 1969; posuresin ordinary atmosphericconditions, but Kennedy 1970b; Stettenheim 1972, 1976; suchsurfaces will becomewetted upon extend- Rutschke1976; Rhijn 1977;Schreiber 1977; Jones ed exposureor under increasedpressure. Den- 1978; Winkler 1983). Bernstein and Maxson im fabricsand feathersare water-repellentsur- (1982) discussed Rijke's conclusions in both faces,and Rijke'smodel deals with repellency contexts.Although they measuredthe feather per se. elementsof the Antarctic Blue-eyedShag (PhaI- The textile model (Cassie and Baxter 1944, acrocoraxatriceps) to contrast with Rijke's data Baxter and Cassie 1945) is considered to be the for other speciesof cormorants,they noted stateof the art for the water repellencyof po- Hailman's (1969a, b) objections to the model. roussurfaces (Crisp 1963).The water repellen- Without reference to the underlying process, cy of a surfaceis determinedby whether water McAtee and Stoddard (1945) and Owre (1967) dropletson it beadup (repellent) or flatten and proposed that cormorantsand anhingas have a spread out (wettable) (Fig. 1). Thus, surface wettable plumage or feather coat (nomencla- wettability can be operationally expressedas ture from Humphrey and Parkes 1959; the the angle made by the surfaceand a tangent to feather coat is the aggregateof feathers worn a droplet's curvature at the point of contact, April 1984] WaterRepellency ofFeathers 373

(see Fig. 2). Cassieand Baxter's (1944) mathe- matical statementfor water repellency is:

cos0'u = f• cos0• - f2, (1)

where O'•is the effective receding contactangle on the porous surface; 0R is the intrinsic receding contact angle on the same solid ma- terial as is in the porous surface, but smooth Fig. 2. Crosssection of a pore coveredby a water (i.e. not porous);and f• is an expressionfor the droplet many times larger in size; r is the radius of ratio of solid/watercontact and f2 for air/water the cylindersand d is •,5the distancebetween adja- contact. Thus, equation (1) is an expressionfor •centcylinders. 0 is an angleused to derivef, andf2. the shape of a water droplet on a porous sur- (The meniscusof water is ignored.) facecaused by the affinity of the solid material for water (0•) and by the structure, or topog- raphy,of the surface(f, and f2).From geometric measuredthrough the liquid (see Fig. 1). This properties of the structure, Cassie and Baxter is the contact angle O;it is intrinsicto the sur- (1944) derived the following equations for f• face material. As Fig. 1 shows, larger angles and f2: describe more spherical droplets, and there- fore, more water-repellent surfaces. f• = [,rr/(r + d)][1 - (0A/180ø)] (2) Cassieand Baxter (1944) showed that surface and porosity changesthe intrinsic contactangle to a larger effective contact angle (0') because f2 = 1 - r sin OA/(r + d), (3) poresunder the droplet add air spacesfor which water moleculeshave little affinity. Cassieand where r is the radiusof the cylinders,and d is Baxter termed the new angle the "apparent one-half the distance between adjacent cylin- contactangle" (0h)and Rijke used this expres- ders (seeFig. 2). 0h is the intrinsic advancing sion as well as "effective contactangle." I use contactangle made on a nonporoussolid of the the latter term. same material as is in the porous surface. The Two kinds of intrinsic contactangles are im- effective advancingcontact angle 0'^ can be cal- portant to Cassieand Baxter's(1944) equations. culated by substituting 0h for 0u in equation They can be visualized in rain drops individ- (1). These equations,though not intuitive, have ually trickling down a dirty window pane. The gained credibility by applicationin diversedis- rounded advancing front of a droplet has a ciplines (Disc. Faraday Soc. 1948, Advancesin larger angle, the advancing contact angle 0h Chem. 1964, Moilliet 1963, and Zografi pers. (not to be confusedwith Rijke's effective con- comm.). Their derivations are available in the tact angle 0h). This advancing angie deter- given references.Cassie and Baxter (1944: 549) mines the penetration of water into porous sur- showed that, if 0h and 0• are treated as con- faces such as feathers (Cassie 1948). The stants,their model predictslarger 0'ufor larger following tail of the droplet hasa much smaller valuesof (r + d)/r (i.e. differently gaugedgrids angle, the receding contact angle 0u, that de- of the samematerial). I term this expressionthe termines the water repellency of such surfaces. "structuralindex," or simply "index." Note that Rijke discussedintrinsic contact angles but did the index gives a dimensionlessvalue. not differentially label them in his equations. The variable terms in the three equations are The structuralfactors in the textile equations applied to feathers as follows. In a general describe the porous surface, which Cassieand sense, the elements of feather structure can be Baxter (1944) illustrated as a crosssection where considered analogousto the parallel cylinders a seriesof circlesrepresent parallel cylinders of the textile model (Fig. 2). The rami [nomen- (i.e. the threads)(Fig. 2). Specifically,these fac- clature from Lucas and Stettenheim (1972); the tors describe the areas of solid/water contact barb is a primary branch,or ramus,plus its dis- (i.e. on the cylinders)and air/water contact(i.e. tal and proximal barbules] radiate from the over the pore) per unit planar area of surface rachisin approximatelyparallel rows. Between 374 A.M. ELOWSON [Auk, Vol. 101 adjacent rami are distal and proximal barbules TABLE1. Rijke's data (see text for details). each in roughly parallel rows that have vary- ing angles with respect to the rami. Smaller Species (r + d)/r feather elements such as barbicels do not con- Mallard (Anasplatyrhynchos) 5.9 tribute significantly to the feather/water sur- African Shelduck (Tadorna cana) 5.8 face. Therefore, the ramus or barbule radius is Reed Cormorant (Phalacrocorax africanus) 4.3 r and 1/2the distancebetween adjacentrami or Bank Cormorant (P. neglectus) 4.5 barbulesis d. Note that feather structure gen- Cape Cormorant (P. capensis) 4.4 erates two indices, one for rami and another Great Cormorant (P. carbo) 4.8 for barbules.The intrinsic contactangles 0^ and African Darter (Anhingatufa) 4.5 0Rfor feathersare thoseformed by water drop- lets on the feather rachis (nonporous surface). Rijke's data.--Rijke tested one prediction-- whether the (r + d)/r values of the rami are prise a large portion of the feather surface, a larger in species that do not assume spread- credible test of structure-dependentwater re- wing postures (ducks) than in those that do pellency should evaluate them as well. In sum, (cormorants and anhingas). His results (Table Rijke's data are not sufficient to support his 1), show that the indicesof thesebehaviorally conclusions, because they represent several distinct groups can differ by as little as 1.0 (see feather types, have too few values for ducks, Table 1, the difference between the data for the are not replicable, and do not contain barbule African Shelduck and Great Cormorant). measurements. In addition to the problems Hailman (1969a, Moreover, as Hailman (1969a,b) pointed out, b) noted, Rijke's data are an inadequate test of Rijke did not present water repellency in op- his hypothesisfor several reasons.First, they erational terms. Rijke calculated the effective represent.more than one feather type. He stat- contactangles for only two species:143 ø for the ed that the African Shelduck index was from Mallard, which does not show spread-wing wing-feather measurementsand all of the oth- postures,and 121ø for the African Darter, which er indices were from breast feathers. In a later does. In these calculationsRijke assumedwith- footnote, however, Rijke (1970: 473) noted that out explanation that 0^is constantat 90ø and 0R the African Darter value (see Table 1) was for at 60ø. Also, without explanation, Rijke (1968: a quill (remex), and it should instead be listed 188) concluded that 121ø "is too small to effect as 11.0 for a breast feather. Second, Rijke con- indefinite pearling off." trasted the indices of five cormorant and an- I calculatedthe 0'R using Rijke's data and his hinga specieswith only two anatid indices, valuesfor 0Rand 0^in equations1-3 and found which were from entirely different types of 134 ø for the Mallard and 127 ø for the African feathers. This is not compelling evidence that Darter. Hence, thesespecies theoretically differ ducks do not adopt spread-wing posturesbe- by 7ø, not 22ø as Rijke reported--not a large causetheir larger structuralindices confer upon distinction. Possible differences in the water them water-repellent feathers.Third, Rijke did repellency of these two speciesare also con- not report where he measuredon the feathers. founded by the 0'Rcalculated with the one bar- This is an unfortunate omission, because the bule index Rijke gave: 4.7 for the Mallard. In rami and barbulesvary in crosssection and dis- this case, the 0'Ris 128ø, very nearly that pre- tance apart within and among feathers and dicted with Rijke's data for the African Darter. amongspecies (Chandler 1916,Lucas and Stet- Even more confusingis the angle calculatedfor tenheim 1972, Stettenheim 1976). Therefore, r the African Darter when Rijke's footnoted re- and d are variable within species,and the small vision of its index changed from 4.5 to 11.0. numerical difference Rijke observed between Then this specieshas an 0'Rof 147ø--largerthan ducks on the one hand and cormorants and an- the duck. In sum, the outcome of Rijke's incor- hingas on the other may be due to structural rect calculations and muddled data is the lack differences between feather types or regions of an operational definition for feather water and not taxonomic differences in structure. repellency. Fourth, the data in Table ! are based on mea- Predictions.--IfRijke's hypothesis were true, surementsof rami only. Becausebarbules corn- then the (r + d)/r values for rami of birds that April 1984] WaterRepellency of Feathers 375

TABLE2. Speciesfor which ramusand barbule dimensionswere measured.•

Spread-wing References Species behaviorb for behavior Common Loon (Gavia immer) N Palmer (1962) Pied-billed Grebe (Podilymbuspodiceps) N Palmer (1962) Reed Cormorant (Phalacrocoraxafricanus) Y Siegfried et al. (1975) Double-crested Cormorant (P. auritus) Y Lewis (1929) Anhinga (Anhingaanhinga) Y Clark (1969) African Darter (A. rufa) Y Cramp and Simmons(1977) Brown Pelican (Pelecanusoccidentalis) O Schreiber (1977) American White Pelican (P. erythrorhynchos) O Schaller (1964) Magnificent Frigatebird (Fregatamagnificens) O Cramp and Simmons (1977) Mallard (Anasplatyrhynchos) N McKinney (1965) EuropeanShelduck (Tadorna tadorna) N McKinney (1965) White-winged Scoter (Melanitta fusca) N McKinney (1965) Ruddy duck (Oxyurajamaicensis) N McKinney (1965) Osprey (Pandionhaliaetus) O Elowson-Haley (1982) Measurementstaken in each of six feather categories(see text). Refers to assumptionof spread-wing posture by wet birds: Y - predictably; O = occasionally;N - never.

predictably assumespread-wing postureswhen Each stub was grounded with silver paint (#1481, wet should be smaller than those of speciesthat Ernest F. Fullam, Schenectady,NY) and shaded with occasionallyassume the posture,which, in turn, a 0.02-0.03-gm-thicklayer of 60% gold and 40% pal- should be smaller than those of speciesthat ladium in a vacuum vaporizer (Denton). I took mi- never show this behavior. The sameprediction crographsat 100x with a JapanElectron'Optics Lab- oratory SEM Model JSM-U3 fitted with a' Polaroid should also be true for indices of the barbules. camera using Polaroid Land 4 x 5 film (Type 551 Also, if the theoretical textile model is viable Positive-Negative). with respectto feathers,then calculatedcontact On each micrograph, I outlined an area roughly angles should predict those observed. parallel with and 400-500 gm from the rachis and, with a metric rule, took 2 (when structures were so

METHODS AND MATERIALS large as to preclude more) to 8 measurementsof the rami and barbules (the latter at 50 am from the ra- Measurementof r andd.--I measuredthe ramusand mus).On mostmicrographs I could not measureboth barbule structureof 14 speciesin six feather cate- r and d at one spot,thereby finding its index. There- gories:breast, back, and four areasof a primary (re- fore, the indicespresented herein are calculatedfrom mex) (outer vane proximally and distally and inner the mean r and d values for each sample.As a mea- vane proximallyand distally). Unavoidably,the par- sure of sample variability in the index, I found the ticular (number) remex I used varied from speciesto maximumand minimum indicesfor eachsample giv- species.Table 2 lists the speciesand notes whether en all the possible(r + d)/r values from the separate or not and how consistentlythey assumespread-wing (not mean) r and d measurements. postures,based on the literature listed. In addition Measurementof contactangles.--I measuredcontact to these species,I also studied the feathers of the anglesby the sessiledrop method(Bigelow et al. 1946) Dipper (Cincluscinclus) in all six categories.The barø with a specialapparatus designed by B. Johnson(1982) bule and ramusstructure was so very different from and an opticalsystem commercially available through the 14 nonpasserinespecies I studied, however, that Gaertner Scientific Corporation of Chicago, Illinois. it was not possibleto measure the r and d dimen- The apparatusdelivered water droplets of known sions,let alone contrastthe indices of Dipper feath- volume onto feather samples(1 cm2) in a housing of ers to those of the other species. constant temperature and humidity. After delivering I used scanningelectron microscopy(SEM) to mea- a droplet of triple-distilled water, I allowed a 30-s sure the feather elements, because I could not resolve equilibration time and then took 4-6 readingsof the them with light microscopywith either reflectedor contactangle from each of the two visible sides of transmittedlight. To preparesamples for SEM, I cut the droplet.As is standardwith this technique,I con- 1-cm2 piecesfrom the feathersof one individual in sideredthe advancingangle to be that formed 30 s each species,taking care to cut in corresponding after delivery of the drop and the receding contact placesfor all species.I mounted the uniquely num- angle to be that formed 30 s after someof the liquid bered sampleswith the dorsal (i.e. outermost) side was removed. The feathers were not washed or treat- up on aluminum SEM stubswith double-sticktape. ed before this procedure. 376 A.M. ELOWSON [Auk,Vol. 101

Due to my limited use of the apparatus,I was able also apparent in their mean values (presented to measure0^, 0 R, 0'^, and 0'Rfor only the following in the Appendix). For example, in the Pied- Mallard and Reed Cormorant samples(cut from areas billed Grebe contrast the r values for the rami contiguous with those used for the SEM work): the in distal primary outer and inner vanes--11.65 primary rachises(0^ and 0• only), the breast feather •m to 22.50 •m, respectively. Also, the d values vanes;and the proximal primary inner vanes. I used the mean 0^ and 0Rvalues to calculatethe predicted for the rami in the Mallard proximal primary 0'^ and 0'Rfor these two species. remigesvary from 86.67 •m on the outer vane Statisticalanalysis.--I used nonparametric tests to to 145.00 •m on the inner vane. Lucas and Stet- avoid making assumptionsconcerning distributions. tenheim (1972: 259) reported equal or greater ! analyzed whether or not the structuralindex dif- variation within the same vane. fered significantlyamong the three groups(see Table In addition to supporting my view that the 2) of specieswith one-tailed Mann-Whitney U-Tests structuraldifference Rijke observedmay be due (Siegel 1956). Eachanalysis was a pair-wise compar- to differencesamong feathers and not among ison (within one feather category) with an Ho such taxa,the variability of structureevident in the that the (r + d)/r values did not significantly differ. micrographs and Appendix data challenge The level of significance(a) was 0.05. Rijke's hypothesis for another important but subtlereason. The model assumesparallel rows RESULTS of perfect cylinders, becausethe equationsfor f• and f2 are derived from a geometricanalysis Thefeather/textile analogy.--Figure 3 presents of the diagramatic model in Fig. 2. Because micrographsof the 6 feather categoriesfor 3 feather structuredoes not reasonablyconform species,representing the 3 behavioral groups to this physicalmodel, however, the reliability listed in Table 2. In all but two micrographs, of the textile equationsin predicting its contact the widest elements are rami, and the numer- angles, hence water repellency, is in doubt. ous smaller elements are barbules (BK and BR Index differencesin relationto spread-wingbe- in group B are the exceptions,where the widest havioraldifferences.--The (r + d)/r values I cal- element in each is the rachis). Note that bar- culated from the SEM measurementsare pre- bules constitute most of the feather surface area. sentedin Fig. 4. The figure is relatively complex With respectto the feather/textile analogy, and requires a few notes of explanation. In- Fig. 3 indicates that feather structure is much dices from the 4 primary and 2 body contour more complex than the ideal porous surface feather categoriesare presentedfor all species shown in Fig. 2. Note that the distal barbules in blocks(a) to (f). The speciesare grouped (in- branching from one ramus often cover the dis- dicated by stripes,stipples, or clear) according tally adjacent ramus--covered rami cannot to their spread-wing behavior as indicated in contribute to superficially determined water Table 2. Superimposedon the data for the rami repellency. The micrographs reveal that the are dashed and solid lines that correspond re- shapeof the rami varies (see especiallygroups spectivelyto Rijke's largest index for a species A and B) from somewhatflattened in the prox- with reputedly wettable feathers and to his imal primary (POV and PIV) to somewhat smallest index for a species with reputedly rounded on the distal primary (DOV and DIV). water-repellent feathers. The range indications Therefore, the r value of a ramus varies over at the end of each bar designate the maximum the feather surface. The d dimensions of rami and minimum (r + d)/r values given the vari- are also highly variable, especiallyon the pri- ability of r and d in each micrograph. maries (contrast the outer and inner vanes in Consider first Rijke's delimiting values in- any one row of micrographs). Stettenheim dicatedby the dashedand solid lines. Note that (1976) has likewise commented that the rami for many speciesmy data do not consistently of a feather vary in their angle of attachment support his water repellent/wettable dichoto- to the rachis.Figure 3 alsoreveals that the bar- my. For example, the indicies of the rami in bules do not have the uniform structure fun- the proximal primary remiges of the Mallard, damental to the textile model. Note that the European Shelduck (both, for outer vane data, barbule d distances in the breast and back Fig. 4b), and the Ruddy Duck (inner vane, Fig. feathersof all three speciesvary considerably. 4a) are within the range of thoseRijke reported The variability of r and d within and among as causingwettable feathers,yet these species feather types or regions of a given speciesis do not assumespread-wing postures.Further- April1984] WaterRepellency ofFeathers 377

, ;' . I '-. t"

' ,• ,, '• / • '•' , , •, '• ,,•, , , • •,- i • ' •K ,.' "'•/' • • "t', . • ,, •,.• /' ..,,

Fig. 3. SEM Eicrographs(100x) of the six feather categoriesfrom thr• species(see Table 2 as to their spread-wingbehavior): (g) •e• Cormorant,(B) Brown •elican, and (C) Whitewi•g• S•tet. The feather cat•ories a•e: DOV, distal priEa• outer vane; DIV, distal priEa• inner vane; BK, back;POV, proximal ptima• outervane; •[V, proximalptiEa• inner vane;and B•, brest. •11 the Eicrographsshow the dorsal surfaceof the f•thet •Eples, which are orient• with their distal e•d up•tmost o• the page. 378 A.M.ELOWSON [Auk,Vol. 101

I0 5 0 5 15 5 0 5 I0 15

O. Anhinga b. African Oarter Dbl-cr. Cormorant Reed Cormorant Brown Pelican Am. Wh. Pelican Mgft. Frigatebird Osprey

Pied-b. Grebe Mallard Eur. $helduck Wh.-w. Scorer Ruddy Duck

Anhmga African Darter Dbl.-cr. Cormorant Reed Cormorant Brown Pelican Am Wh. Pelican Mgft. Fr•gateb•rd Osprey Cammon Loon Pied-b. Grebe Mallard Eur. Shelduck

Wh.-w. Scorer Ruddy Duck

Anh*nga Aftman Darter Obl.-cr. Cormorant Reed Cormorant :%) 0 Brown Pehcan I- Am. Wh Pelican z 0 Mgft. Frigatebird Osprey )- Common Loon c• 0 Pied-b Grebe Mallard Eur. Shelduck Wh.-w. Scarer Ruddy Duck

I0 5 0 5 I0 15 5 0 5 I0 t5 Borbules Raml Borbules Roml (r+d)/r (r +d)/r Fig.4. Theindex for rami and barbules (respectively right and left of the heavy lines at (0) based on SEM measurements.The six feather categories are: (a) PIV, (b) POV,(c) DIV, (d) DOV,(e) breast, and (f) back.For furtherexplanation, see text and legend to Fig. 3. April1984] WaterRepellency ofFeathers 379

TABLE3. Probabilities from pair-wise one-tailed TABLE 4. Probabilities from pair-wise one-tailed Mann-Whitney U-testsof (r + d)/r values for the Mann-Whitney U-testsof (r + d)/r values for the rami. barbules.

Feather Pairs of bird groupscompared • Feather Pairs of bird groupscompared a categories Y/N Y/O N/O categories Y/N Y/O N/O Breast NS b NS NS Breast NS b (NS) • NS Back (NS) c NS NS Back NS NS NS Primary a Primary a DIV (NS) NS NS DIV NS NS NS DOV 0.032 NS NS DOV NS NS NS PIV 0.005 0.014 (NS) PIV NS NS (NS) POV 0.005 0.028 NS POV NS NS NS

' Y = speciesthat predictablyspread-wing; O = speciesthat occa- For key see Table 3. sionally spread-wing;N = speciesthat never spread-wing. NS = not significant(P > 0.05). • NS = not significant(P > 0.05). (NS) - not significant, but 0.06 > P > 0.05 (but see text). ß(NS) = Not significant,but 0.06 > P > 0.05. Categoriesas defined in legend to Fig. 3. d DIV, DOV, PIV, POV as defined in legend to Fig. 3.

study have "wettable" primaries and most have more,the datafor t•aedistal primaries show "wettable" body feathers. that the Double-crestedCormorant has a larger My statistical analysis of the (r + d)/r data ramusindex, thereforesupposedly more water- for barbules (Table 4) overwhelmingly rejects repellent feathers, than does the Pied-billed Rijke's hypothesis.Although two comparisons Grebe, when their inner vane data are com- gave equivocal results (see Table 4), these do pared (Fig. 4c); yet the opposite is true of the not even weakly support Rijke's hypothesis, data for their outer vanes(Fig. 4d). Finally, note because in both cases the indices that should that Rijke's distinction between "water repel- have been the smallest(i.e. of speciesthat per- lent" and "wettable" indices is meaningless form spread-wing postures)were in fact the when applied to the data on the breastand back largest and vice versa. The lack of any signifi- feathers. cantdifference in the indicesof barbulesamong The resultsof Mann-Whitney U-testson the the three groupsof speciesis all the more com- data for the rami in Fig. 4 are presentedin Ta- pelling, because the barbules contribute far ble 3. Of 18 comparisons,only five supported more surface area than the rami. Rijke's conclusionby rejecting the null hypoth- Calculatedand observedeffective contact an- esis. These results are not clear-cut, but they gles.--Do the textile equationsaccurately pre- s•uggestthat his hypothesismay be valid to the dict the measuredwater repellency of feathers, extent that the primariesof duckshave a ramus as Rijke implicitly assumed?The observed val- structure that is theoretically more water re- ues for contactangles on the feather samples pellent than thoseof anhingasand cormorants. from a Reed Cormorant and a Mallard (Table It is pu_zzlingthat Rijke reachedthis conclusion 5) provide some insight into this question. I almost entirely on the basis of breast feather solved for the calculatedangles by using the indices,which do not differ significantlyamong indicesgiven in Fig. 4 for the two speciesand the three groups of birds. the meanvalues for 0Rand 0^given in Table Rijke did not measure barbule structure or 5. The observedangles are not differentiated as define its relative water repellency in terms of to rami and barbules,because the droplets cov- (r + d)/r values. It is reasonable to assume, ered both. If Rijke'sassumption is correct,there however, that the same delimiting values he should be good agreement between the ob- found for the rami should also characterize the served and calculated angles. wettable/water repellent dichotomy of the Consider, first, angles calculated with in- (r + d)/r data for barbules. With this perspec- dices of the rami. In the Mallard, the observed tive, note that, with a few exceptionsin the 0'^and 0'Ron the samplefrom the breastfeath- breast and back feather data, the indices of bar- er differed from those predicted by 17ø and 29ø , bulesare all smaller than 4.8. By Rijke'scriteria, and the observed/predicted0'Rs for its primary 4.8 is too small to generate the large effective remex differ by 24ø. These are large differ- contactangles of beaded droplets. Therefore, ences--in the range of that which Rijke saw as basedon barbulestructure, all 14 speciesof this distinguishingbetween specieswith putative 380 A.M. ELOWSON [Auk, Vol. 101

TABLE5. Comparisonof observedand calculatedadvancing and recedingeffective contact angles. •

Calculated0'^ Observedangles Calculated Species Feather Ramus Barbule 0'A 0'R Ramus Barbule Mallard Breast 154ø 130ø 137ø 121ø 150ø 121ø (0^= 88ø, 0R = 75•) PIVb 148ø 117ø 139ø 119ø 143ø 105ø Reed Cormorant Breast 154ø 134ø 133ø 130ø 149• 126ø (0^= 95ø, 0R = 82•) PIVb 137ø 124ø 155ø 146ø 128ø 113ø All observedangles are mean values. PIV = proximal primary inner vane. water-repellentor wettable feathers.The cal- structure and lack of agreement between pre- culated and observed0'^s on the Mallard pri- dicted and observed effective contact angles. mary, differing by 9ø , are perhapsin the range First,the equationsthat Cassieand Baxter (1944) of agreement.The data for the Reed Cormo- derived and Rijke applied assumeforces such rants also showed little agreement;the differ- as gravityor hydrostaticpressure to be negli- ence between the observed and calculated an- gible. As Crisp (1963) correctly pointed out, glesfor both breastand remexfeather samples however,the ability of a poroussurface to re- was 18ø-21 ø. sistwater penetrationunder externalpressure For the anglescalculated with indicesof the is inverselyproportional to the firstpower of barbules,the breastand primary feather sam- scalesize of the pores.This factoris insignifi- ples of both speciesshow conflictingresults. cant at the water's surface, but pressure in- On the one hand the calculated and observed creaseswith depth. At only 10 m below the angles for the breast feathersof both species surface,a diving bird experiencesan external are in good agreement,differing by no more pressureof 1 atmosphererelative to the sur- than 7ø . On the other hand, however, the pri- face.I questionwhether the textile model can mary featherangles differed by 14ø-33ø, similar reliably predicta differencein the potential to the disparitiesfound for the dataof the rami. wettability of the featherson, say, a Common Aside from the contrast between the ob- Loon and Double-crested Cormorant, both at servedangles and those predicted by Cassieand 10 m underwater. Baxter'sequation, the datain Table5 contradict Second,the equationsoverlook an inherent Rijke'shypothesis for two reasons.First, recall difficulty in dealing with recedingcontact an- that Rijke statedthat an 0'Rvalue of 121ø (for gles.These angles cannot be predictedreliably, the African Darter), was too low to causewater becausetheir size is affected by minute accu- dropletsto pearl-off.Yet, the observed0'Rs of mulations of water in surfacepores and irreg- the Mallard are 121ø and 119ø (see Table 5). ularitiesas the water dropletadvances over that These data, indicate that either Rijke's delim- surface(Cassie 1948, Crisp 1963). For this rea- iting angleis incorrector the textilemodel does son,equation (1) is not a statementof thevalue not reliablydescribe water repellency in feath- of 0'•, but of one in a range of possiblevalues. ers. Second,under Rijke's hypothesis,the ob- Moreover, the accumulations of water can in- served 0'• and 0'A of the Reed Cormorant creaseduring prolongedimmersion such that should be smaller than those of the Mallard. the effectivereceding angle may be consider- Yet, comparingthe angleson equivalentsam- ably reduced.This phenomenoncan be accel- plesfrom the two speciesreveals, with oneex- erated by immersion under pressure(Cassie ception, that the Reed Cormorantangles are 1958: 167). Consideringthese effects,the rele- the larger.The exceptionis the breastfeather vance of the textile water-repellencymodel to 0'^, the Mallard angle being larger by 4ø. If birds that dive and swim under water is doubt- cormorants are more wettable than duck•, these ful. datasuggest that the reasondoes not lie in the Third, Rijke explicitly assumedthat the se- shapeof waterdroplets on their feathers. cretionsof the uropygial gland are neither ex- traordinaryin waterproofingcharacteristics nor DISCUSSION sufficientlyvariable among birds to accountfor speciesdifferences in water repellency.Al- Rijke'sapplication of the textilemodel is un- thoughhe later conceded(Rijke 1970:471) that realistic for reasons over and above variable "the function of the gland oil necessarily in- April1984] WaterRepellency ofFeathers 381 volves, besides lubrication, the basis for a finite behaviorwhen wet. More important,indices of contactangle," Rijke remained convinced that the barbules, which constitute more of the sur- the water repellency of birds is primarily de- face area than rami, show no significantdiffer- pendent upon the structural indices of their encesamong these groupsof birds. Moreover, feather elements. Although the studies by even with empiricallymeasured values for the Madsen (1941) and Fabricius (1959) support variables, Cassieand Baxter's (1944) equations Rijke's view (based on their uropygial gland do not consistentlypredict the observedeffec- extirpation experiments), Stettenheim (1972) tive contact angles on feathers. That is not to has stressedthat they overlookedtwo possibil- say,however, that featherstructure is not a rel- ities: that a thin film of secretion remains on evant factor in the water repellency of the the feathers for some time and that the intrin- feathercoat--Rijke's approach is useful in pro- sic feather lipids themselvesare a possiblefac- moting this issue,but his hypothesisdoes not tor in water repellency. Elder (1945) consid- provide the mechanism. ered the gland essentialto the water repellency Finally, Rijke's conclusionthat wing-drying of the feathering, and this seemslikely given is the function of spread-wing posturesin cor- Langmuir's (1919, Zisman 1964) conclusionthat morants and anhingas is based on an a priori a fatty acid monolayer on a surface consider- assumptionthat the behavior servesonly one ably alters its wetting characteristics. function. Most certainly this is an oversimpli- Despite their apparentagreement, there is an fication. important distinction between Madsen's (1941) and Fabricius' (1959) conclusionsand Rijke's ACKNOWLEDGMENTS point of view. The former were both attempt- ing to explain the water repellency of the Financial support came from the University of feather coat, not of individual feathers as was WisconsinGraduate School and the Department of Rijke. It seemsreasonable to me that character- Zoology Davis Fund. I thank the Department of Zo- istics of the entire feather coat would influence ology Museumstaff for providing skinsand the De- how wet a bird gets.The feather coatsof ducks partmentof Entomologyfor useof their SEM facility. differ strikingly from those of cormorantsand I gratefully acknowledgethe skill of Cheryl Hughes and Donald Chandler, who assisted with illustra- anhingas in their greater thickness. Moreover, tions. G. Zografi and his student BarbaraJohnson of there may alsobe differencesamong these birds the University of Wisconsin School of Pharmacy in the density, distribution, and extent of over- kindly let me use their contact angle equipment. I lap of their feathers,as well as whether or not especiallythank R. Schreiber of the Los Angeles there is laminar or turbulent flow around some County Natural History Museum for supplying me areasof the submergedfeather coat.Therefore, with feathers. I thank S. Kolmes, J. R. Baylis, and if cormorantsand anhingasare wettable,it may especiallyJ.P. Hailman for helpful advice. I am be due to the architecture of their feather coats grateful to P. Stettenheim, O. Owre, and an anony- and not so much that of their feathers. mous reviewer, all of whom contributed to an im- proved manuscript.Last, I thank my family for their unfailing support and patience. CONCLUSION

LITERATURE CITED I reject Rijke's hypothesis that the textile model can explain the mechanismof water re- ADVANCESIN CHEMISTRY.1964. Contact angle, wet- pellency in feathersand, as a consequence,re- tability, and adhesion. Series 43. Proc. Kendall solve the question of why spread-wing pos- Award Symposium. Washington, D.C., Amer. tures are performed by cormorants and Chem. Soc. anhingas and not by other water birds. The tex- BAXTER,S., & A. B. D. CASSIE. 1945. The water re- tile equationsare inappropriate for feathers, pellency of fabricsand a new water repellency because their fundamental ideal model is in- test. J. Textile Inst. 36: T67-T90. BENT, A.C. 1922. Life histories of North American adequateto describefeather structure. My re- petrels and pelicans and their allies. U.S. Natl. suitsshow that the textile model fails to predict Mus. Bull. 121. which speciesspread-wing and which ones do BERNSTEIN,N. P., & S. J. MAXSON. 1982. Absence of not. I found few significant differences among wing-spreading behavior in the Atlantic Blue- the (r + d)/r valuesfor the rami of speciesthat eyed Shag (Phalacrocoraxatriceps bransfieldensis). predictably, occasionally,or never show the Auk 99: 588-589. 382 A.M. ELOWSON [Auk,Vol. 101

BIGPLOW, W. C., D. L. PICKETT, & W. A. ZISMAN. 1946ß KENNEDYßR.J. 1969. Sunbathing behaviour of birds. Oleophobic monolayers. I. Films absorbedfrom Brit. Birds 62: 249-258. solution in non-polar liquids. J. Colloid. Sci. 1: ß 1970a. Direct effects of rain on birds: a re- 513-538. view. Brit. Birds 63: 401-414. CASSIE,A. B.D. 1948. Contactangles. Disc. Faraday ß 1970b. Directional water-shedding proper- Soc. 3: 11-16. ties of feathers. Nature 227: 736-737. ß 1958. Contact angles and the absorption of 1971. Preen gland weights. Ibis 113: 369- liquid. Pp. 166-171 in Surface phenomena in 372. chemistry and biology (J. F. Danielli, K. G. A. ß 1972. The probable function of flexules. Ibis Pankhurst, and A. C. Riddiford, Eds.). New York, 114: 265-266. Pergamon Press. LANGMUIR, I. 1919. The mechanism of the surface ß & S. BAXTER. 1944. Weftability of porous phenomena of flotation. Trans. FaradaySoc. 15: surfaces.Trans. Faraday Soc. 40: 546-551. 62-74. CHANDLER,A. C. 1916. A study of the structure of LEwisßH. F. 1929. The natural history of the Dou- feathers, with reference to their taxonomic sig- ble-crested Cormorant (Phalacrocoraxauritus au- nificance. Univ. California Publ. Zool. 13: 243- titus (Lesson)). Ottawa, Ru-Mi-Lou Books. 446. LucAsß A.M., & P. R. STETTENHEIM. 1972. Avian CLARK,G. A., JR. 1969. Spread-wing posturesin Pel- anatomy: integument. Washington D.C., U.S. ecaniformes, Ciconiiformes, and Falconiformes. Dept. Agr. Handbook 362. Auk 86: 136-139. MADSENßH. 1941. Hvad gor Fuglenes Fjer-Dragt CRAMP, W., & K. E. L. SIMMONSß 1977. Handbook of vandskyende?Dansk. Orn. Foren. Tidsskr.35: 49- the birds of Europeßthe Middle East, and North 59. Africa: the birds of the Western Palearctic, Vol. MAHONEY,S.A. 1981. Some aspectsof the thermal I. New York, Oxford Univ. Press. physiology of Anhingas Anhinga anhinga and CRISP,D.J. 1963. Waterproofing mechanisms in an- Double-crested Cormorants Phalacrocorax auritus. imals and plants. Pp. 416-481 in Waterproofing Pp. 461-470 in Proceedingsof the symposiumon and water-repellency (J. L. Moilliet, Ed.). New birds of the sea and shore, 1979 (J. Cooper, Ed.). York, Elsevier. Cape Town, African Seabird Group. Discussion OF THE FARADAY SOCIETY. 1948. Inter- McATEE, W. L., & H. L. STODDARD. 1945. Wettable action of water and porous materials.No. 3. Lon- water birds. Auk 62: 303-304. don, Gurney and Jackson. MCKINNEY, F. 1965. The comfort movements of ELDER,W. H. 1954. The oil gland of birds. Wilson Anatidae. Behaviour 25: 120-220. Bull. 66: 6-31. MOILLIET,J. L. (ED.). 1963. Waterproofingand water- PLOWSON-HALEY,A.M. 1982. Osprey spreadswings repellency. New York, Elsevier. after fishing. Wilson Bull. 94: 366-368. OWRE,O. T. 1967. Adaptations for locomotion and FABRICIUS,E. 1959. What makes plumage water- feeding in the Anhinga and Double-crestedCor- proof? Rept. Wildfowl Trust 10: 105-113. morant. Ornithol. Monogr. 6: 1-138. GEORGE,W. G., & C. L. CASLER.1972. Subalular ap- PALMER,R. S. (ED.). 1962. Handbook of North terium in birds. Auk 89: 245-262. American birdsß vol. 1. New Havenß Connecti- HAILMAN, J.P. 1969a. A review of "The water re- cut, Yale Univ. Press. pellency and feather structure of cormorantsß RHIJN, J. G. VAN. 1977. Processes in feathers caused Phalacrocoracidae."Bird-Banding 40: 65-67. by bathing in water. Ardea 65: 126-147. 1969b. A review of "The water repellency RIJKE,A.M. 1967. The water repellency and feather and feather structure of cormorants, Phalacro- structure of cormorants, Phalacrocoracidae. Os- coracidae."Bird-Banding 40: 346. trich 38: 163-165. HENNEMANN,W. W., III. 1982. Energeticsand spread- 1968. The water repellency and feather winged behavior of Anhingas in Florida. Con- structureof cormorants,Phalacrocoracidae. J. Exp. dor 84: 91-96. Biol. 48: 185-189. HUMPHREY,P.S., & K.. C. PARKES. 1959. An ap- 1970. Weftability and phylogenetic devel- proach to the study of molts and plumages.Auk opment of feather structurein water birds. J. Exp. 76: 1-31. Biol. 52: 469-479. JOHNSON,B.A. 1982. Studies of advancing and re- ROWANßJ. W., & D. GAGLIARDI.1947. Properties of ceding contactangles. Unpublished M.S. thesis. water repellent fabrics.Amer. Dyestuff Rept. 36: Madison, Wisconsin, Univ. Wisconsin-Madison. 533-540. JONES,P.J. 1978. A possiblefunction of the "wing- RUTSCHKE,E. 1960. Untersuchungen tiber Wasser- drying" posture in the Reed Cormorant Phala- festigkeit und Struktur des Gefieders von crocoraxafricanus. Ibis 120:540-542. Schwimmv6geln. Zool. Jahrb. 87: 441-506. KAHL, M.P. 1971. Spread-wing postures and their --. 1976. Gross structure in bird feathers. Proc. possiblefunction in the Ciconiidae.Auk 88: 715- 16th Intern. Ornithol. Congr.:414-425. 722. SCHALLER,G. B. 1964. Breeding behavior of the April 1984] WaterRepellency ofFeathers 383

White Pelican at Yellowstone Lake, Wyoming. Condor 66: 3-23. SCHREIBER,R.W. 1977. Maintenance behavior and communication in the Brown Pelican. Ornithol. Monogr. 22: 1-78. SIEGEL,$. 1956. Nonparametric statisticsfor the be- havioral sciences. New York, McGraw-Hill. SIEGFRIED,W. R., A. J. WILLIAMS, P. G. t-I. FROST,& J. B. K•NAHAN.1975. Plumageand ecologyof cor- morants. Zool. Africana 10: 183-192. $TETTENHEIM,P. 1972. The integument of birds. Pp. 1-63 in Avian biology,vol. 2 (D. S. Farner and J. R. King, Eds.). New York, Academic Press. ß 1976. Structuraladaptations in feathers.Proc. 16th Intern. Ornithol. Congr.: 385-401. THOMPSON,J. E. 1953. The water-repellency of feathers. Australian J. Sci. 16: 65-66. WINKLER,J•. VON. 1983. Das Fliigelspreitverhalten der Mohrenscharbe Phalacrocoraxniger. J. Orni- thol. 124: 177-186. ZISMAN,W.A. 1964. Relation of equilibrium contact angles to liquid and solid constitution. Pp. 1-51 in Contactangle, wettability and adhesion.Ad- vancesin Chem. Ser. 43. Washington D.C., Amer. Chem. Soc.