RESEARCH REPORTS 497

Copyright � 2004, SEPM (Society for Sedimentary Geology) 0883-1351/04/0019-0497/$3.00

OwlPelletTaphonomy:APreliminaryStudy of thePost-RegurgitationTaphonomicHistory ofPellets in a

Temperate Forest

REBECCA C. TERRYDepartment of the Geophysical Sciences, The University of Chicago, 5734 South Ellis Avenue, Chicago, IL 60637,

Email: rcterry@uchicago.edu

PALAIOS, 2004, V. 19, p. 497–506

Owls are important contributors to the Tertiary small-ver-tebrate fossil record. They concentrate small-vertebrate re-mains by producing pellets rich in skeletal material thatprovide a sample of the small-vertebrate fauna of an area. Acommon assumption is that different predators inflictunique fragmentation and skeletal element representationsignatures, thus providing a method for identifying a fieldassemblage as pellet derived, and possibly identifying thepredator. In addition to the digestive process of pellet for-mation, the taphonomic history of a pellet includes the post-regurgitation processes of weathering, disintegration,transport, and burial, all of which can introduce biasesinto an assemblage and confound paleoecological interpre-tation. Analysis of a modern accumulation of small-verte-brate remains from Great Horned Owl (Bubo virginianus)pellets in a temperate forest environment on San Juan Is-land, Washington, reveals that fragmentation and skeletal-element representation change with residence time on theforest floor as pellets disintegrate and skeletal elements be-come dispersed. Matted hair initially protects the skeletalelements. As the pellet breaks down, the bones become dis-persed, fragmentation of the bones increases (from 99% in-tact bones in intact pellets to 75% intact bones in fully dis-persed pellets), and small, fragile skeletal elements are lost,resulting in a residual concentration of larger, more robustskeletal elements. The spatial distribution of skeletal ele-ments below the roosting site follows a right-skewed, bimod-al pattern. Skeletal elements are preserved in the soil to adepth of three centimeters. Post-regurgitation processeshave the potential to distort the original faunal and skeletalcomposition of pellet-derived assemblages, thus maskingany original predator-specific signatures. Actualistic taph-onomic studies are necessary in order to understand howwell pellet-derived assemblages capture information on lo-cal ecological and environmental conditions. This is a crit-ical question that must be addressed to enable correction forsuch biases before pellet-derived assemblages are used forassessment of small-vertebrate community change and pa-leoenvironmental reconstruction.

INTRODUCTION

Predation has long been recognized as an importantmechanism leading to the concentration of small-verte-brate skeletal remains (Mellett, 1974; Dodson and Wexlar,1979; Maas, 1985; Andrews, 1990). Modern predators in-

clude mammalian carnivores, diurnal birds of prey, andnocturnal owls (Mellett, 1974; Andrews, 1990). Althoughmembers of all raptor families produce pellets (regurgitat-ed oblong masses of the undigested components of a bird’sfood, usually consisting of fur, bones, claws, and teeth),owl pellets are characterized by survival of a higher pro-portion of skeletal material than pellets/scat produced byother predators (both avian and mammalian), and thusare thought to provide a more complete sample of the localfauna (Mayhew, 1977; Andrews, 1983; Hoffman 1998; Ly-man, 1994). Owl pellets also can include climate indicatorssuch as pollen (Fernandez-Jalvo et al., 1996; Scott et al.,1996; Fernandez-Jalvo et al., 1999). Because many owlsshow high roost fidelity (Bent, 1961; Andrews, 1990), pel-let-derived accumulations also can provide a geohistoricaltime series for small-vertebrate communities, and thuscontribute information that is critical to our ability to as-sess small-vertebrate community change (on seasonal tomillennial scales), as well as changes in predator behaviorpatterns and local climate change (Avery, 1995; Hadly,1996; Vigne and Valladas, 1996; Fernandez-Jalvo et al.,1998; Grayson, 2000; Hadly and Maurer, 2001; Avery etal., 2002).

Despite their potential as primary resources for recon-structing paleoclimates and assessing ecological and en-vironmental change, pellet-derived small-vertebrate as-semblages have been underutilized, perhaps because ofuncertainty about their taphonomic history (Fernandez-Jalvo et al., 1998). The taphonomic history of a pellet con-sists of multiple phases, each with particular biases. First,selective hunting behavior of predators, time of predatoractivity, season, vulnerability of prey, and local climateconditions will introduce an initial bias into the speciescomposition of any predator-derived accumulation (Craig-head and Craighead, 1956; Andrews, 1990; Denys et al.,1996; Saavedra and Simonetti, 1998), although thesesame factors may permit identification of rare elements inthe fauna (Avery et al., 2002). Second, the digestive pro-cess of pellet formation results in characteristic fragmen-tation and bone loss (Andrews, 1990). Owl pellets areformed in the stomach because a narrow pyloric openingnear the entrance of the esophagus prohibits large parti-cles from entering the intestines; the undigested materialtravels back up the esophagus and is ejected as a pellet(Bent, 1961; Hoffman, 1988; Andrews, 1990). Third, pellethistory following regurgitation (weathering, transport,disintegration, and burial) has the potential to mask orig-inal species composition and other taphonomic signatures,thus biasing the fossil record.

498 TERRY

FIGURE 1—Maps of San Juan Island and Point Caution, WashingtonState, USA. (A) San Juan Island (redrawn from Morgan, 1960). Insetshows location of San Juan Island in the San Juan Archipelago. (B)Point Caution (redrawn from Friday Harbor quadrangle, USGS 7.5minute series, 1954).

Of these three potential sources of bias, only raptor die-tary preferences and patterns of bone destruction due todigestion have received significant attention (e.g., Craig-head and Craighead, 1956; Dodson and Wexlar, 1979;Hoffman, 1988; Andrews, 1990; Denys et al., 1996; Saave-dra and Simonetti, 1998; Trejo and Guthmann, 2003).Some of these studies have attempted to identify uniquefragmentation and skeletal element representation sig-natures left by different predators (e.g., Dodson and Wex-lar, 1979; Hoffman, 1988; Andrews, 1990), whereas othershave investigated the microscopic effects of digestion onbones and teeth (e.g., Rensberger and Krentz, 1988; An-drews, 1990). Comprehensive study of the post-regurgita-tion processes acting on pellet-derived small-vertebrateassemblages is noticeably absent (but see Andrews, 1990,for limestone-cave deposits). This study presents an initialinvestigation into the post-regurgitation taphonomic his-tory of owl pellets in a temperate-forest environment witha focus on: (1) the pattern of skeletal-element distributionthat has developed on the forest floor below a roosting site,(2) changes in the fragmentation and relative abundanceof skeletal elements, and (3) changes in the taphonomiccondition of skeletal elements as pellets disintegrate andskeletal material is dispersed and incorporated into thesoil.

STUDY AREA

The Friday Harbor Laboratory of the University ofWashington is situated on the eastern side of San Juan Is-land, Washington, and includes both coastline and tem-perate forest (Fig. 1). Point Caution, a headland directlynorth of the lab, is dominated by Western Red Cedar (Thu-ja plicata) and Douglas Fir (Pseudotsuga menziesii) foreststhat have not been logged for over 100 years (Staude, pers.comm., 2002; Guberlet, 1975). Approximately 1.6 kmnorthwest of the lab is a 0.01 km2 grove of Western RedCedar [N48�33.615�, W123�0.891�]. In summer 2002, whenthis study was undertaken, the floor of the grove was lit-tered with more than 1,500 small-vertebrate skeletal ele-ments and more than 20 pellets in various stages of disin-tegration. The largest concentration of skeletal elementsoccurred around the base of one tree, but at least four oth-

er trees in the vicinity also showed separate pellet-derivedbone accumulations around their bases.

The high skeletal content of the pellets as well as pelletsize (average size: 5.6 cm x 3.8 cm x 2.4 cm) and the sizeand taxonomic composition of the prey—primarily thenocturnal Townsend’s Vole (Microtus townsendii), with ashrew (Sorex sp.) and a mole (Scapanus sp.) also present(Glass and Thies, 1997)—suggest that the pellets are froma Great Horned Owl (Bubo virginianus) (Bent, 1961; An-drews, 1990). Other large pellet-producing raptors com-mon to San Juan Island include the diurnal Bald Eagle(Haliaeetus leucocephalus), Red-tailed Hawk (Buteo ja-maicensis), Turkey Vulture (Cathartes aura), NorthernHarrier (Circus cyaneus), and the nocturnal Barn Owl(Tyto alba), Western Screech Owl (Otus kennicottii), andNorthern Saw-whet Owl (Aegolius acadicus). Of the noc-turnal owls present on San Juan Island, Great HornedOwls are the largest, most common, are frequently foundin dense forest habitat, and are the only owls to have beenseen and heard around the Friday Harbor Laboratory inthe last several years (Britton-Simmons, pers. comm.2002, Smith et al., 1997).

MATERIALS AND METHODS

The analyses included in this paper are restricted to theroosting tree in the Western Red Cedar grove with thelargest concentration of skeletal elements around its base.The surface distribution of skeletal elements around theroosting tree was mapped using a north/south-orientedgrid, a handheld compass, tape measure, and six 0.25 m2

squares (50 cm x 50 cm) subdivided into 10 cm increments.Skeletal elements rarely were isolated, typically occur-

ring in discrete (or semi-discrete or dispersed but still spa-tially definable) clusters. Spatially defined occurrences ofbones typically contained at least one skull. Dodson andWexlar (1979) reported that the mean number of pelletscast per meal is close to one for Great Horned Owls, thus,each assemblage was reasonably interpreted as a modifiedpellet.

Assemblages were subdivided into three categories: (1)intact pellets, (2) partially dispersed pellets, and (3) fullydispersed pellets (Fig. 2). Pellets were classified based onthe presence and condition of the matted-hair matrix ofthe pellet. Intact pellets were unbroken and characterizedby smooth, matted fur and an oblong shape. In partiallydispersed pellets, the surface of the pellet was no longersmooth and matted, the pellet was broken, and/or skeletalelements had fallen loose from the pellet. In fully dis-persed pellets, spatially definable assemblages of boneswere characterized by the absence of matted fur. All pelletmaterial was mapped, and skeletal elements identifiedwhen possible. Each quarter-meter quadrat on the surfacemap was assigned a unique coordinate and the distancefrom the center of the tree to the center of each quadratwas calculated. Five assemblages of each type were select-ed randomly and collected by hand. Before collection, theorientation of each bone with respect to the soil wasmarked on the exposed surface of the bone for all skeletalelements free from the matted hair of pellets.

Partial and complete pellets were hydrated overnightand picked apart with forceps under a dissecting scope(magnification 2x to 6x). All elements were identified, and

OWL PELLET TAPHONOMY 499

FIGURE 2—Pellet classification scheme. Pellets were classified based on the presence and condition of the matted hair matrix. (A) Intactpellet. (B) Partially dispersed pellet. (C, D) Fully dispersed pellets.

bone fragmentation and modification states recorded.Fragmentation is defined for this study as breakage re-sulting in loss of information about the function or shapeof a skeletal element. The taphonomic condition of all ele-ments from nine representative assemblages (three fromeach assemblage type) was assessed. Bone modification isused in this study to refer to a combination of chemical dis-solution due to digestive modification and secondaryweathering processes, following Behrensmeyer (1978), in-cluding the effects of contact with/burial in soil (chemicalalteration, etching, and pitting) and physical weatheringagents operating on the forest floor (exposure to wind,rain, temperature change). Bone modification categoriesused in this study are defined as follows: (1) unmodifiedbone—the inner cancellous bone is unexposed, protectedby an intact outer layer of dense compact bone; (2) slightlymodified bone—the cancellous bone is exposed; and (3) ex-tensively modified bone—the outer compact layer is ab-sent and the inner cancellous bone is exposed and eroded.

The slope of the forest floor was mapped using a square-meter grid and a Brunton compass. A layer of mobile co-nifer (cedar) detritus covers the soil. The orientations ofskeletal elements from fully dispersed pellets were record-ed to address the importance of hydrodynamic transport.The pH of the soil beneath the conifer detritus, as well as

the phosphorus, nitrogen, and potassium (potash) contentwere determined using a Lammott Soil Kit. Finally, ex-ploratory excavation of six 20-cm2 sites was done using aknife and spoon. Sites were excavated in 1-cm layers to adepth of 5 to 7 cm. Each layer was collected and wet sieved(250 �m mesh).

Unless otherwise noted, statistical analysis consisted ofthe construction of contingency tables analyzed using Chi-square tests. Significance was established at the � � 0.05level.

RESULTS

Spatial Analysis

The surface of the ground surrounding the roost treeslopes �20�–25� to the northeast. The soil beneath the co-nifer detritus is acidic (pH � 5) and contains traceamounts of phosphorus and nitrogen and a high amount ofpotassium (potash). Pellet-derived material (assemblages)is found on all sides of the roost tree. The total number ofassemblages decreases with distance from the roost tree; ahistogram of assemblages versus distance shows a con-cave, right-skewed distribution. This pattern is similarwhen all assemblages are pooled, as well as when assem-

500 TERRY

FIGURE 3—Number of assemblages pooled and subdivided by typewith distance from the tree. (A) Total number of assemblages. (B)Assemblages subdivided by type. The number of assemblages de-creases following a right-skewed, concave trend. Intact pellets peakin abundance approximately 2 m from the tree and do not persist atdistance. Results are not significant due to the small sample size,which violates the assumptions of the Chi-square test (p � 0.2620; n� 52).

FIGURE 4—Number of bones with distance from the tree. The distri-bution of bones follows a bimodal pattern with a peak in abundanceat approximately 2 m. A two-tailed binomial test for proportions indi-cates significance (p K 0.0001; n � 69).

FIGURE 5—Number of robust and fragile skeletal elements with dis-tance from the tree. Robust elements are present in higher numbersthroughout the distribution than fragile elements, and persist fartherfrom the tree. A two-tailed binomial test for proportions indicates sig-nificance (Robust: p K 0.0001; n � 59; Fragile: p K 0.0001; n � 10).

blages are split on the basis of type (Fig. 3). Pellet abun-dance peaks approximately 2 m from the tree, and intactpellets do not persist at distance from the tree, whereaspartially and fully dispersed pellets do persist. The abovepattern is not significant (p � 0.2620; n � 52, the numberof assemblages), perhaps due to the relatively small sam-ple size for the number of quadrats; expected values of20% of the quadrats were fewer than five, a violation of theassumptions of the Chi-square test.

The total number of bones exposed on the surface dis-plays a bimodal distribution with distance from the tree(Fig. 4). Bones appear to be concentrated in peaks between0 and 0.5 m from the tree, and again between 1.5 and 2.5m from the tree. The 5-m measured distribution radiusaround the tree was divided at 2.5 m, creating an inner binthat contained 25% of the total area, and an outer bin thatcontained 75% of the total area. Skeletal elements werecounted for each of the two zones, and a binomial test forproportions that took the disproportionate areas into ac-count was performed to determine significance of the dis-tribution. Despite the fact that the inner zone representsonly 25% of the total area of the site, 59 bones (85.5% of thetotal number of bones) were found in the inner zone. Thispattern is significantly different from what would be ex-pected if elements were distributed randomly between thetwo zones at a 25:75 ratio (p K 0.0001; n � 69). Note thatthese data include only bones visible on the surface. Skel-etal elements in pellets and disarticulating pellets are notincluded because not all pellets were collected and dissect-ed. Figure 3 shows that pellets are found closer to the treethan dispersed assemblages. Including the bones con-

tained in pellets would increase the sample size, andwould reinforce the pattern described.

Long bones were separated into two categories: (1) ro-bust elements (humerus, pelvis, femur), and (2) fragile el-ements (ulna, radius, tibia). The distribution patterns ofrobust and fragile elements were analyzed separately us-ing the binomial statistical method described above. BothFigure 5 and the disparate sample sizes (see below) indi-cate that robust elements are present in higher numbersthan fragile elements throughout the entire distribution,and persist farther from the tree. The total numbers of ro-bust and fragile skeletal elements in each area were sig-nificantly different at � � 0.05 (Robust: p K 0.0001; n �59; Fragile: p � 0.0001; n � 10). There is no evidence (suchas preferred orientations of bones or runneling of the soilsurface and/or mobile conifer detritus) that would indicatehydrodynamic transport at the site.

Taphonomic Analysis

Relative proportions of all skeletal elements from 15 as-semblages were calculated for each assemblage type (in-tact pellets, partially dispersed pellets, fully dispersed pel-lets; Fig. 6). The relative proportions of different skeletalelements differ significantly between assemblage types (PK 0.0001, n �672). Fully dispersed pellets are dominatedby mandibles (18%), followed closely by vertebrae (15%),

OWL PELLET TAPHONOMY 501

FIGURE 6—Relative proportions of skeletal elements with assem-blage type; proportions differ significantly among assemblage types(p K 0.0001; n � 672); Sk � skull; Ma � mandible; Sc � scapula;V � vertebra; H � humerus; U � ulna; Ra � radius; P � pelvis; F �femur; T � tibia/fibula; Ri � rib; CTP � manus and pes elements. (A)Intact pellets. (B) Partially dispersed pellets. Note (A) and (B) are dom-inated by vertebrae, ribs, and manus and pes elements. (C) Fully dis-persed pellets, which are dominated by more robust skeletal elements.

FIGURE 7—Fragmentation frequency with assemblage type. Fullydispersed pellets contain the highest proportion of fragmented ele-ments. Results are significant (p K 0.0001; n � 351).

skulls (11%), humeri (11%), femora (11%), tibiae (11%),and ribs (11%). Pes and manus elements are not present.Partially dispersed pellets and intact pellets show a muchgreater apparent similarity than either assemblage typeshows to fully dispersed pellets. Partially dispersed andintact pellets have a much lower proportion of robust longbones and mandibles, instead consisting primarily of ele-ments such as manus and pes (25–44%), ribs (24%), andvertebrae (20–24%).

Fragmentation was calculated for nine assemblagesand pooled by assemblage type (Fig. 7). Intact pellets con-tained the highest proportion of whole elements (99%) andthe lowest proportion of fragmented elements (1.0%). Atthe other extreme, fragmentation frequency was highestin fully dispersed pellets (26%). Fragmentation frequencyamong the different assemblage types was found to be sig-nificantly different (p K 0.0001; n � 351).

Different types of skeletal elements display uniquebreakage patterns. Skulls, for example, exhibit a charac-teristic breakage pattern of the cranium in which the pos-terior half of the skull is completely absent (Fig. 8A, B).Skulls that exhibit this type of breakage are found mostfrequently in fully dispersed pellets (23%) and partiallydispersed pellets (30%; Fig. 8C). All skulls found in intactpellets were unbroken. While results were found to be sig-nificant (p � 0.0151; n � 13), the small sample size vio-lates assumptions of the test, making statistical analysisinconclusive but consistent with the observed pattern.

The epiphyseal ends of long bones, which representgrowth centers that become fused during the adult life ofan animal (Swindler, 1998), can provide clues about theage structure of a prey assemblage. Approximately 63% ofall long bones (humerus, femur, tibia) from 15 assemblag-es are missing epiphyses (Fig. 9). Significance of this pat-tern was established using an unpaired t-test (p � 0.0098;n � 46). This suggests a potential age bias towards youn-

502 TERRY

FIGURE 8—Fragmentation patterns of skulls; scale bar in millimeters.(A) Dorsal view of three skulls (Microtus townsendii). Two exhibit acommon posterior breakage pattern. (B) Ventral view of the sameskulls. (C) Frequency of posterior skull breakage. All skulls recoveredfrom intact pellets were unbroken. Results are significant (p � 0.0151;n � 13), however, small sample size violates the assumptions of theChi-square test.

FIGURE 9—Epiphyseal presence and absence in humeri and femoraof Microtus townsendi; scale bar in millimeters. (A) Humerus with prox-imal epiphysis missing. (B) Humerus with proximal epiphysis present.(C) Femur with distal epiphysis missing. (D) Femur with distal epiph-ysis present.

ger individuals in the prey assemblage. No significant dif-ference was found for the frequency of epiphysis loss be-tween the different types of skeletal elements (p �0.0758).

Bone-modification results pooled for all elements by as-semblage type are presented in Figure 10. The differencesare significant (p � 0.0116; n � 191) and suggest that forfully dispersed pellets, more than 60% of the bones exhibitno bone modification, whereas for intact pellets, onlyabout 45% of the bones show no bone modification. Con-versely, about 25% of skeletal elements from fully dis-persed pellets and almost 20% of skeletal elements fromintact pellets show evidence of extensive bone modifica-tion.

Bone modification results for skeletal elements found infully dispersed pellets were compared with results for thesame skeletal elements from intact pellets. Results werenot significant (p � 0.2022; n � 101), indicating no differ-ence in terms of bone modification between elements pre-sent in both assemblage types.

Different skeletal elements show different degrees ofmodification, as do the shaft and ends of a single longbone. For all long bones (with and without epiphyses) forall assemblage types pooled together, the ends of longbones show a significantly higher degree of bone modifi-cation than the shafts (p K 0.0001; n � 43; Fig. 11).

Skeletal elements from fully dispersed pellets are scat-tered on the forest floor, with one side of the bone exposedto air, and the other side exposed to twig litter or soil. Bonemodification for each surface of all bones derived from ful-ly dispersed pellets is shown in Figure 12. Seventy percentof surfaces exposed to the air showed little to no bone mod-ification, whereas only 50% of surfaces exposed to the soilshowed little to no bone modification. However, almost20% of surfaces exposed to the soil show extensive bonemodification, while only 5% of surfaces exposed to the airshow extensive bone modification (p � 0.0449; n � 241).

Excavation ResultsPreliminary excavation of three 20-cm squares to

depths of 5 to 7 cm yielded skeletal elements incorporated

OWL PELLET TAPHONOMY 503

FIGURE 10—Bone modification (A) Examples of different degrees ofbone modification in pelves of Microtus townsendii; N � no bone mod-ification; M � extensive bone modification (cancellous bone is ex-posed on the shaft and eroded on the distal ilium); scale in millimeters.(B) Bone modification by assemblage type. Bones from fully dispersedpellets show the highest frequency of no modification. Results aresignificant (p � 0.0116; n � 191).

FIGURE 11—Bone modification compared for the ends and shafts oflongbones. The ends of longbones show a significantly higher fre-quency of modification than do the shafts (p K 0.0001; n � 43).

FIGURE 12—Bone modification of elements from fully dispersed pel-lets, compared for surfaces exposed to air and soil. Surfaces in con-tact with the soil show a significantly higher frequency of extensivebone modification (p � 0.0449; n � 241).

into the soil to a depth of 2 cm below the soil surface (2skulls, 4 mandibles, 1 pelvis, 1 femur, 1 tibia, 1 vertebra;Fig. 13). All skeletal elements recovered from the soil showextensive bone modification.

DISCUSSION

Skeletal material can be concentrated by both physical(fluvial sorting, wind deflation) and biological (predation)processes. These different concentrating mechanismshave the potential to introduce different biases into an as-semblage, which should then affect paleoecological inter-pretation (Behrensmeyer, 1993). Thus, before a pellet-de-rived small-vertebrate assemblage can be used for paleo-ecological reconstruction, there must be a reliable way torecognize that it is indeed pellet-derived. Previous studies,primarily conducted in laboratory settings, have focusedon fragmentation patterns as the means of identifying pel-let-derived assemblages (Dodson and Wexlar, 1979; Hoff-man, 1988; Saavedra and Simonetti, 1998; Andrews, 1990;Kusmer, 1990). Relative abundance of skeletal materialand characteristic breakage patterns produced in labora-tory settings also have been proposed as methods to facil-itate predator identification from field assemblages (Dod-

son and Wexlar, 1979; Hoffman, 1988; Andrews, 1990). Inhis paper on pellet-derived fragmentation patterns, Hoff-man (1988) briefly acknowledged that post-pellet diage-netic processes might confound the taphonomic signaturesof small-vertebrate assemblages, making definitive pred-ator identification difficult. However, the number of actu-alistic studies of the post-regurgitation history of pellets islimited (but see Andrews, 1990). The results from thisstudy illustrate the importance of understanding howpost-regurgitation processes can bias small-vertebrate as-semblages over time in a temperate-forest environment.

Once a pellet has been regurgitated and is exposed tophysical and chemical weathering processes on the forest

504 TERRY

FIGURE 13—Skeletal elements (Microtus townsendii) recovered from the subsurface; scale in millimeters. (A) Skull from 1 cm below the soilsurface. (B) Mandible from 2 cm below the soil surface. (C) Mandible and pelvis from 1 cm below the soil surface. (D) Mandibles from the topof the soil surface below the leaf litter.

floor, the matted hair in the pellet protects the skeletal el-ements contained inside (Andrews, 1990). Assuming thedifferent pellet types designated in this study (intact pel-lets, partially dispersed pellets, and fully dispersed pel-lets) represent steps in the disintegration process overtime, both the relative proportion of skeletal elements pre-sent in an assemblage and the degree of fragmentationchanged as a result of time exposed on the forest floor. Asthe pellets broke down, bones became dispersed, and frag-mentation increased (Fig. 7). Evidence for transport anddispersal by hydrodynamic processes is lacking, thus, theincreased dispersal and fragmentation of skeletal materi-al must be due to other factors, such as scavenging.

Dodson and Wexlar (1979) observed a high frequency ofposterior breakage in skulls recovered from Great HornedOwl pellets in a laboratory setting and presented this as apossible predator identification tool. However, they alsoreported feeding observations for Great Horned Owls, inwhich the owls swallow mice whole and headfirst, makingno attempt to extract the brain of the animal. In compari-son, the skulls recovered from intact pellets in this studyprimarily were unbroken, consistent with the feeding ob-

servations of Dodson and Wexlar (1979), while brokenskulls were found in partially dispersed pellets and fullydispersed pellets on the forest floor (Fig. 8). This suggeststhat skulls become fragmented more frequently duringpost-regurgitation processes such as scavenging; and thatcaution should be used if employing skull breakage as atool for predator identification.

Based on this study, as pellets disintegrate, the skeletalcomposition of the assemblage shifts from a high propor-tion of small, fragile elements to a high proportion of larg-er, more robust elements, because the fragile elements arepreferentially lost (Figs. 5, 6). Skeletal elements from fullydispersed pellets (which potentially have spent the mosttime in contact with the forest floor) show the highest fre-quency of no bone modification. Elements from intact pel-lets, however, show lower proportions of no modificationthan bones from dispersed pellets and high frequencies ofboth slight and extensive bone modification (Fig. 10). Apossible explanation is that the matted hair in a pellet ini-tially shields the smaller and fragile bone fractions (whichare perhaps more intensely modified by digestion). Uponpellet disintegration, these smaller, fragile, extensively

OWL PELLET TAPHONOMY 505

modified elements then become exposed to physical- andchemical-weathering processes on the forest floor. Ar-mour-Chelu and Andrews (1994) documented significantburial and lateral transport of small skeletal elements dueto earthworm bioturbation. This, coupled with potentialeffects of preferential destruction due to higher modifica-tion, would result in a residual concentration of the larger,robust, skeletal elements that were less extensively modi-fied by digestion. Of the bones that do remain on the forestfloor, it should be noted that the surface in contact withthe soil shows a higher degree of bone modification thanthe surface that is exposed to the air (Fig. 12). This pref-erential damage is consistent with a variety of actualisticstudies of weathering of larger bones (e.g., Behrensmeyerand Hill, 1980).

The distinctive characteristics of fragmentation andskeletal element representation thus change over time.This presents a problem if these characteristics are to beused to determine whether an assemblage is predator-de-rived, and to identify the predator. Alternate methods foridentifying an assemblage as pellet-derived deserve moreattention. As expected, in this study, the density of skele-tal elements decreases with distance from the tree. Theright-skewed, bimodal distribution of skeletal elementsseen below the roost site, with a distribution radius of in-tact pellets approximately 2 m from the tree (Figs. 3, 4), ismost likely the result of a reflecting boundary (the treetrunk) and preferred roosting position of the owl. As pel-lets are regurgitated and tumble through the densebranches to the ground below, scattering off the branchesand the trunk would yield a peak in abundance at a mod-erate distance from the trunk. If surface patterns also arepreserved at depth (especially in environments more con-ducive to preservation than acidic, moist, coastal-forestsoils), they could be powerful indicators that an assem-blage is pellet-derived. Alternative techniques for identi-fying whether an assemblage is predator-derived that de-serve further study include microscopic investigation ofchemical etching, pitting, and dissolution that occurs withdigestion (and taxonomic differences in preservation po-tential). Microscopic analyses also are needed to distin-guish the effects of digestion from secondary chemical dis-solution that occurs as bones come in contact with the soil.Finally, actualistic and microscopic studies also are need-ed to better understand the confounding effects of scav-enging on dispersal, skeletal-element representation, andfragmentation patterns. These analyses are beyond thescope of this work.

CONCLUSIONS

(1) Pellets and skeletal elements become dispersed withdistance from the tree. The pattern of dispersal of skeletalelements follows a right-skewed bimodal distribution.

(2) The relative proportions of skeletal elements changeas pellets disintegrate, thus this is not a reliable criterionfor identifying the pellet origin of and predator responsiblefor (sub)fossil assemblages. Intact pellets are dominatedby small fragile elements, which are preferentially lostwith pellet disintegration over time.

(3) Frequency of fragmentation increases as pellets be-come dispersed, and individual breakage patterns (such asloss of the cranium in skulls) can be incurred due to post-

regurgitation processes such as scavenging. Thus, frag-mentation is not a reliable indicator for predator identifi-cation.

(4) The frequency of intense bone modification is highestin pellets and lowest in dispersed assemblages. This likelyreflects preferential destruction of smaller, more fragileskeletal elements, which may be modified more intenselydue to digestive processes, upon being released from thepellet.

(5) The taphonomic history of pellet-derived small-ver-tebrate assemblages is more complex than commonly ac-knowledged. Since post-regurgitation processes distortthe original skeletal composition of pellet-derived assem-blages, actualistic studies are necessary in order to under-stand and correct for this bias, leading to more accurateassessments of small-vertebrate community change andpaleoenvironmental reconstruction.

ACKNOWLEDGEMENTS

This work was completed as part of the 2002 Marine Ta-phonomy course at the Friday Harbor Laboratory of theUniversity of Washington. I would like to give specialthanks to Michael LaBarbera and Michal Kowalewski fortheir enthusiasm, patience, and invaluable assistance andtheir willingness to let me work on terrestrial vertebratesin the context of a marine invertebrate taphonomy course.I would also like to thank David Eberth, Karen Chin, andPeter Andrews for their helpful reviews of this manu-script, as well as Susan Kidwell, Matthew Kosnik, andMark Terry for their initial comments on the paper. Final-ly, much thanks is also due to Katie LaBarbera for herfield assistance.

REFERENCES

ANDREWS, P., 1990, Owls, Caves and Fossils: University of ChicagoPress, Chicago, 231 p.

ARMOUR-CHELU, M., and ANDREWS, P., 1994, Some effects of biotur-bation by earthworms (Oligochaeta) on archaeological sites: Jour-nal of Archaeological Science, v. 21, p. 433–443.

AVERY, D.M., 1995, Micromammalian studies: information from thepast of relevance to the future: Transactions of the Royal Society ofSouth Africa, v. 50, p. 41–47.

AVERY, D.M., AVERY, G. and ROBERTS, A., 2002, A contribution frombarn owl pellets to known micromammalian distributions inKwaZulu-Natal, South Africa: African Zoology, v. 37, p. 131–140.

BEHRENSMEYER, A.K., 1978, Taphonomic and ecologic information onbone weathering: Paleobiology, v. 4, p. 150–162.

BEHRENSMEYER, A.K., 1993, The bones of Amboseli: bone assemblag-es and ecological change in a modern African ecosystem: NationalGeographic Research, v. 9, p. 402–421.

BEHRENSMEYER, A.K., and HILL, A.P., eds. 1980, Fossils in the Mak-ing: Vertebrate Taphonomy and Paleoecology: University of Chi-cago Press, Chicago, 338 p.

BENT, A.C., 1961, Life Histories of North American Birds of Prey,Part Two: Dover Publications, Inc., New York, 482 p.

CRAIGHEAD, J.J., and CRAIGHEAD, F.C., JR., 1956, Hawks, Owls andWildlife: Wildlife Management Institute, Washington, D.C., 443p.

DENYS, C., DAUPHIN, Y., RZEBIK-KOWALSKA, B., and KOWALSKI, K.1996. Taphonomic study of Algerian owl pellet assemblages anddifferential preservation of some rodents: palaeontological impli-cations: Acta Zoologica Cracoviensia, v. 39, p. 103–116.

DODSON, P., and WEXLAR, D., 1979, Taphonomic investigations of owlpellets: Paleobiology, v. 5, p. 275–284.

FERNANDEZ-JALVO, Y., DENYS, C., ANDREWS, P., WILLIAMS, T., DAU-PHIN, Y., and HUMPHREY, L., 1998, Taphonomy and palaeoecology

506 TERRY

of Olduvai Bed-I (Pleistocene, Tanzania): Journal of Human Evo-lution, v. 34, p. 137–172.

FERNANDEZ-JALVO, Y., SCOTT, L., and DENYS, C., 1996, Pollen com-position in owl pellets and their environmental implications:Comptes Rendus de L’academie des Sciences, Paris, v. 323, p. 259–265.

FERNANDEZ-JALVO, Y., SCOTT, L., and DENYS, C., 1999, Taphonomyofpollen associated with predation: Palaeogeography, Palaeoclima-tology, Palaeoecology, v. 149, p. 271–282.

GLASS, B.P., and THIES, M.L., 1997, A Key to the Skulls of NorthAmerican Mammals, Third Edition: Self-published, Monte L.Thies, Sam Houston State University, Department of BiologicalSciences, Huntsville, TX, unpaged.

GRAYSON, D.K., 2000, Mammalian responses to middle Holocene cli-matic change in the Great Basin of the western United States:Journal Of Biogeography, v. 27, p. 181–192.

GUBERLET, M.L., 1975, The Windows to His World: The Story of Tre-vor Kincaid: Pacific Books, Publishers, Palo Alto, 287 p.

HADLY, E.A., 1996, Influence of late Holocene climate on northernRocky Mountain mammals: Quaternary Research, v. 46, p. 298–310.

HADLY, E.A., and MAURER, B.A., 2001, Spatial and temporal patternsof species diversity in montane mammal communities of westernNorth America: Evolutionary Ecology Research, v. 3, p. 477–486.

HOFFMAN, R., 1988, The contribution of raptorial birds to patterningin small mammal assemblages: Paleobiology, v. 14, p. 81–90.

KUSMER, K.D., 1990, Taphonomy of owl pellet deposition: Journal ofPaleontology, v. 64, p. 629–637.

LYMAN, R.L., 1994, Relative abundances of skeletal specimens andtaphonomic analysis of vertebrate remains: PALAIOS, v. 9, p.288–298.

MAAS, M.C., 1985, Taphonomy of a late Eocene microvertebrate local-

ity, Wind River Basin, Wyoming (U.S.A.): Palaeogeography, Pa-laeoclimatology, Palaeoecology, v. 52, p. 123–142.

MAYHEW, D.F., 1977, Avian predators as accumulators of fossil mam-mal material: Boreas, v. 6, p. 25–31.

MELLETT, J.S., 1974, Scatological origin of microvertebrate fossil ac-cumulations: Science, v. 185, p. 349–350.

MORGAN, C.T., 1960, The San Juan Story: San Juan Industries, Fri-day Harbor, 28 p.

RENSBERGER, J.M., and KRENTZ, H.B., 1988, Microscopic effects ofpredator digestion on the surfaces of bones and teeth: ScanningMicroscopy, v. 2, p. 1541–1551.

SAAVEDRA, B., and SIMONETTI, J.A., 1998, Small mammal taphono-my: intraspecific bone assemblage comparison between South andNorth American Barn Owl, Tyto alba, populations: Journal of Ar-chaeological Science, v. 25, p. 165–170.

SCOTT, L., FERNANDEZ-JALVO, Y., and DENYS, C., 1996, Owl pellets,pollen, and the palaeoenvironment: South African Journal of Sci-ence, v. 92, p. 223–224.

SMITH, M.R., MATTOCKS, P.W., JR., and CASSIDY, K.M., 1997, Breed-ing birds of Washington State, volume 4: in Cassidy, K.M., Grue,C.E., Smith, M.R., and Dvornich, K.M., eds., Washington GapAnalysis—Final Report: Seattle Audubon Society Publications inZoology No. 1, Seattle, 538 p.

SWINDLER, D.R., 1998, Introduction to the Primates: University ofWashington Press, Seattle, 284 p.

TREJO, A., and GUTHMANN, N., 2003, Owl selection on size and sexclasses of rodents, activity and microhabitat use of prey: Journalof Mammology, v. 84, p. 652–658.

VIGNE, J.D., and VALLADAS, H., 1996, Small mammal fossil assem-blages as indicators of environmental change in northern Corsicaduring the last 2500 years: Journal of Archaeological Science, v.23, p. 119–215.

ACCEPTED FEBRUARY 20, 2004