Sensitivity of Taphonomic Signatures to Sample Size, Sieve ...geosci.uchicago.edu/pdfs/kidwell/Kidwell-2001-Sensitivity.pdf · Sensitivity of Taphonomic Signatures to Sample Size,

26

Copyright � 2001, SEPM (Society for Sedimentary Geology) 0883-1351/01/0016-0026/$3.00

Sensitivity of Taphonomic Signatures to Sample Size,Sieve Size, Damage Scoring System, and Target Taxa

SUSAN M. KIDWELL, THOMAS A. ROTHFUS, and MAIRI M.R. BESTDepartment of Geophysical Sciences, University of Chicago, 5734 South Ellis Avenue, Chicago, IL 60637

PALAIOS, 2001, V. 16, p. 26–52

The sensitivity of taphonomic signatures to a battery ofcommon sampling and analytic procedures is tested hereusing modern bivalve death assemblages from the San BlasArchipelago, Caribbean Panama, to determine (a) the mag-nitude of methodological artifacts and, thus, the compara-bility of taphofacies patterns among studies; and (b) themost efficient and robust means for acquiring damage pro-files (taphonomic signatures) of death assemblages both an-cient and modern. Damage frequency distributions do notstabilize below sample sizes of 120–150 individuals. Usingdamage to the �8 mm portion of the assemblage as a base-line (interior damage only, fragments included), it is foundthat qualitative trends among environments (higher dam-age levels in reefal skeletal gravel versus mud) and therank-order importance of taphonomic variables per envi-ronment (intensity of damage from encrustation, boring,fine-scale alteration, edge-rounding, fragmentation) are ro-bust to most methodological decisions. The exception is theuse of target taxa: of three genera tested, only one was sen-sitive to the same suite of environmental differences as thetotal-assemblage, and taxa had disparate rank-ordering ofvariables. In contrast to the general robustness of qualita-tive trends, quantitative damage levels are affected signifi-cantly by methodology. Specifically, the measured frequencyof damage is generally lower for finer size fractions and fin-er sieve sizes, for whole shells versus fragments, for taxo-nomically well-resolved specimens, for infaunal versus epi-faunal species regardless of mineralogy, and for interiorsurfaces versus exterior or total surface area of shells. Fullfrequency-distribution data on states of taphonomic dam-age are most powerful for differentiating samples, but if sin-gle-value metrics are desired, the frequency of high-intensi-ty damage is more powerful—and shows less between-oper-ator variance—than presence-absence data or average dam-age state. To maximize the detection of damage and ofbetween-environment differences in taphonomic signature,and to foster between-study comparisons, the following arerecommended: (1) analysis of discrete size-fractions ratherthan broad spectra and, in particular, the separate treat-ment of coarse size fractions (�4 mm); (2) examination ofcomplete assemblages (fragments as well as whole speci-mens; all species or broad subsets of species rather than se-lect taxa); (3) variables scored independently (e.g., encrus-tation v. boring) rather than grouped into summary grades;and (4) evaluation of rank-ordering of variables in plots ofthreshold damage profiles as a complement to ternary taph-ograms.

INTRODUCTION

For the last fifteen years, taphonomists have focused in-creased effort on quantifying variation in post-mortemdamage among environments. The primary aim of suchtaphofacies analysis (sensu Speyer and Brett, 1986) hasbeen to improve the quality of paleoenvironmental inter-pretations by considering the state of preservation of or-ganic remains, including patterns of damage to skeletalhardparts and styles of skeletal concentration. An out-growth and complement to sedimentary microfacies anal-ysis (e.g., Pilkey, 1964; Pilkey et al., 1969, 1979; and seeFlugel, 1982), this information also can help to recognizeexotic material (damage patterns distinct from indigenousmaterial; e.g., Davies et al., 1989; Miller et al., 1992), torank the importance of taphonomic variables (by relativeintensity or frequency of damage in assemblage; e.g., ref-erences in Table 1, and see below), and, inferentially, to es-timate the facies-by-facies reliability of paleobiologic infor-mation, under the assumption (following Johnson, 1960)that assemblages dominated by well-preserved skeletonsare less time-averaged or otherwise modified than thosewith high post-mortem modification (and see review byParsons and Brett, 1991).

‘‘Taphonomic signatures’’ (Davies et al., 1989) now havebeen documented for a variety of groups in both modernand ancient systems (for actualistic examples, see Nebel-sick, 1999, for echinoids; Llewellyn and Messing, 1993, forcrinoids; Pandolfi and Greenstein, 1997, and Perry, 2000,for corals; Smith and Nelson, 1994, for bryozoans; Martinand Liddell, 1991, for foraminifera; Gastaldo et al., 1987,for macroflora). Marine mollusks have received the mostattention, but no consensus has emerged on taphofaciesmethodology. For example, of 20 actualistic molluscan ta-phofacies analyses (Table 1), 12 different sieve sizes havebeen used, and there has been a similarly high degree ofvariation in the taphonomic variables scored (e.g., frag-mentation, rounding, boring), in how damage states arequantified (e.g., number of ‘‘grades’’ per variable andgrade definition), in the portion of the death assemblageexamined, and in methods of data analysis. This diversityis natural given the youth of the quantitative taphofaciesapproach, but makes comparison and synthesis of resultsextremely difficult and, thus, reduces the collective valueof this effort.

Here, as a guide for future studies, the sensitivity ofmolluscan taphonomic signatures is evaluated relative tomethodological decisions during sampling, data-collection,and data-analysis. By applying a battery of methods to asingle set of samples, in this instance bivalve death assem-blages from subtidal mud, sand, and shell gravel across asmall fringing reef in the San Blas Archipelago of Carib-

SENSITIVITY OF TAPHONOMIC SIGNATURES 27

TABLE 1—Summary of methods used in actualistic molluscan taphofacies studies. * each taphonomic variable scored macroscopically (10�)and analyzed independently, unless otherwise noted; damage scored as presence-absence (‘‘�’’) or according to the degree of damage tothe shell (‘‘damage state’’). ‘‘Summary grades’’ are damage categories for individual shells based on multiple rather than single variables.

Study

Sieve sizemm (sizefractionstested)

Inclusion offragments &

specimens notidentifiable tospecies-level

Set(s) oftaxa

analyzed

Surfacesexamined &

comparedTaphonomic

variables

Form ofdata for

individualshells*

Descriptionof assemblage,& analysis of

differencesamong

assemblages

US Atlantic conti-nental shelf (Pilk-ey et al., 1969)

2(2–4)

Included frag-ments, bothidentifiable &unidentifiable

Total shellmaterial insize frac-tion, domi-nantly mol-luscan

Total only Fragmentation,rounding, dis-coloration

� fragmentation;damage state forother variables

Maps of %damage

Puerto Rico conti-nental shelf (Pilk-ey et al., 1979)

1(1–2)


Total shellmaterial insize frac-tion, domi-nantly mol-luscan

Total only Rounding, luster Damage state; also3 summary grades(old dull, old pol-ished, fresh origi-nal)

Histograms of% damage

Georgia shelf pa-limpsest sand(Frey & Howard,1986)

2.5 Only taxonomical-ly identifiablefragments in-cluded

All mollusks Total only Encrustation, bor-ing; retention oforiginal color,gloss; presenceof ligament orperiostracum

2 summary grades(‘‘old-new’’); en-crustation & bor-ing treated sepa-rately

Line-graphs ofnew:oldvalve ratios,and % �damage

Georgia longshoretidal channel andnearshore shelf(Henderson &Frey, 1986)

1.5 Included frag-ments

All mollusks Not speci-fied

Size-sorting, %articulated,right-left ratiosof bivalves; re-tention of origi-nal color, gloss;presence of lig-ament or per-iostracum

2 summary grades(‘‘old-new’’); �data for articula-tion

Line-graphs ofnew:old,right:left,and % artic-ulated spec-imens

Sonora tidal flat(Fursich & Flessa,1987

3 Damage to whole,identifiableshells only

6 target gen-era (3 in-faunal bi-valves, 2epifaunalgastropods,1 infaunalechinoid)

Interioronly

Abrasion, luster,color, fragmen-tation dissolu-tion, boring, en-crustation

Sum of ordinal dam-age states

Bivariateplots andhistogramsof summedscores andof full-fre-quency dataon states

Texas tidal inlet(Davies et al.,1989)

2(2–4, 4–8, 8–

16, �16)

Included frag-ments, bothidentifiable &unidentifiable(tested effects)

All mollusks Total; inte-rior v. ex-terior; 8morpholog-ic subare-as

Dissolution, frag-mentation,abrasion includeedges, ‘‘biotic’’(includes preda-tory drillholes),size-sorting

Damage state Tables of %-state dataand MAN-COVA re-sults

Texas continentalshelf above stormwavebase (Staff &Powell, 1990)

4(4–12, �12)


All mollusks;2 infaunaltarget spp

Total; inte-rior v. ex-terior; 8morpholog-ic subare-as

Disarticulation,fragmentation,dissolution,abrasion in-clude edges,‘‘biotic’’ (in-cludes predato-ry drillholes),right:left &size-sorting

Damage state Tables of %-state dataand MAN-COVA re-sults

Massachusetts tidalflat (Meldahl &Flessa, 1990)

6 Damage to wholespecimens only

single infau-nal bivalve

Total only Fragmentation,edge rounding,exterior abra-sion, alterationof muscle scars,boring, encrus-tation

� damage Histogramsand clusteranalysis of%-� data;multi-di-mensionalscaling toidentifypathways ofdamage ac-crual

28 KIDWELL ET AL.

TABLE 1—Continued.

Study




Set(s) oftaxa

analyzed

Surfacesexamined &

comparedTaphonomic

variables

Form ofdata for

individualshells*


differencesamong

assemblages

Sonora tidal flat(Feige & Furisch,1991)

3 Damage to whole,identifiableshells only

All mollusks;plus sepa-rate sam-ples of 6target mol-lusk genera(4 infaunal,2 epifau-nal)

Interioronly

Loss of ornament(abrasion �bioerosion), lossof color, encrus-tation, boring,fragmentation,articulation,right:left sort-ing, dissolution,maceration

Damage state Bivariateplots of av-erage state(alterationindex) �data,stacked bar-graphs offull-fre-quency data

Texas continentalslope cold seep(Callender & Pow-ell, 1992)

1 Included frag-ments, bothidentifiable &unidentifiable(tested effects)

All mollusks;target spe-cies fromeach biofa-cies

Total; inte-rior v. ex-terior; 8morpho-logic sub-areas

Fragmentation,articulation,right:left &size-sorting,presence of per-iostracum, ‘‘dis-solution’’, pre-cipitation, edgecondition, abra-sion, ‘‘biotic in-teractions’’

Damage state Tables of %-state data,Chi square& nestedMANOVAresults

Texas cold seep,nearshore shelf, &tidal inlet (Callen-der et al., 1992)

4(4–9, 9–14,

14–19, 19–24, �24)


All mollusks Total; inte-rior v. ex-terior; 8morpho-logic sub-areas

Dissolution, abra-sion, edgerounding, frag-mentation, biot-ic interac-tions, artic-ulation, right:left ratios, size-sorting

Damage state Tables of %-state dataChi squareresults

Virgin Islands car-bonate shelf (Par-sons, 1989, 1993)

5 Included frag-ments

All mollusks Total only Disarticulation,fragmentation,encrustation,bioerosion, rootetching, edgerounding, abra-sion, color loss

Damage state Bar-graphs offrequencydata, dis-criminantanalysis ofaveragestates

Baja Californiacheniers (Kowa-lewski et al.,1994)

12.5 Included frag-ments

Single infau-nal bivalvespecies

Total exceptinterior v.exteriorluster

Bioerosion, en-crustation,fragmentation,cracking, exte-rior peeling,edge condition,luster, state ofinterior fea-tures; size-sort-ing, articula-tion, right:left

Damage state Ternary taph-ograms offull-fre-quencydata, multi-variateanalysis ofaveragestates(PCA,CVA); clus-ter analysis& binomialtests ofpooled sam-ples

Germany, Baja &Sonora tidal flats(Kowalewski etal., 1995)

30 mm meansize

Included frag-ments

1–2 infaunalbivalvespecies perstudy area

Total exceptinterior v.exteriorluster

Fragmentation,surface abra-sion, bioerosion,encrustation,luster & colorloss

Damage state Ternary taph-ograms offull-fre-quency data


TABLE 1—Continued.

Study




Set(s) oftaxa

analyzed

Surfacesexamined &

comparedTaphonomic

variables

Form ofdata for

individualshells*


differencesamong

assemblages

Norway low-tidebeach and Sonoraintertidal flat(Cutler & Flessa,1995)


5 spp Nor-way, 7comparablespp Sonora

Interioronly

Microbioerosion,dissolution/maceration,precipitationtextures @ 40�& 100� magni-fication

� damage Bar-graphs of%-� data

Sonora tidal flat(Cutler, 1995)

6 Whole, identifi-able shells only

2 infaunal bi-valve spp

Interioronly

Bioerosion, abra-sion, dissolu-tion

� damage Bar-graphs of%-� data

Florida carbonateshelf (Dent, 1996)


15 mostabundantmollusks; 2epifaunal& 1 infau-nal targetspp

Total only Abrasion, clionidboring, colorloss, encrusta-tion, edgerounding, frag-mentation, fine-scale alteration

Damage state Bar-graphs &ternarytaphogramsof frequencydata; clus-ter analysisof averagestate

Baja California bay(Meldahl et al.,1997b)

15 Whole disarticu-lated shellsonly

Infaunal bi-valves

Interioronly

Color loss, bio-erosion, encrus-tation, fine-scale alteration

4 summary grades Bivariateplots of av-erage state(Taphonom-ic Grade In-dex)

San Blas Archipela-go, Panama innershelf (Best, 2000)

8 Included frag-ments, bothidentifiable &unidentifiable

All bivalves;infaunal v.epifaunal;calcitic v.aragonitic;1 target in-faunal bi-valve

Interioronly; com-parison of3 interiormicro-structuralsectors

Disarticulation,fragmentation,boring, encrus-tation, fine-al-teration, edgemodification,alteration col-ors

Damage state Stacked bar-graphs offrequencydata; bino-mial confi-dence inter-vals;bootstrap-ping; con-tingency;non-metricmulti-di-mensionalscaling

Bocas del Toro, Pan-ama lagoon (Best& Kidwell, 2000a,2000b)

8 Included frag-ments, bothidentifiable &unidentifiable

All bivalves;infaunal v.epifaunal;calcitic v.aragonitic

Interioronly

Disarticulation/fragmentation,boring, encrus-tation, fine-al-teration, edgemodification

Damage state Bar-graphs ofhigh-thresh-old damage;binomialconfidenceintervals

San Blas Archipela-go, Panama inlet(this study)

2(2–4, 4–8, �8)


All bivalves;infaunal v.epifaunal;3 abundanttarget spp

Total, inte-rior v. ex-terior

Disarticulation/Fragmentation,boring, encrus-tation, fine-al-teration, edgemodification

Damage state Bar- and line-graphs ofhigh-thresh-old, �, andaveragestate data;stacked-barand ternarytaphogramsof full-fre-quencydata; bino-mial confi-dence inter-vals

30 KIDWELL ET AL.

FIGURE 1—Bivalve death assemblages were sampled from the top 10 cm of sediment along a cross-reef transect in the cul-de-sac Ulagsukuninlet on the lee-side of the Punta San Blas peninsula, Caribbean Panama (site numbers along top edge of cross-section, which is drawnwithout vertical exaggeration).

bean Panama, it is possible to (a) quantify the magnitudeof methodological artifacts in taphofacies patterns, and (b)identify the most efficient and robust means for acquiringdamage profiles of death assemblages.

STUDY AREA

The San Blas Archipelago lies along the Caribbeancoast of eastern Panama at �9.5� N between the Canaland the Colombian border (Fig. 1). Samples were collectedfrom a transect of reef-flat, back-reef, and down-slope sed-iments in the low-energy head of a small, V-shaped, man-grove-rimmed cul-de-sac (Ulagsukun inlet) on the leecoast of the Punta San Blas peninsula at the western endof the Archipelago (9� 33.02�N, 78� 59.73�W). This transectincludes a fieldsite from the work of Best (2000; site95SB053; see her more detailed map), where sedimenttype, rates of experimental shell degradation, radiocar-

bon-calibration of time-averaging, and pore-water geo-chemistry are all being analyzed (Ku and Walter, 1998;Best et al., 1999). For a general treatment of sediments inthe San Blas Archipelago, see Best (2000, Chapter 2).

The mouth of Ulagsukun inlet is edged on both sides by afringing reef of high coral diversity, which continues aroundthe end of the Punta San Blas (Fig. 1). Coral communitiesex-tend into the inlet as it narrows and shallows along a some-what sinuous course. The inlet appears to be a relict streamvalley, flooded during Holocene sealevel rise, with a deep,central channel and relatively flat subtidal shoulders, whoseedges support the major coral growth (Best, 2000, Chapter2). The inlet is very low-energy away from its mouth, pro-tected by its sinuous plan and narrow, reef-constrictedmouth from squalls on the Golfo, and protected by the pen-insula from Caribbean storms.

Reef development within the inlet becomes patchierwith distance from the mouth, and at the sampled tran-


TABLE 2—Taphonomic variables and damage states used to score bivalve shells in this study (under 10� magnification; modified from Bestand Kidwell, 2000a).

Variable 0 � No Damage 1 � Low Damage 2 � High Damage Notes

Encrustation None Covering 10% shell area Covering � 10% area Total coverage by all taxaBoring None Affecting 10% shell area Affecting � 10% area Non-predatory borings

onlyFine-Scale Surface

AlterationNone Dull or chalky Both chalky & pitted, or

erodedFor shell interiors, area

outside the pallial lineonly

Edge Modification None Chipped Rounded Commissural edge onlyFragmentation Valve still articulated Whole but disarticulated Large or small fragment

sect consists of a �7 meter-wide zone of live coral and skel-etal gravel within a predominantly fine-grained siliciclas-tic environment (Fig. 1; no vertical exaggeration in tran-sect cross-section). Siliciclastic sediment is shed from thePunta San Blas peninsula (see Chapter 2, Best, 2000).Halimeda-sand-rich mud with reef-derived coral debrisaccumulates on the narrow, gently sloping back-reef sea-floor along the mangrove shoreline. These back-reef sedi-ments are sparsely vegetated by patches of Halimeda,Penicillus, and Thalassia, and are heavily burrowed bycallianassid shrimps, which produce abundant mud-draped mounds on the seafloor and shelly sand-filled bur-rows to 50 cm or deeper in cores (Halimeda constitute 85wt-% of � 2 mm bioclasts; sand is �50:50 mix of siliciclas-tic and bioclastic grains). In the transition zone to the reef,Thalassia is replaced by scattered corals, mostly small �50 cm-diameter Porites and Diploria. Reef developmentoccurs along the 5–6-m deep edge of this marginal shoul-der and covers �50% of the seafloor at the line of transect.The central reef zone consists of isolated coral heads (Mon-tastraea, Diploria), small Porites-dominated thickets, andpolytaxic patch reefs up to 5 m diameter, with generallyhigh live-coral coverage. Sediments within this reef-zoneare strongly discolored (dark gray and rust) gravel of coraland molluscan skeletal debris. In situ coral patches andreef-derived coral rubble decrease in size and abundancedown the steep, mud-draped reef-slope (�45 degrees;transect in Fig. 1 is drawn without vertical exaggeration).This slope has an abrupt transition with the flat centralfloor of the inlet (10.5 m maximum depth), which is char-acterized by unvegetated, highly bioturbated, soft organic-rich clayey mud (3.0 wt-% TOC), with admixed bioclasticsand, sparse low callianassid mounds, and abundant liveholothurians and open burrows (1–2 cm diameter) of un-known origin. Site 5 is located on this flat floor near thebase of the reef-slope, and site 3 near the channel axis. Atsite 3, the mud contains only trace gravel �2 mm, which iscomposed of equal quantities of Halimeda (transportedfrom back-reef, along with mangrove litter), molluskshells, and unidentified bioclasts.

METHODS

A series of �5-liter surface sediment samples were tak-en along this transect on SCUBA in January, 1997, usinga large rectangular plastic scoop to excavate the top 10 cmof the seafloor (Fig. 1B). Each sample, which had been

placed in a heavy polypropylene bag for retrieval, was sub-sampled for quantitative grain-size analysis, with the re-mainder of the bulk sediment wet-sieved in the fieldthrough a 2 mm plastic mesh. These skeletal residueswere later washed with fresh water, air dried, and gentlyresieved through 2, 4 and 8 mm mesh in the lab.

Bivalve shells and fragments were picked from splits ofsamples from five sites along this transect: back-reef Hal-imeda-rich muddy sand (site 13); skeletal gravel from thecentral and outer edge of the reef flat (sites 10 and 7; ap-proximate replicates); shelly mud on the inlet-floor nearthe base of the reef-slope (site 5); and clayey mud on thedistal inlet-floor (site 3; experimental site 53 of Best,2000). Bivalve shell abundance, especially specimens �8mm, was relatively low at site 3 and, hence, that samplewas augmented with material collected from the samearea in previous years. In general, few bivalve specimenswere larger than 2 cm in any environment; thus, shell sizewas narrowly constrained even within the �8 mm portion.

Replicate samples were examined from the reef-flat(sites 7 and 10) because earlier studies found highestamong-sample variability within facies that include signif-icant hard substrata (Best and Kidwell, 2000a). Site 7 sed-iment was adjacent to and slightly down-slope from apatch of small (�50 cm) Montastraea annularis colonies,and was very coarse with just a thin (few mm) veneer ofmud on the surface (Fig. 1). Site 10 was rubble between a3–4 m diameter patch reef of Agaricia, M. annularis, andM. cavernosa, and a 2m-diameter reef of Agaricia and M.annularis, in the core of the reef-flat. Reef-slope site 5 andback-reef site13 are lithologically very similar to each oth-er in containing significant mud and reef-derived coral de-bris, but are not environmental replicates.

Each specimen was identified taxonomically to specieslevel when possible, and the degree of damage was scoredfor five macroscopic taphonomic variables, using 10 � ste-reoscopic magnification: encrustation, boring, fine-scalesurface alteration, commissural edge modification, anddisarticulation/fragmentation. In scoring damage, the op-erational criteria of Best and Kidwell (2000a; very similarto those of previous workers, Table 1) were followed, butthe degrees of damage were reduced to three states pervariable (Table 2). Damage to interior and exterior surfac-es of shells was scored separately, unless the specimenwas too damaged for this distinction to be made. Boring re-fers to any non-predatory penetration of the shells, and isprimarily by clionid sponges. Dominant macroscopic en-

32 KIDWELL ET AL.

FIGURE 2—Damage profiles for sites based on the frequency of high-damage states in the �8 mm size fraction of the complete bivalvedeath assemblage. These profiles are the baseline for evaluating dif-ferent methods of analysis, and reflect damage to shell-interiors of allbivalve specimens, including fragments, epifaunal as well as infaunalspecies, and specimens that cannot be identified to species or genuslevel. Note similarity among sites in rank-order importance of the fivevariables, but quantitative differences among environments in the fre-quency of highly damaged shells (95% binomial confidence intervals;a variable whose value lies outside the C.I. of another is significantlydifferent; sites 7 and 10 are replicate samples of the same reef-flatenvironment).

crusters are serpulid worms and bryozoans (both runnerand sheet forms). Although not quantified, the taxonomiccomposition of the fouling community was similar to thatin carbonate and siliciclastic habitats in the Bocas del Toroarea of Caribbean Panama (Best and Kidwell, 2000a; Fig.1). Fine-scale surface alteration denotes macroscopic (10�)dulling (loss of luster), chalkiness, pitting, and erosion orexfoliation of the original shell surface. Only with SEMcan this fine-scale alteration be interpreted confidently asthe product of microbioerosion, mineral dissolution, or-ganic-matrix maceration, and/or physical plucking (SEMof San Blas bivalves by Best, 2000; and see Cutler, 1995;Cutler and Flessa, 1995); thus, these process-terms arenot used. To control for microstructure and in vivo effects,damage to shell interior surfaces was scored only for thearea outside the pallial line (OPL), which is the ‘‘outcroparea’’ of the outer shell layer; damage to shell exteriorswas based on the entire exterior surface (also composed ofthe outer shell layer). This contrasts with previous work-ers who have treated the entire shell as a unit, or havesubdivided it into standard areas related to shell shape(e.g., Davies et al., 1989; others in Table 1). Edge modifi-cation refers to the shell commissure only, and not edgesproduced by fragmentation. SEM indicates that bioero-sion (microboring, grazing) rather than physical abrasionis the key process of rounding in these environments(Best, 2000; and see Best and Kidwell, 2000a). Edge chip-ping is a less severe modification of the commissure andprobably is generated by predators, especially crabs; somechipping may be from handling during collection, ship-ping, and washing. Edge-thinning was negligible. Frag-mentation appears to be largely the product of predation,as it is high in all environments despite their low physicalenergies. Differences among assemblages were evidentduring collection; thus, it is not believed that fragmenta-tion is greatly affected by handling. Fragments withouthingelines (‘‘minor’’ fragments of Staff and Powell, 1990)were treated the same as those including hingelines. Al-though some workers treat it as a separate variable, dis-articulation is treated here as a state within the fragmen-tation category (Table 2).

To permit analysis of co-occurring features on individu-al shells and facilitate re-examination, specimens weregiven unique identification numbers and archived (inshallow boxes using double-sided tape).

RESULTS

Baseline Damage Profiles

Damage to shell interiors in the �8 mm size fractionwas used to establish a baseline damage profile for eachbivalve death assemblage. All specimens of bivalve originwere assessed, including fragments and specimens thatcould not be resolved to species or genus level. The ‘‘taph-onomic signature’’ of each assemblage is expressed as ahigh-intensity damage profile (bar-graphs of Fig. 2), witheach variable (bar) treated independently and only the fre-quency of damage above a certain threshold intensity plot-ted (‘‘threshold damage profile’’). Here, the height of eachbar indicates the percentage (frequency) of shells in thesample that exhibit the highest degree of damage for thatvariable (as defined in Table 2), with 95% binomial confi-

dence intervals following Raup (1991). A variable whosevalue lies outside the confidence interval (C.I.) of anotheris significantly different.

Qualitatively, reef-flat shell gravels (replicate sites 7and 10) exhibit highest overall damage levels, declining tointermediate levels in muddy shell-gravel (back-reef site13 and reef-slope site 5), and the lowest observed levels inshell-poor mud (inlet-floor site 3). The signatures of thesesites show high similarity in the rank-ordering of tapho-nomic variables—in all profiles, encrustation boring fine-scale damage � edge rounding � fragmentation.Quantitatively, however, the profiles differ significantly.Pairwise tests indicate that, with the exception of the site3/13 comparison (the two sites with smallest samplesizes),profiles differ in 3, 4, or all 5 tested variables (Table 3).Sites that are most similar in environmental features areleast dissimilar in damage profile: reef-flat replicate sites7 and 10 differ in only 3 variables, as do reef-rubble-richsites 5 and 13.

In the sections that follow, baseline damage profiles areevaluated with respect to different methods of data collec-tion and analysis.

Use of Finer Size Fractions and Sieve Sizes

For all variables except fragmentation, there is a signif-icant size-fraction effect on taphonomic damage: in bothend-member assemblages (site 7 shell gravel and site 3mud), the 4–8 mm size fraction has the same or lower fre-quency of high-damage shells than �8 mm fraction, and


TABLE 3—Site by site comparison of high-intensity damage profilesof Figure 2, based on damage to the interior surfaces of shells �8mm. In each cell, the results are listed as a string for the 5 test vari-ables: encrustation, boring, fine-scale alteration, edge modification,and fragmentation. • � frequency of damage significantly different at0.95 C.I; n � not significantly different.

Site 10 7 5 3 (n � 54)

13 (n � 140)10 (n � 200)7 (n � 746)5 (n � 733)

• • • • • • • • • •• • nn •

• nn • •• • • • •• • • • •

nn • • n• • • • •• • • • •• n • • •

the 2–4 mm fraction has a significantly lower frequencythan either of the coarser fractions (Fig. 3A-B, Table 4).This effect is most pronounced for boring and encrusta-tion. (If damage to exterior shell surfaces is used, this size-fraction effect is even stronger, especially in shell-gravel ofsite 7, where encrustation, boring, and fine-scale alter-ation all exhibit disparities between size fractions of up to45 percentage points.) In contrast, fragmentation in the 2–4 mm fraction tends to be as high as, or higher than in,coarser fractions.

Because fine fractions comprise far more specimensthan coarse fractions in bulk samples, the taphonomic sig-nature of the finest size-fraction is expected to dominatethe overall damage profile of an assemblage. To test forthis in San Blas samples, data were pooled from size-frac-tions to simulate alternative sieve sizes, and it was foundthat fine sieves do yield lower damage profiles than coarsesieves (Fig.3C-D). [Note: data on the 2–4 and 4–8 mm sizefractions are based on subsamples of the total fine-mate-rial (1/10 to 1/3 of material actually associated with �8mm specimens); this was compensated for in poolingdata]. Using a 4 mm rather than an 8 mm sieve has a rel-atively small effect on damage profiles for site 7 shell grav-el: only encrustation and fragmentation shift significantly(confidence intervals are too large for the �8 mm fractionto determine an effect at site 3). In contrast, adding the 2–4 mm fraction to the others to simulate a 2 mm sieve shiftsmost variables to significantly lower values at both sites.

Thus, the use of finer sieves reduces the measured fre-quency of high-intensity damage in assemblages. Rank-order of variables is conserved, however, and among-sitedifferences in profiles are still evident (Fig. 3C-D).

Sample Size

Samples yield highly variable numbers of specimens,which may be insufficient (or, at the other extreme, over-kill) for confident determination of taphonomic signa-tures. Cumulative sampling curves, plotted during datacollection from successive sets of ten specimens (‘‘collec-tion curves’’), indicate that the true proportions of damagein the assemblage can be approximated with a sample sizeof 50 specimens, but that generally 120–150 specimensare needed to confidently establish damage profiles for thefive variables tested here. Figure 4 presents the results forthe 2–4 mm size fraction at two end-member sites (reef-

flat shell gravel site 7, and inlet-floor mud site 3; compareto Fig. 3C-D).

Increasing the sample size above the 150-specimenthreshold does not change the measured %-frequency ofdamage significantly, but narrows the confidence inter-vals. In this study, the very large sample sizes for somesites permitted statistical power to be maintained whenanalyzing subsets of the total assemblage, e.g., when tar-geting specific taxa or subgroups of taxa (see later sec-tions).

Analysis of Total Shell Surface Versus Interior-Only

The interior and exterior surfaces of most shells exhibitthe same degree of damage from boring, encrustation, andfine-scale surface alteration (cells along diagonals in con-tingency Table 5). When damage on the two surfaces is un-equal, however, exterior surfaces are at least twice as like-ly to carry the greater damage (upper right cells in Table5). Some of this difference may be truly taphonomic; for ex-ample, if the mineralogy, microstructure, or topography ofthe shell exterior is more prone to attack by post-mortemphysical, chemical, and biological processes. However,much of the ‘‘excess’’ damage to shell exteriors may accrueduring the life of the bivalve, especially for taxa that areepifaunal or lack a protective outer periostracal layer. Dis-parity in frequencies of high-damage between shell inte-rior and exterior surfaces is evident in all size fractions(Table 4).

Scoring the extent of shell damage to the total shell sur-face (interior plus exterior; this is the procedure in mosttaphofacies analyses, Table 1) should yield higher %-fre-quencies of damage in an assemblage than when damageis scored on the basis of interior surfaces alone. This is in-deed the case in San Blas assemblages for high-intensitydamage profiles, and these differences are significant inall environments except inlet-floor mud (site 3, low n; Fig-ure 5). This effect is evident in all size-fractions of the bi-valve death assemblage down to 2 mm.

Despite the inflationary effect on %-frequency values,among-site differences in damage profiles are largely ro-bust: (a) edge-rounding and fragmentation are completelyunaffected; (b) the rank-order importance of the three af-fected taphonomic variables is unchanged (frequencies offine-scale alteration exceed those of boring, which exceedthose of encrustation, regardless of whether total shellsurfaces or interiors-only are scored); (c) both scoringmethods detect the same taphonomic trends among sites;and (d) pairwise differences in damage levels among envi-ronments are still statistically significant (sole exceptionis the lack of difference in boring levels between sites 3and 5 for shell interiors).

Thus, although damage to shell exteriors has ambigu-ous origins and systematically increases measured valuesof damage, its inclusion does not obscure among-habitatdifferences that are strictly taphonomic in origin. Thisfinding has particular relevance to studies in lithified sed-iments, where it is commonly difficult to examine shell in-teriors.

Exclusion of Fragments

In the �8 mm size fraction (interior surfaces only), frag-ments yield higher frequencies of all types of damage than

34 KIDWELL ET AL.

FIGURE 3—Effect of size fraction (A-B) and sieve size (C-D) on damage profiles for site 7 shell gravels and site 3 muds; 95% binomial C.I.s,which are too small to plot for all bars in site 7 data. When pooled, the far more numerous and typically less damaged specimens of finer sizefractions reduce the overall frequency of damage in the assemblage. Pooling the 4–8 mm and �8 mm fractions has small effect on damageprofiles (only encrustation and fragmentation change significantly at site 7; no differences at site 3); adding the 2–4 mm fraction has a strongereffect on most variables at both sites.

TABLE 4—Percentage-point difference in the frequency of high-damage states between the �8 mm size fraction and the 4–8 mm and 2–4mm size fractions for reef-flat shell-gravel (site 7) and inlet-floor mud (site 3). For encrustation, boring, and fine-scale damage, values arebased on data from shell interiors; exterior-surface values are in boldface.

Sizefrac-tion

(mm)

Encrustation

4–8 2–4

Boring

4–8 2–4

Fine-ScaleAlteration

4–8 2–4

EdgeRounding

4–8 2–4

Fragmentation

4–8 2–4

Site 7�8 mm

15% lower25% lower

25% lower45% lower

n.s.10% lower

40% lower40% lower

n.s.10% lower

n.s.20% lower

n.s. n.s. 10% lower n.s.

Site 3�8 mm

n.s.10% lower

n.s.15% lower

n.s.n.s.

10% lower10% lower

n.s.n.s.

n.s.n.s.

n.s. n.s. 15% higher 15% higher

whole shells, and these differences are significant in allenvironments tested except site 3 mud (Fig. 6). In SanBlas assemblages, where fragments constitute 50–80% ofspecimens in bivalve death assemblages (Fig. 2A), the ex-clusion of fragments yields damage profiles significantlylower than if all bivalve material is scored (generally by 10

or more percentage points per variable). The rank order ofvariables (shape of the profile) is conserved.

Thus, the intensity of perceived damage in the death as-semblage depends on the proportion of whole and brokenshells. Profiles based on whole shells alone will havedamped values compared to total-assemblage profiles, but


FIGURE 4—For each of the 5 taphonomic variables, damage frequency distributions are confidently stable at sample sizes of 120–150specimens. Data for shell interiors from 2–4 mm size fractions (Fig. 3A-B).

in San Blas assemblages, among-site differences are stillsignificant.

Exclusion of Poorly Identified Shells

Shells that only can be identified as ‘‘bivalve,’’ ratherthan resolved to family or finer taxonomic level, constitute10–20% of specimens in San Blas death assemblages. Allof these specimens are fragments and exhibit significantlyhigher levels of damage, generally by 10 to 45 percentagepoints (except at site 3 mud, with low n; Fig. 7).

Thus, excluding taxonomically poorly resolved bivalvespecimens lowers the damage profile of the assemblage,

but the proportion of such shells in San Blas assemblageswas not sufficient to affect the outcome significantly.Rank-order of variables is conserved in both subsets of theassemblage, as are among-environment trends in damagelevels.

Analysis of Infaunal Shells Only

When all data from the study are considered (i.e., shellsfrom all sites and size fractions are pooled), epifaunalshells exhibit significantly higher frequencies of all kindsof damage except fragmentation, relative to infauna. Thisdifference persists even if only aragonitic epifauna are

36 KIDWELL ET AL.

TABLE 5—Contingency data for damage to interior versus exteriorshell surfaces, for total data (all sites, all size fractions). Most speci-mens show identical levels of damage to both surfaces (boldface di-agonal cells in each matrix), but, when damage is unequal, it is inmost cases heaviest on the exterior surface (numbers in upper righthalves of each matrix).

Exterior Surface Damage

0 1 2

Encrustation Interior:012

757130

6

538768

91

78341351

Boring Interior:012

550154

1

1761015

130

5238791

Fine-Scale Alteration Interior:012

496

130924119

2186

1643

FIGURE 5—For bivalve shells �8 mm at all sites, the frequency ofdamage to the total surface area (white bars) is consistently higherthan that scored for shell interiors only (dark bars) for each of thethree relevant taphonomic variables (encrustation, boring, fine-scalealteration), and these differences are significant at all sites except site3 mud (low n; 95% binomial C.I.).

considered (chamids, arcids, some anadarids; Fig. 8) and,thus, is linked to life-habit rather than to shell mineralo-gy.

Because total death assemblages include not only epi-fauna but also specimens that cannot be classed confident-ly (mostly high-damage fragments), damage assessmentsbased on infauna-only should yield lower damage profilesthan the total assemblage, and this is true in San Blas en-vironments. At sites where epifauna constitute 50% ormore of specimens in the death assemblage (site 7 reef-flatshell gravel, site 5 reef-slope muddy rubble, site 13 back-reef muddy skeletal sand), infauna-only profiles are sig-nificantly lower for 2 to 3 taphonomic variables (encrus-tation, boring, and fine-scale alteration; Fig. 9). Becausethis effect occurs at all sites, among-environment trends indamage levels are conserved.

Restriction of Data to Single Target Taxa

Of the total available assemblage, species of the infau-nal bivalves Laevicardium and Pitar and of epifaunal scal-lops (especially Chlamys) were sufficiently abundantacross multiple habitats to test for taxon-specific damageprofiles in sites 7 (reef-flat shell gravel) and 3 (inlet-floormud), using specimens from all size classes down to 2 mm(Appendix). Each of these target taxa exhibits higher fre-quencies for most variables at site 7 than at site 3, consis-tent with the total death assemblage (compare Fig. 10 to 2mm sieve data in Fig. 3C-D). Laevicardium is most sensi-tive to extrinsic environment, differing in all 5 variablesbetween the two sites, whereas Pitar and Chlamys differsignificantly in only 2 variables (encrustation and boring;the total death assemblage �2 mm differs in 4 variables).Laevicardium also shows a strong offset in damage-levelsbetween the two environments, comparable to the sensi-tivity of the total death assemblages; among-environmentdifferences in the other two target-taxon profiles are mut-ed. Laevicardium constitutes 19% of specimens in the site

3 assemblage and 13% at site 7; thus, it does not approxi-mate total-assemblage damage simply because of numeri-cal dominance. The proportional abundance of the othertaxa are 18 and 7% for Pitar, and 20 and 20% for Chlamys(see Appendix for complete assemblage composition).

The damage profiles of the three taxa differ significant-ly from each other within each site, both in frequency ofdamage and in rank-order of variables (Fig. 10), showing


FIGURE 6—Fragments consistently yield significantly higher frequencies of damage (interior surfaces) than whole shells in the same � 8 mmsize fraction, in all environments tested. 95% binomial C.I.s.

the effect of intrinsic, taxon-specific factors on shell condi-tion. Some of these differences are understandable interms of shell morphology. For example at site 7, Chlamysexhibits less edge-rounding than the others probably be-cause of its exceptionally thin commissure, which is eitherless prone to rounding (versus chipping) or less easy to rec-ognize as rounded under 10� magnification; at both sites7 and 3, Pitar exhibits less fragmentation than Laevicar-dium, probably because most Pitar specimens are small[2–4 mm fraction, versus �8 mm for Laevicardium, andfragmentation is elevated in fine size fractions in general(see earlier section)]. Explanations for other differencesare more obscure (e.g., Pitar’s higher fine-scale alterationand rounding than Laevicardium, despite the same gen-eral microstructure and thinner commissures).

Of the three taxa, Laevicardium damage provides thebest match in level and rank-order to the profile of the to-tal death assemblage in both site 7 and 3 (Fig. 10 versus 2mm sieve data in Fig. 3C-D). Laevicardium profiles alsoagree well with the total-infaunal damage profiles (Fig.10). It is not clear, however, why this particular taxon pro-vides the closest proxy to more inclusive datasets, becauseLaevicardium dominates neither the total death assem-

blage nor the infaunal component of either site (Appen-dix). A second cardiid genus, Americardium, occurs in suf-ficient abundance at site 7 to evaluate, but on all vari-ables, damage to this coarsely ribbed quadrate bivalve dif-fers significantly from Laevicardium (large, smooth-surfacedegg-shape), and matches Pitar (small, smooth-surfaced ve-nerid) on all variables except fragmentation.

Other Forms of Data and Methods of Analysis

Data on damage to individual shells can be gatheredand manipulated in several ways. Key choices in data-gathering include whether taphonomic variables aretreated (1) as a set of linked conditions, that is multi-fac-torial grades sensu Flessa et al. (1993; Meldahl et al.,1997a), or (2) independently (e.g., encrustation versus bor-ing, etc.), as in this and most other studies (Table 1). Keychoices in handling data on damage states include wheth-er (a) the full frequency distribution is analyzed (stackedbar-graphs of Feige and Fursich, 1991; ternary tapho-grams of Kowalewski et al., 1994, 1995), or (b) the infor-mation is reduced to, for example, the frequency of high-intensity damage (as here; % shells in the highest category

38 KIDWELL ET AL.

FIGURE 7—Shells that can be classed only as bivalve in origin (bars labelled ‘‘unid’’) consistently yield significantly higher frequencies ofdamage than specimens that can be identified to species, genus, or family level (bars labelled ‘‘ID’’); �8 mm size fraction, interior damageonly; 95% binomial C.I.s. ‘‘All’’ denotes total sample.

FIGURE 8—Based on analysis of the total San Blas dataset (all sites, all size fractions), shells of epifaunal bivalves exhibit significantly higherfrequencies of all kinds of damage except fragmentation than infaunal shells, even if (as here) epifauna are limited to those with aragoniticshells.


FIGURE 9—Damage profiles generated by infaunal shells only (bars labelled ‘‘Inf’’) are lower than total-assemblage profiles (‘‘All’’), whichinclude epifaunal specimens (‘‘Epi’’) and specimens that cannot be classed to life-habit (mostly highly damaged fragments). Data are for �8mm size fractions, interior-surfaces only; 95% binomial C.I.s.

FIGURE 10—Damage profiles for �2 mm specimens of the infaunal bivalves Laevicardium and Pitar, and the epifaunal scallop Chlamys atsites 7 and 3 (multiple species of each genus; Appendix). Each target taxon detects a decrease in damage-levels from site 7 to site 3, butdiffer significantly from one another within each site in level of damage and rank-order of variables. Laevicardium provides the closest matchto the damage profile of the total death assemblage (�2 mm data in Fig. 3C-D).

on the damage scale), presence-absence (% shells havingany degree of damage), or average damage state (mean de-gree of damage: Parsons, 1993; Meldahl et al., 1997a).

Figure 11 presents the full frequency distribution ofdamage states in the �8 mm size fraction (interior dam-

age only) as stacked bar graphs. (The baseline high-inten-sity damage profile of Fig. 2 uses only the black segmentsof these bars.) Although confidence intervals for San Blasdatasets are sufficiently small that they do not overlapamong bar-segments, this could be a complicating problem

40 KIDWELL ET AL.

FIGURE 11—Stacked-bar graphs displaying the full frequency distribution of damage states (0–2, defined in Table 2) for taphonomic variablesin �8 mm size fractions (interior damage only). The black segments of bars indicate the frequency of high-damage states (2), and are thebasis of the baseline damage profiles of Figure 2. 95% binomial C.I.s. Encrust � encrustation; Fine Alter � fine-scale surface alteration; EdgeMod � edge modification (edge-rounding is state 2); Frag � fragmentation.

in datasets with smaller sample sizes. Stacked bar-graphslike these are effective in revealing similarities amongsites in rank-order of variables as well as significant dif-ferences on a variable-by-variable basis, and any numberof states per variable can be accommodated.

Because degrees of damage in San Blas assemblages arecategorized into three states (bins) per variable, thesedata also can be plotted on ternary diagrams (Fig. 12A-E).Taphograms are very effective in depicting the end-mem-ber nature of the reef-flat shell gravel and the inlet-floormud, the specific natures of these assemblages, and, withthe addition of 95% confidence intervals, the significanceof among-site differences. For all variables except frag-mentation, (a) replicate reef-flat shell gravel sites 7 and 10plot as nearest neighbors to each other and also closest tothe ‘‘high damage’’ pole, (b) inlet-floor mud site 3 plotsmaximally distant from these shell-gravel and closest tothe ‘‘no damage’’ pole, and (c) muddy shell sites 5 and 13plot in intermediate positions and quite close to one anoth-er on most variables. These taphograms also are effectivein demonstrating that even when sites are very similar ornot significantly different in some kinds of high-intensitydamage, they may still separate on the basis of pristine(no-damage) shells (e.g., sites 3, 5, and 13 with respect toboring).

When all variables for all sites are plotted on a singletaphogram (Fig. 12F), the end-member sites (7 and 10,versus 3) occupy non-overlapping areas, again demon-strating the utility of taphograms in distinguishing sam-ples. This plot highlights a disadvantage of taphograms,however, in that similarities among sites in the rank orderof variables are hidden completely, in contrast to the

stacked bar graphs and high-intensity damage profiles.Another disadvantage is that degrees of damage must becategorized into (or otherwise collapsed into) 3 states pervariable (see examples in Kowalewski et al., 1995).

In Figure 13A-B, taphograms of encrustation and bor-ing data are used to display the effect of using differentsize fractions of an assemblage (points within clouds) andof gathering data from total shell surfaces rather thanfrom interior surfaces only (points connected by arrows).On taphograms, 2–4 mm size fractions plot significantlycloser to the ‘‘good’’ (i.e., pristine, no damage) corner thantheir counterpart 4–8 mm and �8 mm fractions for thesame site, and the magnitude of the effect for encrustationand boring is sufficient to move points into different sec-tors of the diagram (data for sites 3 and 7 only; Fig. 13A-B). When data for the �8 mm faction is based on total shellsurface rather than interiors only, the value for each siteshifts significantly toward a higher-damage corner butgenerally stays within the same sector of the taphogram(Fig. 13A-B). Clouds drawn around sets of points reflectthe size of the combined 95% confidence limits for pointsfrom that site.

The divergence of damage levels on target taxa fromeach other and from damage to the total-assemblage (dis-cussed in previous section) is clarified by plotting on ter-nary taphograms (Fig. 13C-F). Five patterns are evident.

(1) Laevicardium is the only taxon that consistently re-tains the two end-member sites in their correct relative po-sitions for all five variables; that is, placing site 3 signifi-cantly closer to a lower-damage corner than site 7. Chla-mys and Pitar each capture these relative positions basedon encrustation and boring levels, but scramble the site


FIGURE 12—Data from Figure 11 replotted on ternary taphogram template of Kowalewski et al. (1995; see other figures for confidenceintervals). Taphograms A-E plot the relative frequencies of damage states in each site for single variables. Taphogram F plots data for allvariables and all sites on a single diagram, underscoring the end-member nature of damage levels in reef-flat shell gravel sites 7 and 10, andinlet-floor mud site 3, which form non-overlapping clouds of points. These diagrams are possible only because degrees of damage werecategorized into three states (bins) per variable (Fig. 11, Table 2).

positions or show no significant difference between thetwo sites when scored for other kinds of damage.

(2) Laevicardium nonetheless is offset significantly fromtotal-assemblage values for some variables, and no targettaxon does better than the others at matching total assem-blage damage (2 mm sieve data) for all variables. This iscontrary to the finding using high-intensity damage,where Laevicardium was a close match to the 2 mm-sievetotal-assemblage data (earlier section; Figs. 10 and 3C-D).

(3) Target taxa display relatively high disparity for allvariables at site 3 compared to their behavior at site 7(compare clouds of points). Fragmentation is the only ex-ceptional variable, with high disparity among taxa at bothsites.

(4) For all variables, site-3 Chlamys shells plot as closeor closer to site-7 Chlamys than to other site-3 target taxaor to the total-assemblage value at site-3, suggesting ei-ther a strong intrinsic control on damage levels (i.e., simi-lar signature regardless of post-mortem environment) orimport of specimens from site 7, where they acquired theirsignature. Consideration of the high-intensity damage

profile (Fig. 10), showing the same basic profile for Chla-mys at both sites but simply higher frequencies of damage,argue that intrinsic effects rather than post-mortemtransport are most important. Site-3 Pitar specimens lieas close or closer to site-7 Pitar than to other site-3 valuesfor only three variables (fine-scale alteration, edge modi-fication, and fragmentation), whereas Laevicardium dam-age levels at the two sites are quite different and closer tothe value for the total-assemblage from which they aredrawn. Thus, intrinsic factors are moderately importantfor Pitar, and least important for Laevicardium, whosedamage levels in the two sites are quite different fromeach other and closer to the value for the total-assemblagefrom which they are drawn.

(5) Consistent with analysis of the total assemblage(Figs. 8 and 9), the epifaunal bivalve Chlamys plots closerto a higher-damage corner than the other two (infaunal)target taxa for 3 of the 5 taphonomic variables.

Because most shells exhibit some degree of fine-scale al-teration, edge modification, and disarticulation/fragmen-tation, presence-absence information is less effective than

42 KIDWELL ET AL.

FIGURE 13—Effect of size-fraction, total surface area, and target taxa on full-frequency data, displayed using taphograms. Clouds drawnaround sets of points for a site reflect the size of the combined 95% confidence intervals for those points. (A and B): At end-member sites 3and 7, damage levels for the fine size-fraction (2–4 mm) plot in lower-damage portions of taphograms than counterpart 4–8 mm and �8 mmfractions, with the effect most pronounced for boring and encrustation (note position within clouds). Data points for size-fractions denoted,respectively, by 2, 4, and 8 attached to site number; damage to interior surface only. Scoring damage to total shell surface area (symbol‘‘t’’) rather than interior-only shifts values toward a higher-damage portion of taphogram (�8 mm size fraction; arrow). (C-F): Damage levelsfor target genera Laevicardium (-L), Pitar (-P), and Chlamys (-C) plot at varied distances from total- assemblage values (based on all �2 mmspecimens; interior damage only; boldface site numbers with 2 suffix). Laevicardium is the only target that captures the relative positioningof total-assemblage damage on the plots, with site 3 specimens closer to low-damage corners than site 7 specimens, but is not a consistentproxy for (nearest-neighbor to) total-assemblage damage values. Site-3 Chlamys plots closer to site-7 Chlamys than to other site-3 target taxaor total-assemblage values, indicating intrinsic control of damage patterns in this epifaunal genus or possible import from site 7.


FIGURE 14—Sensitivity of between-site differences in damage levelsand rank-order of variables as a function of damage metric, plotted asline-graphs (�8 mm size fraction, interior damage only). (A) Frequen-cy of high-intensity damage (same data as Fig. 2, and black segmentsof bars in Fig. 11). (B) Presence-absence data (combined shadedsegments of bars in Fig. 11); note that only 2 variables differ in valueamong sites; rank-orders of variables are preserved; fragmentationhere is disarticulation (see Table 2). (C) Average damage state (cal-culated from full bars in Fig. 11; 0 � no damage, 1 � low damage, 2� high damage from Table 2); muted ability from high-intensity dam-age to differentiate levels of damage at sites, but generates the samerank-ordering of variables and detects similarity in those orders amongsites.

high-intensity damage and full-frequency distributiondata in discriminating the San Blas samples (presence-ab-sence data combines the shaded high-damage and moder-ate-damage segments of bars in Fig. 11). Most pairwisecomparisons of sites differ significantly in only two vari-ables at most (most commonly boring and encrustation),including end-member sites 3 and 7. The lack of resolvingpower of presence-absence data is underscored visually ina line-plot of the same data (Fig. 14B). Compared withhigh-intensity frequency data (plotted in line-graph formin Fig. 14A), which shows damage of all types increasingacross the transect to maximum values in reef-flat shell-gravel sites 7 and 10, only two of the variables exhibit suchtrends in presence-absence data.

‘‘Average damage state’’ (calculated from the data inFig. 11) shows intermediate power to discriminate envi-ronmental differences in damage frequency to high-inten-sity and presence-absence data, as shown in the line-graph of Figure 14C. The rank-order of variables is thesame as for high-intensity data (Fig. 14A), but this wouldnot necessarily occur if the variables differed in numbersof states—that is, if some variables had 2 or 4 or 5 states,instead of all having 3.

Use of Multiple Evaluators

All data used in this study were collected by a single op-erator (TAR). However, to estimate operator error, two ofus (one trained by the other) independently collected datafrom the same sample (two small samples from site 3, �8mm fraction, n � 44). The two operators showed highagreement in assessing the frequency of high degrees ofdamage in the assemblage, but differed significantly inparsing shells among categories of low and no damage. Infact, differences were sufficient for the same assemblageto plot in different sectors in ternary taphograms, wherethese full-frequency data are employed. On the positiveside, the operators were consistent in either under- orover-scoring the pristine (no-damage) category relative tothe other; hence, a compensation factor could have beencalculated.

These results suggest that (a) not surprisingly, data col-lected by a single operator will be superior to that pooledfrom multiple operators, but (b) if multiple operators arenecessary, the most reliable data may be those based onscoring a single, high threshold of damage, rather than thefrequency distribution of all damage states for a variable.The larger the number of states per variable, the largerthe likely error among operators because of subjectivity inestimation. Thus, it is worth stressing the obvious that,when using multiple operators, double-blind tests of con-sistency are essential.

DISCUSSION

The results of these sensitivity tests, and their implica-tions for taphofacies analysis, are summarized in Table 6.Of the variables tested, sample size has received little ex-plicit attention for its effect on taphonomic signatures.Based on the present analysis, samples smaller than 20specimens (whether single sample or set of pooled sam-ples) are insufficient to determine taphonomic signaturesfor multi-species assemblages, especially given the

44 KIDWELL ET AL.

TABLE 6—Summary of test results.

Test Results (for high-intensity damage) Caveats & Implications

Sieve size Lower frequencies of damage (except fragmentation)in finer size fractions (4 mm), but basic shape ofdamage profile and between-environment differenc-es are maintained

Measured damage levels will be lower in studies usingfiner sieves, due to numerical dominance of smallspecimens

Sample size 120–150 specimens necessary for profile to stabilize inmulti-species assemblages

50 shells will give rudimentary information, but sta-tistical power is low

Damage to totalshell vs. shell in-terior only

Shell exteriors sustain equivalent or greater damage,but basic shape of damage profile and between-envi-ronment differences are maintained

Effect is strongest in coarsest grained environments;damage levels based on total shell surface will beinflated relative to those on interiors-only

Exclusion of frag-ments

Fragments exhibit higher frequencies of all damagetypes, but same rank-order of variables as wholeshells

Depending on frequency of fragmentation, damageprofile of assemblage is lowered if score wholeshells only

Exclusion of taxo-nomically poorlyresolved speci-mens

All such specimens are fragments, with higher fre-quencies of damage than shells that can be identi-fied to family, genus, or species

Depending on frequency of such specimens, damageprofile of assemblage is lowered if score well-re-solved specimens only

Focus on infaunalspecies

Infaunal shells exhibit lower frequencies of damagethan epifauna, even when mineralogy is held con-stant (aragonitic taxa only)

Infauna-only profiles are lower than total-assemblageprofiles, but between-environment trends are con-served

Focus on damageto single taxa

Individual taxa detect environmental differences indamage-levels, but only one of three taxa tested isas effective as the total death assemblage

Taxa can yield disparate rank-ordering of variables;intrinsic controls on shell damage reduce the utilityof single taxa as proxies of total-assemblage damage

Other metrics ofassemblage con-dition

Presence-absence data least effective and averagedamage-state intermediate power to distinguishsites

Ternary taphograms make effective use of full-fre-quency data, but provide no rank-order insight onvariables in signature

Operator error Consistent results in recognizing high-damage statesof variables, but significant variance for zero andlow-damage states

If multiple operators to be employed, data on high-in-tensity damage may be most robust

among-taxa heterogeneity in signature evident from ourtarget-taxa analysis (above). At the other extreme, 20,000specimens are almost certainly overkill; this range is en-compassed by studies in Table 1. The 20–30 specimensrule-of-thumb used by many may be sufficient for tapho-nomic characterization of single target taxa. However, thebest overall strategy is to generate collection curves to es-tablish, using material at hand, when an adequate num-ber of specimens has been examined and data collectioncan reasonably stop, rather than targeting a piori a partic-ular number of specimens or volume of sediment. Fromthe suite of San Blas habitats, minimum necessary samplesizes were on the order of 50–100 specimens, depending onthe homogeneity of damage and the number of taphonom-ic variables. Based on rarefaction analysis of other sam-ples from the Bocas del Toro and San Blas areas of Pana-ma, Best (1998; 2000) found that sample sizes of 200 spec-imens are necessary when a fuller array of taphonomicstates (not collapsed to 3 per variable), an additional vari-able (alteration color), and the taxonomic composition ofborers and encrusters are to be quantified.

Differences in damage levels among size fractions andbetween whole shells and fragments have been examinedin molluscan death assemblages by a number of workers.Staff and Powell (1990) used a protocol very similar to theone used here for mollusk assemblages from the Texascontinental shelf (silty siliciclastic sands) and came to sim-ilar results: fragmentation is higher in the fine fraction (4–12 mm), but ‘‘biotic interactions’’ (encrustation and boring,

including predatory drillholes) are higher among largeshells (�12 mm, 2 of 3 sample sites); and fragments andwhole specimens have similar overall signatures, but frag-ments—especially ‘‘minor fragments’’ lacking hinges orwhorl apices—exhibit much greater ‘‘dissolution’’ (fine-scale alteration of all types) and edge rounding. In conti-nental slope cold-seep assemblages, Callender and Powell(1992; Callender et al., 1992) also found higher damageamong fragments from ‘‘dissolution’’ and rounding, butwhole shells had higher frequencies of ‘‘biotic interac-tions.’’ The same patterns pertain in higher-energy and/orcoarser-grained settings: higher levels of damage amonglarger shells (e.g., from borings by Boekschoten, 1966; ofall damage types for bivalve shells �2 cm by Bosence,1979; from ‘‘dissolution’’ by Davies et al., 1989), and higherdamage levels among fragments than among whole shells(e.g., edge-rounding and surface abrasion by Davies et al.,1989; and see Pilkey, 1964, and Pilkey et al., 1969, onsame pattern in palimpsest shelf sands).

Studies of the effect of unidentifiable specimens consis-tently find that their inclusion increases the documenteddamage level for a sample (qualitative statement in Da-vies et al., 1989; data in Staff and Powell, 1990, and Cal-lender et al., 1992; this study, Fig. 7). Davies et al. (1989)and Staff and Powell (1990) recommended including ‘‘un-identifiable’’ material for its greater sensitivity to detailedenvironmental differences. In contrast, ‘‘unidentifiable’’material (all fragments) in the present study was notfound to have greater sensitivity than identifiable materi-


al: damage levels were higher in all environments, butamong-environment differences were proportional tothose for identifiable material. As a caveat, because oper-ators will differ in the cut-off they use between identifiableand unidentifiable material (if the protocol is to ignore un-identifiable material), and because operators also will dif-fer in their skill in identifying ‘‘unidentifiable’’ shells asmolluscan (that is, how far down into ‘‘debris’’ they willventure), unidentifiable material creates a certain degreeof subjectivity in taphofacies analysis regardless of proto-col. That is, the more exclusionary the operator, the lowerthe damage profile. Individual operators should achieveconsistency within a study, but cross-study comparisonswill be less secure. Thus, the key caveat in cross-studycomparisons will be to treat studies that include uniden-tifiable fragments separately from those that use onlyidentifiable fragments, and to partition both of these fromstudies that use only whole shells.

The effect of focusing on only infaunal or epifaunal spec-imens has been evaluated in few studies, but these consis-tently find higher damage among epifaunal specimens (asfound herein; Figs. 8 and 9; and see plots of epifaunalChlamys compared to the two infaunal target taxa in Fig.13). Using the same protocol for samples from Bocas delToro, Panama, Best and Kidwell (2000b) found higherdamage levels among epifaunal shells (interior surfaceonly) in both soft-sediment and hard-substratum environ-ments. Parsons (1993; Parsons and Brett, 1991) citedhigher damage levels in epifaunal than infaunal subsets ofreefal samples from the Virgin Islands (insufficient epi-fauna to test in soft-sediment settings); this included dam-age to shell exteriors that was recognized as potentiallybeing ecological rather than post-mortem in origin. Dent(1995) also reported higher levels of encrustation and bor-ing for two epifaunal target taxa than for an infaunal tax-on in a range of carbonate environments of southern Flor-ida (also includes damage to exterior shell surfaces).

One of the most interesting and tantalizing results isthe effect of individual ‘‘target’’ taxa on perceived damagelevels. As reviewed by Best and Kidwell (2000b), individ-ual taxa typically yield different levels of damage—and insome instances different rank-orders of variables, ortrends along environmental gradients—from the sourceassemblage as a whole and from other target taxa, evenwhen those taxa are of the same life-habit or mineralogy(e.g., Feige and Fursich, 1991; Dent, 1995). Staff and Pow-ell (1990) found that their target taxa (delicate infaunal bi-valves Abra and Tellina) bore very similar signatures toeach other, differing only in the percentage that retainedperiostracum and the degree of edge modification; thesetaxa strongly resembled the source assemblage (perhapsbecause they jointly dominate it, constituting 41% and23% of the assemblage, respectively; ditto target taxa inCallender and Powell, 1992). Thus, the results of the pres-ent study are consistent with previous studies in recogniz-ing high inter-genus variability in taphonomic signature.In the present study, disparity among target taxa is great-er in the fine-grained facies than in the shell-gravel facies(site 3 versus site 7 in Fig. 13C-G), and the magnitude ofthis effect on signature is comparable to that of size-frac-tion (compare with clouds of points for same sites in Fig.13A-B). Consequently, although target taxa are an ex-tremely valuable approach for some purposes (e.g., cali-

brating taphonomic clocks among environments; Meldahlet al., 1997a), their use as a substitute for or as an indica-tor of overall assemblage condition in paleoenvironmentalanalysis requires caution. The raw data in the Appendixunderscore another, secondary effect of the choice of sievesize on taphonomic signature, and that is variation in tax-onomic composition among size fractions, an effect wellknown to benthic ecologists (Reish, 1959; Bachelet, 1990).To the extent that taxonomic composition influences taph-onomic signature, then sieve-size will influence whichpart of that signature is captured or weighted; this effect isin addition to the differential attractiveness of large andsmall specimens to colonizing infestors.

Other studies find that damage to exterior surfaces ofshells are generally as great or greater than damage to in-terior surfaces, based on measures of ‘‘dissolution’’ and/orabrasion (i.e., fine-scale alteration; Davies et al., 1989;Staff and Powell, 1990; Callender et al., 1992; Kowalewskiet al., 1995). These results are consistent with the presentinvestigation. Additionally, there is significantly higherdamage from encrustation and boring to shell exteriors,which has not been tested in previous studies (‘‘biotic in-teractions’’ negligible in frequency, or damage scored onlyfor total surface area).

Ternary taphograms are extremely effective graphicallyin capturing differences and similarities in full-frequencydata, but it is cumbersome to include the confidence inter-vals that would permit the dispersion of points to be eval-uated statistically. Although they use only part of the fre-quency data, threshold damage profiles (e.g., Fig. 2; orline-graph variants of these—Fig. 14A) are most effectivein evaluating, both statistically and visually, the qualita-tive aspects of these taphonomic signatures, namely therank-order importance of variables. Thus, these are avaluable and arguably essential complement to tapho-grams. These require no additional data to be collected,only a regraphing of a subset of data, and the admittedlyhigher space–demand relative to points on a ternary taph-ogram (Kowalewski et al., 1994) can be reduced by plot-ting only one damage-profile per facies (pooled set of sam-ples; e.g., Best and Kidwell, 2000a), rather than one persample as here. Rank-order insights are otherwise diffi-cult to extract from sets of univariate taphograms (e.g.,Fig. 12), from average-state data (which suppress among-site differences), and from pairwise and multiple analysesof variance (e.g., tables as in Staff and Powell, 1990), and,as a result, the ‘‘overall signature’’ of a death assemblageis down-played.

In these San Blas samples, high-intensity damage wasthe most useful threshold for plotting damage profiles be-cause the contrast between reef and mud assemblages indamage intensity was so great. Notice from the full-fre-quency data in Figure 11, however, that to distinguishamong fine-grained sites, a lower threshold would beequally or more effective (e.g., state 1 rather than state 2damage for sites 13, 5, and 3). Other muddy assemblagesfrom the San Blas Archipelago (Best, 2000) and from Bo-cas del Toro (Best and Kidwell, 2000a), for example, havelower intensities of boring and especially encrustation, asdo assemblages from temperate subtidal mud (e.g., negli-gible ‘‘biotic interactions’’ on shells from continental shelfand slope sites summarized in Callender et al., 1992), andtheir signatures might be indistinguishable were only

46 KIDWELL ET AL.

high-intensity damage to be plotted. The taphonomic sig-natures of sites 13 and 5 almost certainly are influenced tosome degree by reef-derived skeletal material, introducedeither via transport or via faunal condensation from habi-tat-patch migration over the period of time-averaging, giv-en the proximity of these sites to the reef. Across this‘‘mini-transect,’’ environments constitute a continuum oftaphonomic processes and ecological inputs: it is impres-sive that taphofacies analysis distinguishes these sites aswell as it does, as some might be lumped in other settings,and signatures are, in fact, less distinct than those of moredisparate environments sampled in the larger San BlasArchipelago (Best, 2000; and compare with Bocas del Toro,Best and Kidwell, 2000a).

Here and in most studies (Table 1), each taphonomicvariable is treated independently, using a separate set ofdamage states to categorize the extent of shell modifica-tion (ordinal taphonomic grades sensu Kowalewski et al.,1994; see Pilkey et al., 1969, 1979, and Davies et al., 1989,for earlier variants). A number of taphofacies studies haveemployed a more summary form of ‘‘taphonomicgrades’’ to evaluate individual shells (and see Brandt,1989, for categorization of entire assemblages). By this al-ternative method, taphonomic data are composed of anumber of qualitative categories, each describing the over-all extent of modification when fragmentation, abrasion,surface alteration, bioerosion, encrusting, etc., are jointlytaken into consideration. Frey and Howard (1986; Hen-derson and Frey, 1986) are an early example, using a spe-cific set of features to score individual shells as ‘‘new’’ or‘‘old’’ (2-grade system based on preservation of original col-or and luster, and persistence of ligament and/or perios-tracum); overall sample condition was characterized bythe ratio of new to old valves. Flessa et al. (1993; Meldahlet al., 1997a) devised grading systems for Gulf of Califor-nia mollusks based on larger arrays of variables and amore finely divided spectrum of damage (4 ordinalgrades); ‘‘average grade’’ was used to describe overall sam-ple condition. Each grade is characterized by a suite oftaphonomic characteristics: e.g., 1 � no visible bioerosion,encrustation, or other alteration; 2 � dulling of originalluster, 10% of bioerosion or encrustation on interiorvalve surface, etc.

A major advantage of collecting such ‘‘summary grade’’taphonomic data is that the preservational quality of in-dividual shells and assemblages can be assessed veryquickly. If researchers use a single, common grading sys-tem, this approach also would establish consistencyamong evaluators and facilitate cross-study comparisons(as it has been used in Gulf of California mollusk studies—Flessa et al., 1993; Meldahl et al., 1997a). Two problematicaspects of this type of data are in (1) the potential forgrouping shells (and samples) that are really quite differ-ent (e.g., with one having high levels of encrustation andthe other high boring but each scoring as grade 4), and (2)the potential for having, for example, bio-infestation of onegrade and fine-scale alteration or abrasion of anothergrade on a single shell. Protocols could be established todeal with this latter situation—for example, having eval-uators always score shells by the higher of the two possiblegrades—but this only skirts the underlying ambiguity.Fundamentally, although all variables are considered atonce, this system stresses damage level and subsumes

rank-order importance of variables. In addition, gradesdeveloped on the basis of suites of damage in one area maynot include the range of alteration or the combinations ofstates that can occur in another area (for example, the dis-parate ranges of damage to shells in temperate soft-sedi-ments versus those in reef assemblages); thus, broadercomparisons of data may be difficult.

In San Blas samples, it was not clear, even in retrospect,that taphonomic variables co-varied as regularly—or atleast in the same way—as in Gulf of California materialsuch that this kind of grading system would be effective.Thus, this approach to analysis was not pursued. Ideally,one could conduct a multivariate analysis of pilot data(with shells scored for independent variables) to search forthe most effective and locally relevant ‘‘natural’’ grades,thereby reducing several states/variables into singlescores. This customized grading system would speed sub-sequent data collection. However, it is not clear that itwould reduce some of the inherent ambiguities of the ap-proach, unless the multivariate analysis proved the exis-tence of distinct classes rather than continua in damage.

Another factor that was not pursued explicitly is thechoice and number of variables considered. By inspection(data in Fig. 11), it is clear that the number (and bestchoice) of variables needed to differentiate facies dependson whether the metric is to be presence-absence, high-in-tensity, or full-frequency data. Increasing the number ofvariables and states per variable would increase one’sability to differentiate environments, but this would in-crease the sample size needed and increase between-op-erator error (greater subjectivity in scoring variables asthe damage spectrum is subdivided more finely). Otherstudies have included specialized variables, such as rootetchings and clionid borings (Parsons, 1993; Dent, 1995),cracking and peeling from subaerial exposure (Kowalews-ki et al., 1994), authigenic carbonate precipitation (Callen-der and Powell, 1992), and acquisition of diagenetic colors(from diverse mineral precipitates; Best, 2000), and theseallow most effective differentiation and environmentalanalysis of facies. For the present sensitivity study, themost commonly used variables in taphofacies analysiswere focused upon in order to make the results mostbroadly relevant.

CONCLUSIONS

Using the �8 mm portion of the bivalve death assem-blage as a baseline (interior damage only, fragments in-cluded), qualitative trends in damage levels across thetransect, as well as rank-order importance of variableswithin environments, are robust to most methodologicalvariants tested here. The exceptions are insufficient sam-ple size (to state the obvious) and, for rank order, the use oftarget taxa as proxies.

In contrast, and generally consistent with previousstudies of some of these factors (Davies et al., 1989; Staffand Powell, 1990; Callender and Powell, 1992), the quan-titative damage levels measured for an assemblage are sig-nificantly affected by virtually every operational decisionin taphofacies analysis. These choices include size faction,sieve size, use of total rather than interior-only damage,inclusion/exclusion of fragments or of taxonomically poor-ly resolved specimens, focus on infauna-only, and analytic


metric, in addition to factors of larger effect (sample sizeand narrowing focus to target taxa). Thus, methodologicalartifacts in taphofacies analysis are significant for quan-titative comparisons of environments, and cross-studyconclusions must be made with caution unless the methodis controlled (e.g., see adjustments made in size fractions,variables, metrics, and/or taxa by Callender et al., 1992,and by Kowalewski et al., 1995). To advance the field,some wider standardization of methods is desirable, as isthe archiving of data in sufficiently raw form so that re-sults can be converted.

Some disciplines have ‘‘regulated’’ their methodology torelatively high degrees (e.g., protocols for sieve size, sam-ple size, and relative abundance cut-offs for ‘‘presence’’ inforaminiferal studies; size-fraction protocols for QFL anal-ysis in sedimentary petrology). Self-regulation has prosand cons, among the latter being (a) adopting standardstoo hastily, and (b) overcoming the inertia of entrenchedstandards, even in the face of new information or ques-tions (for some cautionary examples, see Martin, 1999,p.108–109, and Murray, 2000). That said, and using bothpresent results and those of previous studies, some stan-dardization in the treatment of size fractions would clearlybe beneficial: the variance among studies has so far beenhuge (from 1 mm to 30 mm; Table 1), and size fraction hasa significant effect on outcome, especially in settings withmetazoan infestors. Because molluscan size-frequencydistributions vary considerably, it is specifically recom-mended that, at a minimum, assemblages be analyzed insize fractions (e.g., 2–4, 4–8, etc.), rather than standardiz-ing to a particular mesh size. This would not mean that op-erators ‘‘must’’ examine all size fractions in their samples,but that raw data should be partitioned by size fractionswithin the range of sizes they examine; log base-2 size-units comparable to those of sedimentology are a logicalway to do this (as above). To maximize the detection ofpost-mortem damage, analysts could focus on larger shells(e.g., �4 mm) and include at least identifiable fragments,if not all fragments. If streamlining is necessary, high-in-tensity threshold damage values (e.g., % of shells in eachsample falling into the maximum encrustation category)probably are more useful than average-state values orpresence-absence data alone, at least when the targetedenvironments are quite disparate (e.g., as reef gravel v.mud in the present study). Full frequency data on damagestates are obviously superior, if possible.

These methodological strategies should be relevant toanalysis of skeletal remains beyond those of bivalves, andfor fossil as well as modern material, although these spe-cific effects must be tested. The issue of scoring damage toshell exteriors is narrower in relevance and unsettled. Ev-eryone recognizes that most damage to shell exteriors isambiguous in origin, and analyses consistently find thatthis damage equals or exceeds unambiguously post-mor-tem damage to shell interiors. However, despite the desir-ability of interior-only data in principle, this may well beimpractical for much fossil material (but this deservestesting; e.g., Permian bivalves with internal features;Sim-oes and Kowalewski, 1998).

Although it was not the primary purpose of this study,San Blas damage patterns indicate selectivities in tapho-nomic processes (see also most of the studies in Table 1).For example, post-mortem infestation is more intense in

reefal shell gravel and on large shells in all environments(testable via time-lapse experiments); low-level fine-scalealteration is pervasive rather than environment-specific,implying microbial agents (testable via SEM); and frag-ments have the highest damage frequencies, consistentwith them constituting on average older cohorts thanwhole shells in the same assemblage (testable via direct-dating). The differing sensitivities of target taxa to envi-ronmental conditions underscores the imprint of intrinsicbiological factors on the taphonomic signatures of sometaxa more than others, even within a single life-habit orshell type, and in some environments more than others.This is clearly something that can be inferred from fossilmaterial without reference to actualistic experiments, butsimply via analysis of, for example, ternary taphogramplots (cf. Fig. 13C-G). The implication of Figure 13, for ex-ample, that bivalve taxa have more disparate signaturesin mud than in shell gravel is extremely tantalizing: is thisunique to Ulagsuken inlet? Thus, to maximize their valueto paleontologic and sedimentologic/stratigraphic investi-gations, actualistic taphofacies studies need broader tar-gets than single species or genera, moving toward clade-level taphonomy and even taxon-free taphonomy (e.g., in-fauna versus epifauna; microstructure- or mineralogy-specific patterns).

ACKNOWLEDGMENTS

We are grateful to Inez Campbell, Olga Barrio, and Ni-gel Waltho for field assistance; to Rachel Pillar, MeredithDudley, Elizabeth Bassity, and Noel Heim for curatorialassistance in Chicago; and to the Smithsonian TropicalResearch Institute for logistical assistance in Panama. Wealso thank D. Jablonski, R.E. Martin, E.N. Powell, and ananonymous reviewer for helpful reviews. Research sup-ported by NSF-EAR-9628345.

REFERENCES

BACHELET, G., 1990, The choice of a sieving mesh size in the quanti-tative assessment of marine macrobenthos—a necessary compro-mise between aims and constraints: Marine Environmental Re-search, v. 30, p. 21–35.

BEST, M.M.R., 1998, Taxonomic and taphonomic rarefaction: A met-ric for assessing sampling and divergence of taphonomic signa-ture: Canadian Paleontology Conference—VIII, Collingwood, On-tario, Canada, Program and Abstracts v. 8, p. 8.

BEST, M.M.R., 2000, Fates of skeletal carbonate in tropical marine sil-iciclastic and carbonate sediments, Panama: Unpublished Ph.D.Dissertation, University of Chicago, 309 p.

BEST, M.M.R., and KIDWELL, S.M., 2000a, Bivalve taphonomy in trop-ical mixed siliciclastic-carbonate settings: I. Environmental vari-ation in shell condition: Paleobiology, v. 26, p. 80–102.

BEST, M.M.R., and S.M. KIDWELL, 2000b, Bivalve taphonomy in trop-ical mixed siliciclastic-carbonate settings: II. Effect of bivalve lifehabits and shell types: Paleobiology, v. 26, p.103–115.

BEST, M.M.R., KIDWELL, S.M., KU, T., and WALTER, L.M., 1999, Therole of microbial iron reduction in the preservation of skeletal car-bonate: Bivalve taphonomy and porewater geochemistry in tropi-cal siliciclastics vs. carbonates: Geological Society of America, Ab-stracts with Programs v. 31, no. 7, p. A419.

BOEKSCHOTEN, G.T., 1966, Shell borings of sessile epibiontic organ-isms as paleoecological guides (with examples from the Dutch

48 KIDWELL ET AL.

coast): Palaeogeography, Palaeoclimatology, Palaeoecology, v. 2,p. 333–379.

BOSENCE, D.W.J., 1979, Live and dead faunas from coralline algalgravels, County Galway: Palaeontology, v. 22, p. 449–478.

BRANDT, D.S., 1989, Taphonomic grades as a classification for fossil-iferous assemblages and implications for paleoecology: PALAIOS,v. 4, p. 303–309.

CALLENDER, W.R., and POWELL, E.N., 1992, Taphonomic signature ofpetroleum seep assemblages on the Louisiana upper continentalslope: Recognition of autochthonous shell beds in the fossil record:PALAIOS, v. 7, p. 388–408.

CALLENDER, W.R., POWELL, E.N., STAFF, G.M., and DAVIES, D.J.,1992, Distinguishing autochthony, parautochthony and allo-chthony using taphofacies analysis: Can cold seep assemblages bediscriminated from assemblages of the nearshore and continentalshelf?: PALAIOS, v. 7, p. 409–421.

CUTLER, A.H., 1995, Taphonomic implications of shell surface tex-tures in Bahia la Choya, northern Gulf of California: Palaeogeog-raphy, Palaeoclimatology, Palaeoecology, v. 114, p. 219–240.

CUTLER, A.H., and FLESSA, K.W., 1995, Bioerosion, dissolution andprecipitation as taphonomic agents at high and low latitudes:Senckenbergiana maritima, v. 25, p. 115–121.

DAVIES, D.J., POWELL, E.N., and STANTON, R.J., 1989, Taphonomicsignature as a function of environmental process: Shells and shellbeds in a hurricane-influenced inlet on the Texas coast: Palaeo-geography, Palaeoclimatology, Palaeoecology, v. 72, p. 317–356.

DENT, S.R., 1995, A taphofacies model of the recent South Floridacontinental shelf: A new perspective for a classic, exposed carbon-ate environment.: Unpublished Ph.D. Dissertation, University ofCincinnati, Cincinnati, 204 p.

FEIGE, A., and FURSICH, F.T., 1991, Taphonomy of the Recent mol-luscs of Bahia la Choya (Gulf of California, Sonora, Mexico): Zit-teliana (Munich), v. 18, p. 89–113.

FLESSA, K.W., CUTLER, A.H., and MELDAHL, K.H., 1993, Time and ta-phonomy: quantitative estimates of time-averaging and strati-graphic disorder in a shallow marine habitat: Paleobiology, v. 19,p. 266–286.

FLUGEL, E., 1982, Microfacies analysis of limestones: Springer-Ver-lag, Berlin, 633 p.

FREY, R.W., and HOWARD, J.D., 1986, Taphonomic characteristics ofoffshore mollusk shells, Sapelo Island, Georgia: Tulane Studies inGeology and Paleontology, v. 19, p. 51–61.

FURSICH, F.T., and FLESSA, K.W., 1987, Taphonomy of tidal flat mol-luscs in the Northern Gulf of California: Paleoenvironmentalanalysis despite the perils of preservation: PALAIOS, v. 2, p. 543–559.

GASTALDO, R.A., DOUGLASS, D.P., and MCCARROLL, S.M., 1987, Ori-gin, characteristics, and provenance of plant macrodetritus in aHolocene crevasse splay, Mobile Delta, Alabama: PALAIOS, v. 2,p. 229–240.

HENDERSON, S.W., and FREY, R.W., 1986, Taphonomic redistributionof mollusk shells in a tidal inlet channel, Sapelo Island, Georgia:PALAIOS, v. 1, p. 3–16.

JOHNSON, R.G., 1960, Models and methods for analysis of the mode offormation of fossil assemblages: Geological Society of AmericaBulletin, v. 71, p. 1075–1086.

KOWALEWSKI, M., FLESSA, K.W., and AGGEN, J.A., 1994, Taphofaciesanalysis of Recent shelly cheniers (beach ridges), northeasternBaja California, Mexico: Facies, v. 41, p. 209–242.

KOWALEWSKI, M., FLESSA, K.W., and HALLMAN, D.P., 1995, Ternarytaphograms: Triangular diagrams applied to taphonomic analy-sis: PALAIOS, v. 10, p. 478–483.

KU, T.C.W., and WALTER, L.M., 1998, Pore water geochemistry ofshallow marine carbonates vs tropical siliciclastics: Carbon, sulfurand iron systematics and implications for carbonate preservation:

Geological Society of America Abstracts with Programs, v. 30, p.A374.

LLEWELLYN, G., and MESSING, C.G., 1993, Compositional and tapho-nomic variations in modern crinoid-rich sediments from the deep-water margin of a carbonate bank: PALAIOS, v. 8, p. 554–573.

MARTIN, R.E., 1999, Taphonomy: A Process Approach: CambridgeUniversity Press, Cambridge, 508 p.

MARTIN, R.E., and LIDDELL, W.D., 1991, Taphonomy of foraminiferain modern carbonate environments: implications for the formationof foraminiferal assemblages: in DONOVAN, S.K., ed., The Process-es of Fossilization: Belhaven Press, London, p. 170–193.

MELDAHL, K.H., and FLESSA, K.W., 1990, Taphonomic pathways andcomparative biofacies and taphofacies in a Recent intertidal/shal-low shelf environment: Lethaia, v. 23, p. 43–60.

MELDAHL, K.H., FLESSA, K.W., and CUTLER, A.H., 1997a, Time-aver-aging and postmortem skeletal survival in benthic fossil assem-blages: Quantitative comparisons among Holocene environments:Paleobiology, v. 23, p. 207–229.

MELDAHL, K.H., YAJIMOVICH, O.G., EMPEDOCLES, C.D., GUSTAFSON,C.S., HIDALGO, M.M., and REARDON, T.W., 1997b, Holocene sedi-ments and molluscan faunas of Bahıa Concepcıon: A modern an-alog to Neogene rift basins of the Gulf of California: Geological So-ciety of America Special Paper 318, p. 39–56.

MILLER, A.I., LLEWELLYN, G., PARSONS, K.M., CUMMINS, H., BOARD-MAN, M.R., GREENSTEIN, B.J., and JACOBS, D.K., 1992, Effect ofHurricane Hugo on molluscan skeletal distributions, Salt RiverBay, St. Croix, U.S. Virgin Islands: Geology, v. 20, p. 23–26.

MURRAY, J.W., 2000, The enigma of the continued use of total assem-blages in ecological studies of benthic Foraminifera: Journal of Fo-raminiferal Resarch, v. 30, p. 244–245.

NEBELSICK, J.H., 1999, Taphonomy of Clypeaster fragments: Preser-vation and taphofacies: Lethaia, v. 32, p. 241–252.

PANDOLFI, J.M., and GREENSTEIN, B.J., 1997, Taphonomic alterationof reef corals: Effects of reef environment and coral growth form. I.The Great Barrier Reef: PALAIOS, v. 12, p. 27–42.

PARSONS, K.M., 1989, Taphonomy as an indicator of environment:Smuggler’s Cove, St. Croix, U.S.V.I.: in HUBBARD, D.K., ed., Ter-restrial and marine ecology of St. Croix, U.S. Virgin Islands: WestIndies Laboratory Special Publication 8, p. 135–143.

PARSONS, K.M., 1993, Taphonomic attributes of mollusks as predic-tors of environment of deposition in modern carbonate systems:northeastern Caribbean: Unpublished Ph.D. Thesis, University ofRochester, Rochester, 434 p.

PARSONS, K.M., and BRETT, C.E., 1991, Taphonomic processes and bi-ases in modern marine environments: An actualistic perspectiveon fossil assemblage preservation: in DONOVAN, S.K., ed., The Pro-cesses of Fossilization: Columbia University Press, New York, p.22–65.

PERRY, C.T., 2000, Factors controlling sediment preservation on anorth Jamaica fringing reef: a process-eased approach to microfa-cies analysis: Journal of Sedimentary Research, v. 70, p. 633–648.

PILKEY, O.H., 1964, The size distribution and mineralogy of the car-bonate fraction of United States south Atlantic shelf and upperslope sediments: Marine Geology, v. 2, p. 121–136.

PILKEY, O.H., BLACKWELDER, B.W., DOYLE, L.J., ESTES, E., and TER-LECKY, P.M., 1969, Aspects of carbonate sedimentation on the At-lantic continental shelf off the southern United States: Journal ofSedimentary Petrology, v. 39, p. 744–768.

PILKEY, O.H., FIERMAN, E.I., and TRUMBULL, J.V.A., 1979, Relation-ship between physical condition of the carbonate fraction and sed-iment environments: Northern Puerto Rico Shelf: SedimentaryGeology, v. 24, p. 283–290.

RAUP, D.M., 1991, The future of analytical paleobiology: in GILINSKY,N.L., and SIGNOR, P.W., eds., Analytical Paleobiology: ShortCourse Notes, Paleontological Society, Knoxville, p. 207–216.

REISH, D.J., 1959, A discussion of the importance of the screen size in


washing quantitative marine bottom samples: Ecology, v. 40, p.307–309.

SIMOES, M.G., and KOWALEWSKI, M., 1998, Shell beds as paleoecolog-ical puzzles: A case study from the Upper Permian of the ParanaBasin, Brazil: Facies, v. 38, p. 175–196.

SMITH, A.M., and NELSON, C.S., 1994, Selectivity in sea-floor process-es: Taphonomy of bryozoans: in HAYWARD, P.J., RYLAND, J.S., andTAYLOR, P.D., eds., Biology and palaeobiology of bryozoans: Pro-ceedings 9th International Bryozoology Conference 1992: Olsenand Olsen, Fredensborg, p. 177–180.

SPEYER, S.E., and BRETT, C.E., 1986, Trilobite taphonomy andMiddleDevonian taphofacies: PALAIOS, v. 1, p. 312–327.

STAFF, G.M, and POWELL, E.N., 1990, Local variability of taphonomicattributes in a parautochthonous assemblage: Can taphonomicsignature distinguish a heterogeneous environment?: Journal ofPaleontology, v. 64, p. 648–658.

ACCEPTED OCTOBER 3, 2000

50 KIDWELL ET AL.

APPENDIX.Ranked % abundance of species in each size fraction, and in totalsample for sites 3 and 7, for all identifiable specimens (whole andfragments). Target taxa in boldface. Number of specimens indicat-ed by n; these specimens are from splits of the total available sam-ple for some size fractions (split indicated in parens). At site 3, thesmall number of specimens available in the 4–8 mm size fractionof the original sample were supplemented with specimens fromtwo replicate samples near the same site (samples Ds116 andDs117).

Site 3 Raw Data

% �8 mm size fraction

20.4112.2410.208.168.164.084.084.084.082.042.042.04

Laevicardium cf. L. pictumLaevicardium laevigatumChlamys muscosusPitar cf. P. arestusChlamys sp. 129Ostrea sp.Microcardium cf. M. perambileMacominae cf. Macoma tageliformisDosinia concentricaPinna attenuataAmericardia cf. A. mediaTagelus cf. T plebeius

2.042.042.042.042.042.042.042.042.04

Pecten ss. cf. P. chazelisi?Amusium sp.Pitar sp. 70Codakia sp. 74Cyclinella tenuisthraciid cf. Cyathodonta semirugosaLima limamactrid cf. MactraTellina (Angulus) sp. 139

n � 49 (entire sample used)

% 4–8 mm size fraction

27.2718.189.099.099.099.099.099.09

Nuculana acutaLaevicardium pictumLaevicardium laevigatumCorbula caribaeaCodakia (Ctena) sp.Dosinia concentricapectinidtellinid



19.7713.9512.796.985.815.814.653.493.492.332.332.332.332.331.161.161.161.161.16

Pitar sp. 70Laevicardium sp.Chlamys muscosuspectinidNuculana acutaChlamys sp. 129Corbula caribaeaCodakia (Ctena) sp.Barbatia sp. 141Tellina (Angulus) probinaTellina (Angulus) euvitreaPapyridea cf. P. soleniformisPitar cf. P. fulminatusCorbula (Varicorbula) sp. indet.Tellina (Angulus) sp.Trachycardium cf. T. muricatumAnadara tricenicostata?Americardia cf. A. mediaCardiomya 2 spp

1.161.161.161.161.16

Nuculana acutathraciid cf. Cyathodonta semirugosaLepton sp. indet.Nucula dalmasiarcid

n � 86 (1/8 of sample used)

APPENDIX.Continued.

Supplement from Sample DS116

% 4–8 mm

12.7711.709.577.456.386.385.325.324.263.192.132.13

Laevicardium sp.Pitar sp. 70Chlamys sp. 129Chlamys muscosusChlamys groupLaevicardium cf. L. pictumAmericardia cf. A. mediapectinidPapyridea cf. P. soleniformisthraciid cf. Cyathodonta semirugosaChama congregataCorbula caribaea

2.132.132.132.132.131.061.061.061.061.061.061.061.061.061.061.061.06

mactrid cf. MactrapteriidDiplodonta cf. D. (Phlyctiderma) notataarcidtellinidNuculana acutaOstrea sp.Anadara tricenicostata?Tellina (Angulus) probinaArca cf. A. imbricataDosinia concentricaLaevicardium pictumAnadara sp. 119Plicatula sp. 134Lyropecten sp.Tellina cf. T. (Tellinella)mytilid


Supplement from Sample Ds117

% 4–8 mm

22.8611.437.865.715.003.573.573.572.862.862.862.862.142.141.431.431.431.431.431.431.43

Pitar sp. 70Laevicardium sp.Chlamys muscosusChlamys sp. 129thraciid cf. Cyathodonta semirugosaPapyridea cf. P. soleniformismactrid cf. MactraTellina (Merisca) sp.Laevicardium cf. L. pictumAmericardia cf. A. mediaLaevicardium pictumpectinidCorbula caribaeaCodakia (Ctena) sp.Nuculana acutaChama congregataLima scabra tenerapteriidLyropecten sp.arcidchamid

0.710.710.710.710.710.710.710.710.710.710.710.710.710.710.71

Ostrea sp.Tellina (Angulus) probinaChama macerophyllaArca cf. A. imbricataCorbula (Juliacorbula) aequivalvisIsognomon sp.Pitar cf. P. fulminatusArca cf. A. zebraBarbatia sp. 118Anadara sp. 119Chione paphiaChlamys sp. 142Periglypta cf. P. listeriDiplodonta cf. D. (Phylctiderma) notatatellinid



APPENDIX.Continued.

% Total data for site 3

15.8010.368.81

Pitar sp. 70Laevicardium sp.Chlamys muscosus

5.184.404.153.373.113.112.852.852.592.592.071.811.811.811.551.301.301.041.041.041.04

Laevicardium cf. L. pictumChlamys sp. 129pectinidpteriidNuculana acutathraciid cf. Cyathodonta semirugosaAmericardia cf. A. mediaPapyridea cf. P. soleniformisCorbula caribaeaOstrea sp.mactrid cf. MactraLaevicardium laevigatumCodakia (Ctena) sp.Laevicardium pictumChlamys groupTellina (Merisca) sp.arcidChama congregataPitar cf. P. arestusTellina (Angulus) probinaDosinia concentrica

1.040.780.780.780.520.520.520.520.520.520.520.520.520.520.260.260.260.260.260.260.26

tellinidPitar cf. P. fulminatusBarbatia sp. 141Lyropecten sp.Anadara tricenicostata?Microcardium cf. M. perambileLima scabra teneraTellina (Angulus) euvitreaArca cf. A. imbricataMacominae cf. Macoma tageliformisAnadara sp. 119Diplodenta cf. D. (Phlytiderma) notataCorbula (Varicorbula) sp. indet.chamidPinna attenuataTellina (Angulus) sp.Trachycardium cf. T. muricatumChama macerophyllaTagelus cf. T. plebeiusCardiomya 2 spp.Pecten ss. cf. P. chazelisi?

0.260.260.260.260.260.260.260.260.260.260.260.260.260.260.260.260.260.26

Amusium sp.Corbula (Juliacorbula) aequivalvisCodakia sp. 74Isognomon sp.Cyclinella tenuisLima limaArca cf. A. zebraBarbatia sp. 118Chione paphiaPlicatula sp. 134Tellina (Angulus) sp. 139Chlamys sp. 142Periglypta cf. P. listeriTellina cf. T. (Tellinella)mytilidDiplodonta cf. D. (Phylctiderma) notataLepton sp. indet.Nucula dalmasi

n � 386

APPENDIX.Continued.

Site 7 Raw Data

% �8 mm size fraction

8.888.227.897.896.586.586.094.934.613.782.802.632.472.142.141.811.641.481.48

Americardia cf. A. mediapectinidLaevicardium laevigatumarcidPapyridea cf. P. soleniformisChlamys sp. 129Chama macerophyllamactrid cf. MactraChlamys muscosusLaevicardium sp.Lima scabraChlamys groupChama congregataPitar sp. 70pteriidAnadara notabilismactridLaevicardium cf. L. pictumArca cf. A. imbricata

1.321.321.150.990.990.820.660.490.490.490.490.490.490.330.330.330.330.330.330.330.33

Lima scabra teneraPitar cf. P. fulminatusArca sp. 117Barbatia sp. 32Barbatia sp. 115chamidMacoma sp. cf. M. tageliformisPinna attenuataAnadara (Cunearca) brasilianaMacominae cf. Macoma tageliformisArca cf. A. zebraAnadara cf. A. notabilisPododesma rudisOstrea sp.Pitar cf. P. arestusMicrocardium cf. M. perambileLithophaga sp.Cyclinella tenuisCodakia (Ctena) orbiculataBarbatia sp. 118limid

0.160.160.160.160.160.160.160.160.160.160.160.160.160.160.160.160.160.160.16

Trachycardium cf. T. muricatumCorbula caribaeaDiplodonta sp. cf. D. punctataTellina (Eurytellina) nitensTellina (Arcopagia) faustaIsognomon sp.Callista (Macrocallista) sp. cf. C. maculatalucinidLopha fronsTagelus cf. T. plebeiusCodakia (Ctena) sp.Laevicardium pictumAnadara sp. 119Tellina (Merisca) sp.cardiidPlicatula sp. 134Chlamys sp. 135Tellina (Angulus) sp. 139tellinid


52 KIDWELL ET AL.

APPENDIX.Continued.


6.256.256.256.255553.753.753.752.52.52.52.52.52.51.251.251.251.25

Chama congregataLima scabra teneraArca cf. A. imbricatapteriidPinna attenuataBarbatia sp. 32Laevicardium cf. L. laevigatumTrachycardium cf. T. muricatumCorbula caribaeaAnadara sp. 119Lithophaga sp.Lopha fronsBarbatia sp. 115Lima cf. L. pellucidaBarbatia sp. 141Periglypta cf. P. listeriLaevicardium laevigatumChione cancellata?Ostrea sp.Chlamys group

1.251.251.251.251.251.251.251.251.251.251.251.251.251.251.251.251.251.251.251.251.25

Americardia cf. A. mediaChama macerophyllaPitar sp. 70Isognomon sp.Lima scabraPapyridea cf. P. soleniformisAnadara notabilisPitar cf. P. fulminatusLima limaCodakia (Ctena) sp.mactrid cf. MactraSolecurtus sp. cf. S. cumingianusChlamys muscosusChlamys sp. 129cardiidTellina (Angulus) sp. 139Barbatia sp. 140Chlamys sp. 142Lyropecten sp.arcidpectinid

1.251.25

Laevicardium sp.tellinid



31.6512.667.597.597.596.335.065.062.532.532.531.271.271.271.271.27

Pitar sp. 70Laevicardium sp.Papyridea cf. P. soleniformismactrid cf. MactraChlamys muscosusCorbula caribaeaCodakia (Ctena) sp.Barbatia sp. 141Barbatia sp. 115Crassinella sp. 149arcidTellina (Scisulla) similisBarbatia sp. 118pteriidChlamys sp. 129Lyropecten sp.

1.271.27

pectinidchamid


APPENDIX.Continued.

% Total data for site 7

7.176.786.656.396.135.485.084.954.824.564.432.612.482.352.221.83

Americardia cf. A. mediapectinidarcidLaevicardium laevigatumPapyridea cf. P. soleniformisChlamys sp. 129Pitar sp. 70Chama macerophyllamactrid cf. MactraChlamys muscosusLaevicardium sp.Chama congregatapteriidLima scabraChlamys groupArca cf. A. imbricata

1.691.561.301.301.301.171.171.170.910.910.780.780.650.520.520.520.520.520.390.390.39

Lima scabra teneraAnadara notabilisBarbatia sp. 32Barbatia sp. 115mactridCorbula caribaeaLaevicardium cf. L. pictumPitar cf. P. fulminatusPinna attenuataArca sp. 117Codakia (Ctena) sp.Barbatia sp. 141chamidTrachycardium cf. T. muricatumLithophaga sp.Laevicardium cf. L. laevigatumAnadara sp. 119Macoma sp. cf. M. tageliformisOstrea sp.Anadara (Cunearca) brasilianaLopha frons

0.390.390.390.390.390.390.260.260.260.260.260.260.260.260.260.260.260.260.130.130.13

Macominae cf. Macoma tageliformisArca cf. A. zebraBarbatia sp. 118Anadara cf. A. notabilisPododesma rudistellinidPitar cf. P. arestusMicrocardium cf. M. perambileIsognomon sp.Cyclinella tenuisCodakia (Ctena) orbiculataLima cf. L. pellucidacardiidTelina (Angulus) sp. 139Lyropecten sp.Periglypta cf. P. listeriCrassinella sp. 149limidChione cancellata?Diplodonta sp. cf. D. punctataTellina (Eurytellina) nitens

0.130.130.130.130.130.130.130.130.130.130.130.130.13

Tellina (Arcopagia) faustaCallista (Macrocallista) sp. cf. C. maculatalucinidTagelus cf. T. plebeiusLima limaTellina (Scisulla) similisLaevicardium pictumSolecurtus sp. cf. S. cumingianusTellina (Merisca) sp.Plicatula sp. 134Chlamys sp. 135Barbatia sp. 140Chlamys sp. 142

n � 767

Documents

Sensitivity of Taphonomic Signatures to Sample Size, Sieve ...geosci.uchicago.edu/pdfs/kidwell/Kidwell-2001-Sensitivity.pdf · Sensitivity of Taphonomic Signatures to Sample Size,