Bioturbation in the Deep Ocean Robert A

Limno!. Oceanogr., 37(1), 1992, 90-104 © 1992, by the American Society of Limnology and Oceanography, Inc. Experimental tests for particle size-dependent bioturbation in the deep ocean Robert A. Wheatcroft' School of Oceanography, WB- 10, University of Washington, Seattle 98195

Abstract The potential for particle size-dependent bioturbation rates was experimentally tested at 1,240 m in the Santa Catalina Basin (eastern Pacific). Spherical glass bead tracers in five size classes (8- 16, 17-31, 32-62, 63-125, and 126-420 m) were spread over the sediment surface and tube cored 997 d later. Downcore concentrations of glass beads were enumerated in each of the five size categories and Page's L-test was used to test the null hypothesis of equal vertical penetration of all size classes of tracer. In all cores the null hypothesis was rejected; finer tracers penetrated deeper into the sediment. In two of the three cores, vertical biodiffusivities were computed from concen- tration profiles of downcore tracers. These also showed size dependence, with biodiffusivities ranging from 1 cm2 yr' for the 8-1 6-pm fraction to 0.1 cm2 yr' for the 1 25-420-pm size class. These data demonstrate that vertical bioturbation rates are particle size-dependent in Santa Catalina Basin. The likely cause is preferential ingestion and downward transport of fine particles by deposit feeders.

Nearly all particles that reach the floor ofplacement of sediment grains also destroys the ocean are displaced several times by an- or blurs sedimentary signals (Wheatcroft imals before they are buried to become part1990). Therefore, a better understanding of of the sedimentary record. This mixing ofbioturbation, especially its rates, styles, and sediment or bioturbation has profound ef-variability would benefit a wide range of fects on a wide range of phenomena. Ratesaquatic disciplines. of organic matter decomposition, for ex- In the past scientists have approached ample, as well as the dissolution of nearlybioturbation from one of two directions. On all sedimentary constituents (e.g. CaCO3 andthe one hand, benthic ecologists (e.g. Rhoads Si02) are markedly influenced by the rate 1967) have typically focused on a single spe- of sediment mixing (Berner 1980). Similar-cies (usually from a shallow-water, temper- ly, the distribution of solid and liquid phaseate site) and observed rates and styles of nutrients is mediated by bioturbation, im-sediment reworking under a range of con- plying that there are likely to be strong feed- ditions (e.g. changing temperature). Follow- backs between the style and rate of mixinging nearly two decades of such autecologic and the distribution of infauna. The dis-studies there is a large body of literature (summarized by Thayer 1983; Wheatcroft et al. 1990) demonstrating that sediment 'Present address: Applied Ocean Physics and En- displacement is tremendously complex. gineering Department, Woods Hole Oceanographic In- stitution, Woods Hole, Massachusetts 02543. Deposit feeding and hence particle displace- ment rates have been shown to be tem- Acknowledgments I thank C. Smith and P. Jumars for aid during the perature- (Rhoads 1967), particle-shape- initiation of this research and dialogue throughout the (Whitlatch 1974), and particle-size- (Powell entire project. This work could not have been accom- 1977) dependent. Moreover, there seem to plished without the aid and expertise of the master and be ontogenetic changes in a species' particle crew of the RV Atlantis II, especially the Alvin group. selectivity, as well as interactions between Help with on-deck processing of cores was provided by D. Penry, R. Pope, L. Self, and a host of others. different environmental cues, that further Previous versions of the manuscript benefited from complicate mixing. comments by J. Bourgeois, C. A. Butman, P. Jumars, The other approach to bioturbation has and A. Nowell. I thank all of these people. been taken primarily by geochemists who This work was supported by Office of Naval Re- search contract N000l4-90-J- 1078 to P. Jumars andrequire a community-wide estimate of the A. Nowell, and is contribution 1889 from the School sediment mixing rate to characterize the role of Oceanography, University of Washington. of bioturbation as a mass transfer mecha- 90 Particle size-dependent bioturbation 91 nism. This goal is achieved by making var- to be associated with finer particles. Coch- ious simplifying models of bioturbation.ran (1985) also found a disagreement be- Currently the most widely used model istween mixing coefficients calculated from based on an analogy with standard Fickianthe profiles of the bomb-produced radioiso- diffusion (Guinasso and Schink 1975;topes, '37Cs and 239'240Pu vs. 210Pb. He sug- Wheatcroft et al. 1990). Community-widegested differential mixing of different-sized mixing rates are obtained by measuring theparticles, but a!so stated that selective mo- vertical concentration profiles of variousbility of the radionuclides cou!d not be ruled tracers and fitting theoretical profiles to theout. An additional independent line of ev- observed gradients. The best-fit curves con-idence that suggests mixing is size-depen- tain various parameter values, one of whichdent comes from the studies of CaCO3 ages is biodiffusivity (Db). This coefficient rep-in the mixing layer. Berger and Johnson resents all of the myriad sediment-displac-(1978) hypothesized that discrepancies be- ing activities of the benthos at a given site,tween 14C- and t8O-derived dates might re- and thus, much of the complexity observedsult from greater mixing of the bulk car- by benthic ecologists is incorporated intobonate fraction relative to the large Db. An untested and extremely importantforaminiferans from which the oxygen rec- question is whether that complexity willord was derived. manifest itself in the community-wide mea- An explicit test for size dependency in sures of mixing (e.g. biodiffusivities) in adeep-sea mixing was made by Ruddiman et systematic and predictable manner. That is,al. (1980). They measured the vertical dis- is it necessary or even possible to explicitlypersion of different-sized ash particles and, incorporate biology into the biodiffusionon the basis of visual comparisons of the coefficient? concentration profiles, cautiously conclud- This question is addressed by focusing oned that mixing was not size-dependent. Since an issue that has a long history (consideringthen, their data have been statistically rean- the youthfulness of this field) in studies ofalyzed by Wheatcroft and Jumars (1987), bioturbation: the question of particle sizewho found that there was indeed greater dependence in deep-sea mixing rates. Asmixing of the fine fraction. Although the previously mentioned, several autecologicresults of their reanalysis and the inferential laboratory studies have demonstrated pref-evidence discussed above qualitatively sug- erential ingestion of fine particles by depositgest mixing rates are size-dependent in the feeders (see Taghon 1989). In addition, manydeep sea, the magnitude of the relationship head-down deposit feeders preferentiallyis unclear. My primary goal here is to de- ingest fine-grained sediments resulting inscribe results of an a priori test for particle biogenic graded bedding, whereby one pass- size-dependent bioturbation in the deep es from fine to coarse grains moving downocean. A secondary goal is to discuss animal into the sediment. These observations,activities that displace sediment within the however, have been made on a limited sub- context of their potential for particle size- set of animals, mainly temperate, shallow-dependent mixing. water deposit feeders in sandy environ- ments. To what extent these shallow-waterBackground observations apply to deep-sea bioturbation All animal activities that displace sedi- rates is presently unknown. ment have the potential to produce size- There is some inferential evidence thatdependent transport rates. The direction of deep-sea mixing is indeed size-dependent.bias (i.e. preference for fine or coarse par- Severalrecentradioisotopicstudiesticles), however, often differs within and be- (DeMaster and Cochran 1982; Stordal etal. tween activities; thus, there is the possibility 1985) have reported disparities betweenthat the integrated community-wide effect mixing coefficients derived from various ra-could be neutral. That is why tests for par- dionuclides (e.g. 210Pb vs. 325i or 210Pb vs. ticle size-dependent mixing must be made 239'240Pu). In both cases the radionuc!ide that in the field, where the summed activities of yielded the higher mixing rates was inferredall animals in a given community can be 92 Wheatcrofi assessed. Size-dependent mixing rates (par- gravitationally forced vertical migration oc- ticle migration, sensu Williams and Khancur. The first is downward migration of fine 1973) can be produced actively or passively. particles (termed percolation, Williams and Passive particle migration often occurs inKhan 1973), due to the greater tendency of industrial mixing (Williams and Khan 1973;the fines to settle into voids created in the Stephens and Bridgwater 1978), where it ismoving particle matrix. This process in turn typically caused by spatial heterogeneity inresults in a second upward migration of the local strain rate and resultant porositycoarse particles because the weight of coarse field. Some animal activities (e.g. burrow- particles causes an increase in pressure be- ing) may cause passive particle migration.low it that stops the particle from moving Most animal activities contain an activedownward. As particles are jostled, more component, however, whereby the animalfines fall under the coarse particles and are preferentially displaces particles of a givenlocked into position (Williams and Khan size. 1973). Thus, the coarse particles migrate During tube building, many invertebratesupward. Elements of the coarse-particle mi- (especially polychaetes) exhibit scrupulousgration mechanism have been used to ex- selection of particles based on various at-plain the tendency for Mn nodules to re- tributes (e.g. shape, specific gravity, or size). main at the sediment surface (Sanderson Field observations of the polychaete Owe- 1985) and the observed upward migration niafusiformis have revealed concentrationsof very coarse glacial debris in deep-sea sed- of tabular grains of hornblende in its tubeiments (McCave 1988). The difference in at least 25-fold over ambient levels (Fagerparticle sizes in both systems was large 1964). Similarly, in laboratory tests the same(- 100-1,000-fold), however, so the likeli- species displayed a 100% preference for an- hood of passive particle migration in more gular over spherical glass particles (Self andequitably sized deposits is unclear. Jumars 1988). Whitlatch and Weinberg Several independent lines of reasoning (1982) observed strong selection of largersuggest that if there is any passive particle particles by the tube-building pectinaridmigration due to locomotion it should be polychaete Cistenides gouldii. Other speciestoward greater downward transport of the of tube-building, shallow-water poly- coarse fraction. First, sediments in nature chaetes, however, exhibit a weak preferenceare not subject to the high-frequency vibra- toward finer particles (Self and Jumarstions experienced in industrial mixing. Thus, 1988). Thus, the integrated effect of variouslocalized high shear zones that result in en- tube builders on particle size-dependenthanced particle migration rates (Williams mixing will likely depend on the speciesand Khan 1973) are not likely to form. More composition of that community and is dif-importantly, most marine sediments are not ficult to predict a priori. Until it can becohesionless. Rather, they are complex shown otherwise, in the equitable com-mixtures of wet, cohesive, organically bound munities characteristic of the deep sea, tubeparticles. Thus, downward percolation of building will be considered to have a neutralfine particles is unlikely in aqueous sedi- effect on particle migration. Test-buildingments where the gravitational force on fine protistans (e.g. foraminiferans and xeno-particles could be balanced by cohesive or phyophores), however, display strong selec-adhesive forces. Only coarse (but not too tion for coarse particles (pers. obs.). In manycoarse; Sanderson 1985; McCave 1988) deep-sea settings protistans dominate mac-particles are likely to display preferential rofaunal biomass; therefore one might ex-downward migration due to burrowing, since pect greater residence times of coarse par- for them gravity overcomes the cohesive- ticles in the surface mixing layer. adhesive bonds. This effect was demon- Animal locomotion (e.g. crawling andstrated by Clifton (1984) who observed that burrowing) has the potential to produce pas- the burrowing mechanism of a nephytid sive, size-dependent vertical mixing. In co- polychaete fluidized the sediment enough to hesionless, polydisperse systems of particlesallow the coarser and denser particles to set- undergoing industrial mixing, two types oftle lower in the deposit. That sediments in Particle size-dependent bioturbation 93 general may be fluidized enough to allowstudies of particle selection, as well as a priori coarse-particle migration or at least reori-tests of size selectivity (e.g. Taghon 1982). entation is tentatively supported by ob- The most frequent finding is of preferential ervations by Chernow et al. (1986), whoselection for smaller particles (see Taghon explained the minimal influence of biotur-1989). More recent experimental studies bation on magnetic fabrics by invoking sim- with glass beads of varying sizes and specific ilar fluidization. Thus, if burrowing has anygravities and animals from several phyla effect on particle migration it may be towardsupport earlier findings of peak selectivity preferential downward movement of thefor small particles (10-20 jim) (Self and Ju- coarse fraction. mars 1988). Preferential downward transport of coarse There are exceptions to the pattern of particles due to the activities of surface-greater selection of the fine fraction, where- crawling animals is also likely. There is aby coarser particles are ingested preferen- higher probability for crawlers to contacttially (e.g. Whitlatch 1974; Fenchel et al. large particles, due solely to geometrical1975; Whitlatch and Weinberg 1982). In considerations. Thus, larger particles aresome instances, enhanced selection of the more likely to be pushed into open tubescoarse fraction occurs when the majority of and burrows where they are transportedfood is associated with that size class (e.g. downward (Wheatcroft et al. 1990). Again,diatoms, Fenchel et al. 1975). In other cases particles that are very large relative to am- it is probably caused by morphological con- bient sediments will be avoided completelystraints of the specific feeding appendage by crawling animals. (Whitlatch and Weinberg 1982). In any case, Of all animal activities, deposit feedingselection for coarse particles is the excep- has been postulated to be the dominant sed-tion, and deposit feeding normally will re- iment transport mechanism (Thayer 1983;sult in enhanced displacement of fines. Wheatcroft et al. 1990). Thus, if preferential A higher displacement frequency (or mixing of particles occurs, it is most likelyshorter rest period in the nomenclature of due to this activity. The direction of particleWheatcroft et al. 1990) for fine particles does size preference is also most clear for thisnot necessarily mean that they will pene- animal activity. Deposit feeders on the wholetrate farther into the sediments in a given actively select and thus displace finer par-unit of time because, at least in shallow- ticles. The reasons are both behavioral andwater environments, many deposit feeders morphological. First, the long-standing ob-(both surface and subsurface types) egest servation that, at a given water depth, de-material on the sediment surface. The ram- posit-feeder standing stock increases withifications of such mixing have been ex- decreasing particle size of the deposit sug-plored, via first-order Markov models, by gests that more food is to be found in finerJumars et al. (1981), who found that under sediments (Sanders 1960). With decreasingsome circumstances fine-grained particles particle size, the ratio of grain surface areawill be preferentially kept in the surface to grain volume increases, providing the po-mixing layer. Thus it is conceivable that tential for more food per unit of grain vol-coarse particles will penetrate into the sed- ume (assuming that most food is surfaceiment at faster rates. associated). Quantitative measures of the relationship between sediment surface areaStudy site and microbial abundance and organic C The study site (33°12'N, 1 l8°30'W) is at (DeFlaun and Mayer 1983) support this ex- 1,240 m on the floor of Santa Catalina Basin planation down to particle sizes of - 10 m.(SCB), a bathyal basin in the California bor- The above observations were put into anderland tectonic province. SCB sediments optimal foraging context by Taghon et al.consist of clayey-silts with a mean disag- (1978) who predicted that deposit feedersgregated diameter of 4 tm. The coarse should, under most conditions, preferen-fraction (>62 m) is composed of radiolar- tially ingest small particles. ians, foraminiferans, quartz, and lithic frag- There have been numerous empiricalments. Bottom water is sufficiently well ox- 94 Wheatcrofi ygenated (0.4 ml liter-') and organic C fluxulation by Alvin. At the sea floor, the bead high enough to support an abundant andspreaders were removed from the basket in reasonably diverse bottom fauna (Jumarsan upright position, held50 cm off the 1976; Smith and Hamilton 1983; C. Smithbottom and upended. Continual shaking by etal. 1986). the manipulator arm for 1-2 mm dispersed The macrofauna is dominated by mem- 200 ml of beads onto 1 m2 of seafloor. bers of the polychaete families Paraonidae,The use of more sophisticated devices that Cirratulidae, and Cossuridae (Jumars 1976;purportedly yield monolayers of tracer par- C. Smith et al. 1986). There is very littleticles on the sediment surface is unnecessary natural history information on these groups,because only the vertical gradient of particle especially from the deep sea Limited ob-concentration is of interest. Moreover, such servations on shallow-water congenericsdevices undoubtedly disturb the near-sur- suggest that many of these worms are de-face infauna and epifauna to a much greater posit feeders, which studies of gut contentsextent than the bead spreader described here. from SCB animals support (Penry and Ju- Treatments of two different durations mars 1990). The megafauna is stronglywere deployed. A long-term experiment was dominated (>99%) by the epifaunal brittleinitiated on 5 February 1985 (dive 1,512) star Ophiophthalmus normani (Smith andand cored 997 d later on 30 October 1987 Hamilton 1983). This animal is probably(dive 1,937) with 6.7-cm-i.d.-diameter an omnivore, capable of deposit and sus-acrylic tube corers. A second, short-term pension feeding, as well as preying on var-treatment was introduced on 12 December ious meiofauna. It is highly mobile and1986 (dive 1,779) and cored 319 d later on probably responsible for the high near-sur-31 October 1987 (dive 1,938). A control face horizontal bioturbation rates measuredexperiment, consisting of bead spreading in this basin (Wheatcroft 1991). Other epi-followed by immediate coring, was per- benthic megafauna consists of two speciesformed on 10 November 1987 (dive 1,949). of gastropod and various echinoderms (e.g.Details of the navigation system used to re- holothuroids, asteroids, and ophiuroids)locate and core the small-scale treatments (Smith and Hamilton 1983). Although it hason the deep-sea floor are given elsewhere not been sampled in box cores, there is ev-(Wheatcroft 1991). idence (visual observations, photographs, On the surface, cores were processed and large mounds and broad surface de-within 4 h by siphoning the supernatant wa- pressions) for the presence of a subsurfaceter and placing the corers onto extruders. echiuran worm that mixes particles advec-After extruding a vertical interval of 1 cm, tively (C. Smith et al. 1986; Smith in press).a thin (0.5 mm) sheet of metal was used to Thus, treatments in this study were locatedvertically isolate the uppermost slab of mud. away from the mounds-depression com-A 4.2-cm-diameter plastic fence was then plexes to avoid the complications of large-pressed into the center of the subsample and scale (>10 cm) advective mixing. the outer ring of mud discarded. The central disk of mud, ostensibly free of vertical Field and laboratory procedures contamination, was then transferred to a The tracers used in this study are spher-sample bottle. The approximately equal- ical glass beads ranging in size from 8 tovolume samples (14 cm3) were collected to 420 tm. To isolate the effect of size on bio- 10-cm depth. turbation rate, I held other attributes of the The method used to enumerate the trac- particles such as specific gravity (2.4) anders consists of randomly settling grains onto surface roughness (very smooth) constant.microscope slides where they can be count- The beads were spread on the sediment sur-ed (method of Moore 1973 as modified by face with the DSV Alvin and "salt shakers":Laws 1983). The samples were sonicated in 1-liter freezer containers with plastictheir bottles for 3 mm with a 250-W Vi- screening (1.2-mm mesh) across their open-bracell ultrasonicator. Pretests showed that ings. Aluminum-T handles were bolted tothis step completely disaggregated all fecal the sides of the bead spreaders for manip-pellets and other aggregates without affect- Particle size-dependent bioturbation 95 ing the glass beads. After transfer to 600-miinventory was then used to normalize the beakers, samples were split into three sub-number counts from each depth interval; samples, each 2/27 (7.4%) of the total sam- thus, concentrations reported in the ensuing ple. These replicate subsamples were sievedtables and figures represent the proportion through 62-Lm Nytex screens. The >62-m (i.e.relative concentration) of the total fraction was sieved onto 8-nm Nucieporenumber of beads of that size class at that polycarbonate filters and transferred onto 5 depth interval estimated in each core. x 7.5-cm glass slides to await counting. The Two paths were used to arrive at an over- <62-itm fraction was poured into 3,000-mlall comparison of particle size-specific mix- beakers filled with tapwater, stirred vigor- ing intensity. The first method simply com- ously, and allowed to settle for 30 mm (seepares the weighted-mean depth of the Moore 1973). Calculations with Stokes' lawconcentration profiles between size classes. suggest that two-thirds of the 8-sm particlesMore intensely displaced size fractions would have transitted the 15 cm of watershould be, on average, moved farther down in the beakers in that time. The supernatantinto the sediment, yielding larger weighted- water was then siphoned to below the mi- mean depths. This technique makes no as- croscope slide stages (see text-figure1 ofsumptions about how (e.g. biodiffusion or Laws 1983), and 250-W infrared heat lamps advection) the particles arrive at their ob- were positioned above the stages. Evapo- served destinations. ration of the remaining water was complete The second method of analysis uses the in <1 h. The slide preparation techniquediffusion analogy of bioturbation to arrive yielded randomly distributed samples, asat a biodiffusion coefficient representative indicated by x2 test for agreement with aof the intensity of vertical particle displace- Poisson distribution (Sokal and Rohlf 1981)ment (Guinasso and Schink 1975; Wheat- on randomly selected slides. croft et al. 1990). Long-term sediment ac- The <62-nm fraction was counted withcumulation rates in SCB (Schwalbach and a Zeiss compound microscope (25 x withGorsline 1985) suggest that during the max- 10 x oculars) and transmitted light. To min-imal length (997 d) of this experiment <0.5 imize subjectivity I assigned subsamplesmm of sediment would have accumulated. random letters and counted blindly in anThus, sedimentation can be neglected as a arbitrary order. Only after counting all depthvertical mass transfer mechanism, and the intervals from a given core were the resultsconservation equation for a nondecaying, tallied. Six stratified random transects wereparticulate tracer undergoing diffusive mix- made across each slide, and the glass beadsing is that were at least halfway within a 280-am oc wide ocular grid were counted into one of (1) three size classes: 8-16, 17-31, and 32-62 tm. Counting each of the two slides in eachwhere C is tracer concentration, t is time, beaker resulted in 36 transects. They formedand z is depth into the sediment. Downcore the basis of the size-specific particle con-porosity variations have been neglected. For centrations and the standard errors at eachan instantaneous, plane source (i.e. at t = 0, depth level. The coarse fraction was count-C = 0 at z > 0 and C = 1 at z = 0), with ed into two size fractions (63-125 and 126-reflection at a boundary (i.e. 8C/8z = 0 at 420 jim) on the glass slides with the Zeissz = 0) the solution to Eq. 1 is (Crank 1975) microscope (10 x with 10 x oculars). 1 Numerical and statistical methods C(z) exp(z2/4Dbt). (2) (lrDbt)1/2 The transects yielded an estimate of the number of tracers (± 1 SE, N 36) in each The biodiffusion coefficient that yields the of five size classes at each depth interval.best-fit concentration profile was computed Within each core the results were summedby iteratively solving Eq. 2 with different to provide an estimate of the total inventoryvalues of Db. The resultant theoretical con- of glass beads of that size. The size-specificcentration profiles were then compared to 96 Wheatcrofi

Table 1.Normalized glass bead concentrations for Statistical comparisons between either the each size class in the three cores. weighted mean displacement (method 1) or

Size class (JLm) the best-fit biodiffusivity (method 2) of each Depth size class were performed with various non- (cm) 8-16 17-31 32-62 63-125 126-420 parametric tests. Page's modification of the Core 1,937-1 ordered alternative test (Hollander and 0-1 0.368 0.529 0.612 0.470 0.741 Wolfe 1973) was used to test the null hy- 1-2 0.270 0.374 0.331 0.452 0.231 2-3 0.140 0.051 0.032 0.030 0.021 pothesis of no differences between treat- 3-4 0,083 0.024 0.007 0.034 0 ments (i.e. tracer sizes) on various attributes 4-5 0.052 0.012 0.007 0.007 0 (i.e. Db or mean vertical displacement) 5-6 0.052 0.007 0 0.004 0.007 against the alternative of a natural ordering 6-7 0.026 0.004 0.010 0.001 0 of treatments with respect to the attributes. 7-8 0.008 0 0 0.002 0 8-9 0 0 0 0 0 A nonparametric test for correlation, Spear- Core 1,937-2 man's p, was used to test for significance 0-1 0.484 0.834 0.846 0.896 0.983 between computed mixing rates and particle 1-2 0.124 0.051 0.049 0.029 0.017 size. 2-3 0.088 0.027 0.022 0.026 0 3-4 0.071 0.017 0.010 0.002 0 Results 4-5 0.082 0.028 0.019 0.010 0 In all three of the long-term cores (the 5-6 0.056 0.032 0.043 0.033 0 6-7 0.028 0.004 0 0.002 0 short-term experiment was missed during 7-8 0.024 0.004 0.007 0.00 1 0 attempted recovery) there are strong indi- 8-9 0.042 0.004 0.005 0.001 0 cations of an increase in mixing intensity Core 1,937-3 with decreasing tracer size. In each core, 0-1 0.340 0.489 0.607 0.694 0.792 there is an increasing fraction of fine tracer 1-2 0.295 0.259 0.187 0.123 0.034 relative to the coarser beads at increasing 2-3 0.193 0.195 0.149 0.103 0.044 sediment depths (Table 1). Page's L-test for 3-4 0.098 0.030 0.029 0.042 0.070 4-5 0.050 0.024 0.028 0.034 0.060 ordered alternatives rejects the null hypoth- 5-6 0.013 0.002 0 0.001 0 esis of equal depth-weighted means for the 6-7 0.012 0 0 0 0 various tracer size fractions at a high level 7-8 0 0.001 0 0 0 of significance (P=0.001) (Table 2). The 8-9 0 0 0 0 0 control core, on the other hand, revealed that <0.1% of any tracer size class pene- trated through the surface (0-1 cm) layer. the observed data, and the biodiffusivity thatThese results suggest that, whatever the yielded the smallest median residual wasmechanism of mass transfer, it is not likely chosen. The best-fit technique is analogous a sampling artifact and it is particle size- to a logarithmic transformation in that itdependent. assigns an equal weighting to all depth in- In two (1,937-1 and 1,937-3) of the three tervals and thus removes the dispropor- cores the vertical concentration profiles can tionate influence of near-surface concentra- be fitted with the biodiffusive model of sed- tions. iment mixing (Eq. 1 and 2). Vertical biodif-

Table 2.Summary of depth-weighted mean displacements(z,in cm) for each size class, In parentheses are the within-core ranks used to compute Page's L-test for ordered alternatives. The null hypothesis is z5 z4 and the alternative is z5 > z4 > > z2 > z1; R, is the rank sum of each size class.

Size class (tim) Core 8-16 17-31 32-62 63-125 126-420 1,937-1 1.98(5) 1.15(3) 1.01(2) 1.18(4) 0.78(1) 1,937-2 2.30(5) 1.01(4) 1.00(3) 0.82(2) 0.52(1) 1,937-3 1.8 1(5) 1.35(4) 1.18(3) 1.11(2) 1.07(1) R5=15 R4=11 R3=8 R2=8 R1=3 Particle size-dependent bioturbation 97

RELATIVE CONCENTRATION

Fig. 1.Core 1,937-1: Observed vertical profiles of tracer concentration (± 1 SE, N = 36) for each size class. The solid lines represent the theoretical concentration profiles of the best-fit biodiffusivities. fusivities for core 1,937-1 range from 1.05in concentration at a sediment depth of 4- cm2 yr' for the 8-16-tm fraction to 0.156 cm (Table 1, Fig. 3). There are at least cm2 yr' for the coarse (126-420 jim) frac-two possible mechanisms that could ac- tion, a variation of about an order of mag-count for such a pattern. First, the glass beads nitude (Fig. 1). Intermediate tracer size frac-may have fallen into open burrows or tubes tions yield intermediate Db values, althoughwhen spread initially onto the sediment sur- the trend is not monotonic. The 63-1 25-timface or later were pushed into tubes or bur- size class of tracers penetrates deeper intorows by mobile epibenthic megafauna (e.g. the sediment and yields a D,, greater thanthe brittle star 0. normani). If this mech- the next smaller size class of tracers (Fig. 1).anism was responsible, any preferential A similar trend in Db values was found intransport of a size class should tend toward core 1,937-3 (Fig. 2). In that core, the finestthe coarser fraction for reasons discussed tracer (8-16 m) yields a biodiffusivity ofabove. The second mechanism is via a pro- 0.95 cm2 yr', whereas the 63-125-jim frac-cess termed reverse conveyor-belt deposit tion yields a Db of 0.45 cm2 yr'. Note thatfeeding (figure I-3f, Wheatcroft et al. 1990), the coarsest fraction of tracer in core 1,937-3 whereby particles are ingested at the surface does not display a diffusive profile (Fig. 2),and egested at some depth in the sediment. but instead shows a subsurface peak in con-Such a nonlocal transport mechanism has centration. been proposed by J. Smith et al. (1986) to In contrast to the overall good fit of the account for subsurface peaks in 239'240Pu and tracer concentration data in cores 1,937-1 210Pb profiles in NE Atlantic sediments. At and 1,937-3 to various biodiffusivities, datatheir site the responsible animal is likely a from core 1,937-2 do not conform to a sim- sipunculan worm, which is the macrofaunal ple biodiffusive model. In all but the coars-community dominant and is known to feed est tracer fraction (which yields a Db of 0.05 in a reverse conveyor-belt mode (Rutgers cm2 yr1), there is a local subsurface peakvan der Loeff and Lavaleye 1986). In SCB 98 Wheatcrofi

RELATIVE CONCENTRATION 0.001,,91 ,,,91J

F!1 6- DBO.45cm2yr'

32-62 im

10-

Fig. 2.As Fig. 1, but for core 1,937-3. The coarse fraction (126-420 m) could not be fitted with a simple biodiffusive model. it is more difficult to nominate a responsibleof their simulations there is a similar, sub- species, but it is interesting to note that6%surface peak. The interesting point is that of SCB macrofauna are members of the cir-with increasing background biodiffusive ratulid polychaete genus Tharyx (C. Smithmixing rates, the subsurface peak broadens et al. 1986). Congenerics in shallow watervertically and decreases in magnitude (fig- have been observed to feed in a reverse con-ures 6-10, J. Smith et al. 1986). This peak veyor-belt mode (Myers 1977). Regardlessbroadening and damping appears to occur of what mechanism is responsible, the datain core 1,937-2 as well (Fig. 3), where there themselves suggest preferential selection ofis an increase of16-fold from the 3-4-cm the finer fraction, since there is a weak trendinterval to the 5-6-cm interval in the coarse of increasing tracer concentration in the 4-(63-125 m) fraction, but succeedingly less- 6-cm depth interval with decreasing particleer increases (4-fold and 2-fold) at finer tracer size (Table 1). sizes. Thus, there is sound evidence that Even though it is not possible to fit a sim-biodiffusion rate increases with decreasing pie vertical biodiffusivity to the data fromtracer size in this core as well. core 1,937-2, several lines of evidence sup- Page's L-test for ordered alternatives is port greater mixing intensity of the fine again used to test the null hypothesis (Table fraction in this core as well. First, the depth-3) of equal Db values against the alternative weighted concentration increases with de- of a predetermined unidirectional trend in creasing particle size (Table 2). Second, thebiodiffusivities (Hollander and Wolfe 1973). shapes of the profiles qualitatively matchDue to fewer data points, the results are less those produced by J. Smith et al. (1986).significant than the outcome of the weight- They derived a coupled biodiffusion-re-ed-mean displacement data, but the null hy- verse conveyor-belt model that yields a suitepothesis is still rejected (P = 0.04). Thus, of theoretical concentration profiles that are there would seem to be a statistically sig- very similar to the core 1,937-2 data. In allnificant trend in the size-specific mixing rates Particle size-dependent bioturbation 99

RELATIVE CONCENTRATION 0.001

5.

17-31 zm

10-

0.001 0.'t 0.1

8

5-.

z 63-125 sm

to Fig. 3.As Fig. 1, but for core 1,937-2. The solid line in the l26-42O-em fraction is the theoretical concen- tration profile for the best-fit biodiffusivity. Profiles of the other tracer size fractions could not be fitted with a simple biodiffusive model.

(Fig. 4). A nonparametric measure of cor-nificant at P < 0.01. Bioturbation is particle relation, Spearman's p, was used to sepa-size-dependent in SCB. rately test for negative correlation between particle size and Db in cores 1,937-1 andDiscussion 1,937-3. The test yields values of Spear- The strength of the negative relationship man's p of 0.90 and 0.95, which are observed between sediment mixing rate and significant at the 0.05 and 0.01 levels. Fish-particle size requires a discussion of poten- er's method of combining independenttial artifacts in the deployment and enu- probabilities (Sokal and Rohlf 1981) pro-meration of the tracers. Artifacts could be vides an overall test of significance. The re- introduced during three phases of this in- sults indicate a negative correlation between vestigation: bead spreading, coring, and particle size and biodiffusivity that is sig-counting. During their initial deployment

Table 3. Summary of particle size-specific biodiffusivities (Db, cm2 yr). In parentheses are the within-core ranks used to compute Page's L-test for ordered alternatives. Note that the 1 26-42O-m fraction has been excluded from this analysis. ND indicates that a biodiffusion coefficient could not be fitted to the observed concentration profile. Table 2 gives the explanation of null and alternative hypotheses.

Size class (Mm)

Core 8-16 17-31 32-62 63-125 126-420 1,937-1 1.05(4) 0.65(3) 0.45(1) 0.60(2) 0.15 1,937-2 ND ND ND ND 0.05 1,937-3 0.95(4) 0.60(3) 0.45(1.5) 0.45(1.5) ND R48 R3=6 R2=2.5 R1=3.5 100 Wheatcrofi

were more easily masked. Thus, one would expect that with increasing depth the finer tracers would be missed more often relative to the coarse fractions. All of these argu- ments suggest minimal artifacts due to the 0 experimental setup and that vertical bio- turbation in SCB is strongly particle size- dependent. 0 Earlier it was argued that deposit feeding is the only animal activity that unequivo- 0 cally favors preferential mixing of finer par- ticles. Thus, the results reported here lend support to previous contentions (Thayer 1983; Wheatcroft et al. 1990) that deposit 0.01-V feeding dominates particle transport in the 10 100 MEAN TRACER SIZE(pm) abyss. The strength and even direction of the relationship still remains surprising, Fig. 4. A log-log plot of observed particle size-spe- cific biodiffusivities vs. mean tracer size. Core 1,937- however, given the textural complexities of 1-0; core 1,937-2-0; core 1,937-3Li. marine sediments and the feeding mechan- ics of deposit feeders. Marine sediments in general, and SCB sediments in particular, from the bead spreaders, settling velocitiesare complex mixtures of fecal pellets of var- of the tracers would differ considerably. ious sizes and shapes, organically bound ag- Thus, there is the possibility of differentialgregates, living and dead protistans and oth- initial penetration into the sediments. Anyer skeletal components, as well as various such bias would be toward greater penetra- rock and mineral fragments ranging in size tion of the coarse tracers, but it is unlikelyfrom <1 im to 1-2 mm. When this com- that penetration would exceed the 1 cm nec- plex mixture is processed at the high rates essary to reach the next lowest depth inter-(0.1 to >20 Hz, Miller 1984; R. F. L. Self val. Although the leading edge of the tubepers. comm.) typical of most deposit feed- corers is beveled to minimize vertical sub-ing, any selection on the basis of a single duction of sediment, invariably some sur-cue (i.e. size) is remarkable. face sediment is entrained downward. If this Particle selection by deposit feeders is at effect occurred, it would also favor subduc-least a two-step process (Jumars et al. 1982; tion of the larger tracers because of theirWhitlatch 1989) made up of the probability greater per-particle area. As detailed in theof contact followed by the probability of methods section and demonstrated by theretention. For surface deposit feeders that control core, however, discarding the outerpat the sediment with various types of feed- ring of sediment at each depth intervaling appendages, there is a greater probabil- seemed to minimize problems with verticality of contacting larger particles due purely subduction. If more coarse tracers wereto the greater per particle area they cover somehow missed during counting at greaterrelative to fine particles (Jumars et al. 1982). depths in the sediment, then the patternsThe same is true to a lesser degree for sub- observed might be artifacts. Again, any po-surface gulpers (Whitlatch 1989) that ingest tential bias is most probably in the oppositea discrete mouthful of sediment. Thus, be- direction (i.e. counting of coarse particlescause the probability of contact favors larger relative to fines was favored at increasingparticles over small ones, the probability of depth). This direction of bias is because theretention must be the dominant mechanism tracers were counted if and only if they were operating to explain the results here. The unequivocally identified as glass beads. Be- mechanics leading to retention are less eas- cause of decreasing porosity with greaterily generalized and depend on particle size depth resulting in greater amounts of veryas well as on other factors including shape fine silt and clay, the smaller glass beads(Whitlatch 1974), specific gravity and sur- Particle size-dependent bioturbation 101 face roughness (Self and Jumars 1978), or-sequently the animal can return to reingest ganic coating (Taghon 1982), and the ad-this material and the accompanying micro- hesive strength of mucus covering thebial community that may have flourished feeding appendage (Jumars et al. 1982). Thein the steep geochemical gradients that sur- results reported here strongly suggest thatround macrofaunal tubes and burrows (All- particle size is a dominant cue in retention,er and Aller 1986). because only particle size (and its derived It is also conceivable that the present ex- variables, e.g. surface area or volume) var-periment, although long (997 d) by most ied in this experiment. standards, was not of sufficient duration for Recall from the discussion above thatparticles to move into the zone where head- many shallow-water deposit feeders egestdown deposit feeders might have been feed- material on the sediment surface, thus aing. Thus, the fine particles never were re- possibility was that fines would be prefer-turned tothesurface. The rate of entially kept in the surface mixing layer. Inbioadvective overturn may be much higher experiments similar to this one performedat Whitlatch's site in Panama Basin. Al- in the abyssal Panama Basin, coarse parti- though all of the above arguments are plau- cles (-80 m) penetrated deeper into thesible they remain speculative and under- sediment than fines (R. B. Whitlatch pers.score two important points. First, present comm.). The likely explanation for this pat-knowledge of the natural history of deep- tern is that deposit feeders (surface or sub- ocean animals is generally lacking. Thus, we surface) in Panama Basin egest material onare often faced with the unsatisfactory sit- the sediment surface. In contrast, the resultsuation of applying information concerning reported here demonstrate that fine parti-sediment displacement activities derived cles penetrate farther into the sediment andfrom shallow-water species to their deep- suggests that surface egestion is not preva-ocean cousins. Given that the deep ocean lent in SCB. is in some ways fundamentally different than There are grounds to believe that depositany shallow-water site, it is reasonable to feeders in the deep ocean in general do notexpect novel behaviors or at the very least egest material on the sediment surface, andshifts in the importance of various etholo- that Whitlatch's results are exceptional. Thegies (Jumars et al. 1990). The second, equal- reasons behind this tentative assertion arely important point is that bioturbation is twofold: first, in most areas of the deep sea,highly variable in the deep sea. Simple, one- including SCB, near-bottom current veloc-dimensional diffusive models that treat all ities are not sufficient to remove fecal pelletsparticles the same will not accurately por- placed on the sediment surface. Thus, overtray the transport kinetics of deep-sea de- time the animal would fill its feeding areaposits. Sediments in the deep ocean show with feces and this would likely have a neg- both large- and small-scale spatial changes ative impact, because, at least in shallowin mixing mode and rate. The next gener- water, there is some evidence that accu-ation of early diagenetic models must strive mulation of feces adjacent to an animal'sto incorporate greater biological realism (cf. tube depresses the feeding rate of that ani-Smith in press). mal (Miller and Jumars 1986; C. A. Butman Results obtained here can be generalized pers. comm.). Shallow-water currents aretentatively to other marine environments. sufficiently vigorous to remove pellets, thusVertical biodiffusivities are likely to be even avoiding this complication. The second rea-more dependent on particle size in deeper son that deep-sea deposit feeders may notabyssal, oligotrophic areas, whereas in shal- egest material on the sediment surface islow-water sediments mixing rates may be that there may be some benefit in seques-independent of particle size. Two observa- tering feces from other heterotrophs (Ju-tions are offered to support these conjec- mars et al. 1990). The easiest way to do this tures. First, the whole premise of size-de- is to place material at some depth withinpendent mixing rests on the assumption that the sediment, because most animals live infood for deposit feeders is surface associ- the upper few centimeters of sediment. Sub-ated, and, because finer particles have great- 102 Wheatcroft er surface areas per unit of sediment vol-particles of the same size. They do, how- ume, more food is associated with them. Ifever, possess different decay periods. It may there are other sources of food (e.g. plantbe that a recently proposed (Smith et al. detritus), then trends in those food sourcesunpubl.) process, termed age-dependent might influence the degree of size selectiv-mixing, has a greater effect on Db values ity. Thus, the observed decreases in thederived from radionuclides with similar amount of terrestrial plant detritus with in-distributions of particle size. Age-depen- creasing depth and distance from land sug-dent mixing postulates that radionuclides gest that particle size selective pressures maywith shorter half-lives are associated with increase in deeper waters. That is, in shallowlabile organic matter for a greater propor- continental shelf sediments an appreciabletion of their lives (Smith et al. unpubl.). amount of food for deposit feeders is notThus, they are more intensively mixed rel- strictly controlled by surface area-to-vol-ative to longer lived radiotracers. Clearly, ume ratios. In such settings, other cues, suchmuch additional work is needed to assess as specific gravity or surface texture maythe prevailing cues for particle selection and dominate over particle size. Secondly, therehence displacement. This study indicates is reason to believe that the decreasing sizeunequivocally that particle size is an im- of abyssal animals (Jumars et al. 1990) mayportant one. influence overall selectivity, since, at least The implications of particle size-depen- for shallow-water species, larger animalsdent mixing are clearest for studies of deep- seem to exhibit less particle size selectivitysea biostratigraphy and paleoclimatology. (Self and Jumars 1988). From the viewpoint of these disciplines, Particle size-dependent mixing has ram- bioturbation acts as a low-pass filter that ifications for the redistribution of variousblurs and alters the apparent timing and am- particle-associated radionuclides and theirplitude of recorded oceanographic, bio- resultant biodifflisivities, although in a morestratigraphic, and climatologic events equivocal manner than one might initially(Loubere 1989; Wheatcroft 1990). Recent- assume. Recall that several earlier studiesly, attention has been devoted to removing comparing biodiffusion coefficients derivedthe effects of sediment mixing from strati- from various sources rationalized differ-graphic signals (Berger et al. 1978; Schiffel- ences in observed mixing rates by invokingbein 1985). These unmixing or deconvo- particle size-dependent mixing. For exam-lution algorithms use estimates of mixing- ple, DeMaster and Cochran (1982) arguedlayer thickness (Lb) and the biodiffusivity convincingly that the higher Db values de-(Db) to parameterize an impulse response rived from 210Pb compared to 32Si profilesfunction that is in turn used to deconvolve were due to the association of 210Pb withthe signal of interest (Schiffelbein 1985). An the clay fraction, whereas 32Si was associ-explicit assumption is that Lb and Db, and ated primarily with large, slowly mixed ra-thus the impulse response function, do not diolarians. Similarly, recently reported dif-vary among sedimentary constituents. The ferences in mixing rates derived from 210Pbresults obtained in this study suggest that and 230Th profiles vs. Db values computedsuch assumptions are not tenable. Mixing from the penetration rate of planktonic fo- rate decreases by at least a factor of four raminiferans into a recently emplaced tur-between tracer sizes from -'20 to 200 tm bidite were explained as resulting fromsuggesting similar differences in mixing rates greater mixing of the radionuclides associ-of different sized microfossils. There is also ated with the fines relative to the larger Fo-the strong possibility, although this study raminifera (Thomson et al. 1988). Both sce-did not address it, that mixing rates and narios are supported by the findings of thismixing-layer thicknesses of still larger (0.5- study. It is less easy to invoke particle size-2 mm) microfossils are even less. Thus, fu- dependent mixing to reconcile differencesture attempts at deconvolving the mixing in 210Pb and 239'240Pu-derived biodiffusiv-function and modeling the effect bioturba- ities (Stordal et al. 1985), because these ra- tion has on microfossil microhabitats should dionuclides are generally associated with thetake into account this added complexity and Particle size-dependent bioturbation 103 may even profit from the explicit inclusion tative estimates of biological mixing rates in abys- of size-dependent bioturbation rates. sal sediments. J. Geophys. Res. 80: 3032-3043. HOLLANDER, M., AND D. A. WOLFE. 1973. Nonpara- In closing, it is important to reiterate the metric statistical methods. Wiley. main conceptual point of this paper. Bio- JuMARs, P. A. 1976. Deep-sea species diversity: Does turbation can and must be approached from it have a characteristic scale? J. Mar. Res. 34: 217- both viewpoints discussed in the introduc- 246. tion (i.e. from species-specific, autecologic L. M. MAYER, J. W. DEMING, J. A. BAlwss, AND R. A. WHEATCROFT.1990.Deep-sea de- studies and community-wide studies). Only posit-feeding strategies suggested by environmen- by using what is known concerning individ- tal and feeding constraints. Phil. Trans. R. Soc. ual styles and rates of sediment reworking Lond. Ser. A 331: 85-101. to formulate and test hypotheses regarding A. R. M. NOWELL, AND R. F. L. SELF1981. A simple model of flow-sediment-organism inter- various aspects of bioturbation as repre- action. Mar. Geol. 42: 155-172. sented by community-wide measures of ,rnj . 1982. Mechanics of par- mixing (e.g. biodiffusivities) will a better un- ticle selection by tentaculate deposit feeders. J. derstanding of sediment mixing evolve. Exp. Mar. Biol. Ecol. 64: 47-70. LAWS, R. A. 1983. Preparing strewn slides for quan- References titative microscopical analysis: A test using cali- brated microspheres. Micropaleontology 29: 60- ALLER, J. Y., AND R. C. ALLER. 1986. Evidence for 65. localized enhancement of biological activity as- LOUBERE, P.1989. Bioturbation and sedimentation sociated with tube and burrow structures in deep- rate control of benthic microfossil taxon abun- sea sediments at the HEBBLE site, western North dances in surface sediments: A theoretical ap- Atlantic. Deep-Sea Res. 33: 755-790. proach to the analysis of species microhabitats. BERGER, W. H., AND R. F. JOHNSON. 1978. On the Mar. Micropaleontol. 14: 317-325. thickness and the '4C age of the mixed layer in MCCAVE, I. N. 1988. Biological pumping upwards of deep-sea carbonates. Earth Planet. Sci. Lett. 41: the coarse fraction of deep-sea sediments. J. Sed- 223-227. iment. Petrol. 58: 148-158. AND J. S. KILLINGLEY.1978. "Un- MILLER, D. C.1984. Mechanical post-capture par- mixing" of the deep-sea record and the deglacial ticle selection by suspension and deposit-feeding meltwater spike. Nature 269: 661-663. Corophium. J. Exp. Mar. Biol. Ecol. 82: 59-76. BERNER, R. A. 1980. Early diagenesis: A theoretical AND P. A. JuMAJs. 1986.Pellet accumula- approach. Princeton. tion, sediment supply, and crowding as determi- CHERNOW, R. M., R. W. FREY, ANr B. B. ELwooD. nants of surface deposit-feeding rate in Pseudo- 1986. Biogenic effects on development of mag- polydora kempi japonica Imajima & Hartman netic fabrics in coastal Georgia sediments. J. Sed- (Polychaeta: Spionidae). J. Exp. Mar. Biol. Ecol. iment. Petrol. 56: 160-172. 99: 1-17. CLIFTON, T. R. 1984. Heavy mineral concentration MOORE, T. C. 1973. Method of randomly distributing at the bottom of polychaete traces in sand sedi- grains for microscopic examination. J. Sediment. ment. J. Sediment. Petrol. 54: 15 1-153. Petrol. 43: 904-906. COCHRAN, J. K. 1985. Particle mixing rates in sedi- MYERS, A. C. 1977. Sediment processing in a marine ments of the eastern equatorial Pacific: Evidence subtidal sandy bottom community. 2. Physical as- from 210Pb, 239.24o and '37Cs distributions at pects. J. Mar. Res. 35: 633-647. MANOP sites. Geochim. Cosmochim. Acta 49: PENRY, D. L., AND P. A. JUMARS. 1990. Gut archi- 1195-1210. tecture, digestive constraints and feeding ecology CRANK, J.1975. The mathematics of diffusion. Clar- of deposit-feeding and carnivorous polychaetes. endon. Oecologia 82: 1-11. DEFLAUN, M. F., ANr L. M. MAYER. 1983. Rela- POWELL, E. N. 1977. Particle size selection and sed- tionships between bacteria and grain surfaces in iment reworking in a funnel feeder, Leptosynapta intertidal sediments. Limnol Oceanogr. 28: 873- tenuis (Holothuroidea, Synaptidae). Tnt. Rev. Ge- 881. samten Hydrobiol. 62: 385-408. DEMASTER, D. J.,mJ. K. COCHRAN. 1982. Particle RHOADS, D. C. 1967. Biogenic reworking of intertidal mixing rates in deep-sea sediments determined and subtidal sediments in Barnstable Harbor and from excess 210Pb and 32Si profiles. Earth Planet. Buzzards Bay, Massachusetts. J. Geol. 75: 461- Sci. Lett. 61: 257-271. 476. EAGER, E. W. 1964. Marine sediments: Effects of a RUDDIMAN, W. F., AND OTHERS. 1980. Tests for size tube-building polychaete. Science 143: 356-359. and shape dependency in deep-sea mixing. Sedi- FENCHEL, T., L. H. KOFOED, AND A. LAPPALAINEN. ment. Geol. 25: 257-276. 1975. Particle size-selection of two deposit feed- RUTGERS VAN DER LOEFF, M. M., AND M. S.S. ers: The amphipod Corophium volutator and the LAvAi.Eye.1986. Sediments, fauna and the dis- prosobranch Hydrobia ulvae. Mar. Biol. 30: 119- persal of radionuclides at the N.E. Atlantic dump- 128. site for low-level radioactive waste. Neth. Inst. Sea GUINASSO, N. L., AND D. R. SCHINK. 1975. Quanti- Res. 134 p. 104 Wheatcroft

SANDERS, H. L.1960. Benthic studies in Buzzards 1989. Modeling deposit feeding, p. 223-246. Bay. 3. The structure of the soft-bottom com- In G. Lopez etal. [eds.], Ecology of marine deposit munity. Limnol. Oceanogr. 5: 138-153. feeders. Springer. SANDERSON, B.1985. How bioturbation supports R. F. L. SELF, AND P. A. JUMARS. 1978. Pre- manganese nodules at the sediment-water inter- dicting particle selection by deposit feeders: A face. Deep-Sea Res. 32: 1281-1285. model and its implications. Limnol. Oceanogr. 23: SCHIFFELBEIN, P. 1985. Extracting the benthic mixing 75 2-7 59. impulse response function: A constrained decon- THAYER, C. W. 1983. Sediment-mediated biological volution technique. Mar. Geol. 64: 313-336. disturbance and the evolution of marine benthos, SCHWALBACH, J. R., AND D. S. GORSLINE. 1985. Ho- p. 479-625. In M. J. S. Tevesz and P. L. McCall locene sediment budgets for the basins of the Cal- [eds.], Biotic interactions in recent and fossil com- ifornia continental borderland. J. Sediment. Pe- munities. Plenum. trol. 55: 829-842. THOMSON, J., S. COLLEY, AND P. P. E. WEAVER. 1988. SELF, R. F. L., AN P. A. JUMARS. 1978. New resource Bioturbation into a recently emplaced deep-sea axes for deposit feeders? J. Mar. Res. 36: 627-641. turbidite as revealed by 210Pb, 230Th,, and AND . 1988. Cross-phyletic patterns of planktonic Foraminifera distributions. Earth particle selection by deposit feeders. J. Mar. Res. Planet. Sci. Lett. 90: 157-173. 46: 119-143. WHEATCROFT, R. A. 1990. Preservation potential of SMITH, C. R. In press. Factors controlling bioturba- sedimentary event layers. Geology 18: 843-845. tion in deep-sea sediments and their relation to 1991. Conservative tracer study of horizontal models of carbon diagenesis. In G. T. Rowe [ed.], sediment mixing rates in a bathyal basin, Califor- Deep-sea food chains: Relation to the global car- nia borderland. J. Mar. Res. 49: 565-568. bon cycle. Kiuwer. AND P. A. Juins. 1987. Statistical re-anal- AND S. C. HAMILTON. 1983.Epibenthic ysis for size dependency in deep-sea mixing. Mar. megafauna of a bathyal basin off southern Cali- Geol. 77: 157-163. fornia: Patterns of abundance, biomass, and dis- C. R. SMITH, AND A. R. M. NOWELL. persion. Deep-Sea Res. 30: 907-928. 1990. A mechanistic view of the particulate bio- P. A. JUMARS, AND D. J. DEMASTER. 1986. diffusion coefficient: Step lengths, rest periods and In situ studies of megafaunal mounds indicate rap- transport directions. J. Mar. Res. 48: 177-207. id sediment turnover and community response at WHITLATCH, R. B.1974. Food-resource partitioning the deep-sea floor. Nature 323: 25 1-253. in the deposit-feeding polychaete Pectinaria gould/i. SMITH, J. N., B. P. BOUDREAU, ANDY. NOSHKIN. 1986. Biol. Bull. 147: 27-235. Plutonium and 210Pb distributions in northeast At- 1989. On some mechanistic approaches to lantic sediments: Subsurface anomalies caused by the study of deposit feeding polychaetes, p. 291- non-local mixing. Earth Planet. Sci. Lett. 81: 15- 308. In G. Lopez et al. [eds.], Ecology of marine 28. deposit feeders. Springer. SOKAL, R. R., AND F. J. ROHLF.1981.Biometry. AND J. R. WEINBERG.1982. Factors influ- Freeman. encing particle selection and feeding rate in the STEPHENS, D. J., AND J. BRIDOWATER. 1978. The mix- polychaete Cistenides (Pectinaria) gouldii. Mar. ing and segregation ofcohesionless particulate ma- Biol. 71: 33-40. terials Part 2. Microscopic mechanisms for par- WILLIAMS, J. C., AND M. I. KAHN. 1973. Mixing and ticles differing in size. Powder Technol. 21: 29- segregation of particulate solids of different par- 44. ticle size. Chem. Eng. 269: 19-25. STORDAL, M. C., AND OTHERS.1985.Quantitative evaluation of bioturbation in deep ocean sedi- ments. 2. Comparison of rates determined by 210Pb and 239'240Pu. Mar. Chem. 17: 99-114. TAGHON, G. L.1982. Optimal foraging by deposit- Submitted: 14 November 1990 feeding invertebrates: Roles of particle size and Accepted: 6 August 1991 organic coating. Oecologia 52: 295-304. Revised: 5 September 1991