© 1999 Macmillan Magazines Ltd

are ingested with lizards — rather thandirectly from the plant — is supported byobservation and by a strong correlationbetween lizard remains and Lycium seeds inthe pellets. The seeds in the shrike pellets hada higher germination rate (64%) than thosefrom lizard droppings (50%) or directlyfrom the plant (54%), showing that theirexperience in passing through two vectorshad increased their potential for immediategermination.

Lizards are an effective dispersal systemfor Lycium on Alegranza, but movementof the plant between islands using the lizardas a vector is likely to be rare (although notimpossible5). Movement of the seeds fromlizard to bird, however, presents a range ofopportunities for island-dwelling plants, giv-ing them a far greater chance of moving tonew islands. Because the seeds are retained inthe shrike’s gizzard for less than an hour, feed-ing and flight must follow in rapid successionif dispersal by this means is to be effective.Like so many other events in island biology,given enough time it will probably happen.

The significance of predators as sec-ondary dispersers of fruits and seeds is likelyto be limited to specific situations. In Britain,

the sparrowhawk (Accipiter nisus) is knownto exploit situations where flocks of fruit-eating birds are feeding6. So, again, thepredator could easily ingest seeds with theprey. But the transfer of seeds from one birdto another in this way is unlikely to havemuch of an effect on seed dispersal. On Ale-granza, Lycium seeds have also been found inkestrel (Falco tinunculus) pellets, but thesemay have been derived from fruit-eatingbirds rather than lizards. Only where seedtransfer involves increased mobility andrange (as in the plant–lizard–bird sequence)will new dimensions of dispersal be opened.Predation can then assume a biogeographi-cally significant role.Peter D. Moore is in the Division of Life Sciences,King’s College, Campden Hill Road, LondonW8 7AH, UK.e-mail: peter.moore@kcl.ac.uk1. Nogales, M., Delgado, J. D. & Medina, F. M. J. Ecol. 86, 866–871

(1998).

2. van der Pijl, L. Principles of Dispersal in Higher Plants (Springer,

Berlin, 1982).

3. Fridriksson, S. Surtsey (Wiley, New York, 1975).

4. Whittaker, R. J. & Jones, S. H. J. Biogeogr. 21, 245–258

(1994).

5. Cendsky, E. J., Hodge, K. & Dudley, J. Nature 395, 556 (1998).

6. Snow, B. & Snow, D. Birds and Berries (T. & A. D. Poyser,

Calton, 1988).

its oxidation state), was transported signifi-cant distances through a groundwater sys-tem, and they argue that colloidal specieswere responsible.

Colloid-facilitated transport of contami-nants has become a sort of Gordian knot forenvironmental scientists. Field studies haveoften been subject to sampling artefacts,such as colloid mobilization through exces-sive well-pumping rates, and it has beendifficult to reconcile field observations withtheory and model simulations. (In 1988, areviewer of a manuscript4, on which I was anauthor, implicating a colloidal intermediatein the scavenging of metals in marine sys-tems, complained: “The problem develop-ing in the literature with colloids is that theyare blamed or claimed for everything thatcan’t be explained”.) Since then, colloid-facilitated contaminant transport has gainedbroad acceptance. Nonetheless, few studieshave unequivocally demonstrated its signifi-cance in the field.

Figure 2 compares two- and three-phasecontaminant transport models. Figure 2ashows a low-solubility contaminant distrib-uted between an aqueous phase and immo-bile aquifer solids (the macroparticles). Col-loidal materials can increase the apparent sol-ubility of low-solubility contaminants (Fig.2b) if those contaminants strongly associatewith the colloids and the colloids minimallyinteract with the stationary macroparticlephase. Because colloid groundwater concen-trations are typically quite low, the contami-nants most likely to be transported by col-loidal materials are of extremely low solubili-ty and strongly partition to mobile non-aqueous phase materials. In this respect, Pu isan ideal candidate; others include pesticides(for instance, DDT and dieldrin), poly-nuclear aromatic hydrocarbons, other lightactinides (for example, thorium) and manytransition metals (such as cobalt).

Groundwater colloids originate from twosources5: through mobilization of existingcolloid-sized (1 nm to 1 mm) materials inaquifer systems following chemical pertur-

news and views

NATURE | VOL 397 | 7 JANUARY 1999 | www.nature.com 23

Perhaps nowhere have the details ofcontaminant transport in ground-water systems been more contentious

than in the area of nuclear waste disposal.Over the past decade, the discovery of col-loidal forms of actinides, such as plutonium(Pu), has often been at the centre of concernover underground storage of radionuclides.On page 56 of this issue1 Kersting et al. pro-vide an illustration of the striking influencecolloids2 may have on contaminant trans-port. The authors have studied groundwatermigration of Pu from a nuclear detonationsite in Nevada. However, the particular sig-nificance of their report lies in reinforcinga general awareness of colloid-facilitatedcontaminant transport.

As late as the early 1980s, a groundwatercontaminant was generally assumed to occu-py one of two phases: a stationary phase con-sisting of aquifer solids; and a mobile, aque-ous phase that serves as the medium for themovement of dissolved chemical species. Insuch a system, the rate of contaminant trans-port depends on the groundwater transportvelocity, of course, but also on the distribu-tion of the contaminant between the twophases. The greater the extent to which acontaminant partitions into the immobilephase, the slower its average transportvelocity in the ground water (Fig. 1).

The unexpected appearance of low-

solubility contaminants some distance fromknown sources, or sooner than would beexpected from their solubility, led to exami-nation of the possible involvement of non-aqueous, mobile colloids in contaminanttransport3. Invoking colloids to explain suchobservations gave rise to three-phase modelsof contaminant transport (Fig. 2, overleaf).The results of Kersting et al. reinforce theneed for such models. In their work, they useisotopic ‘fingerprinting’ to demonstrate thatPu, an element with extremely low aqueoussolubility (as low as 10117 M, depending on

Geochemistry

Colloidal culprits in contaminationBruce D. Honeyman

Dissolvedcontaminant

(mobile)

Sorbedcontaminant(stationary)

Water Water

a b

Macroparticle Macroparticle

Figure 1 Contaminant transport in a simple two-phase groundwater system. a, High sorption, andtherefore low contaminant solubility; b, low sorption, high contaminant solubility. Macroparticlesare the stationary components of a groundwater aquifer and include clays, metal oxides such asquartz, and particulate organic matter. The extent to which contaminant molecules are sorpted tomacroparticles (by adsorption, surface precipitation or absorption) regulates the rate of contaminanttransfer through groundwater systems for a given water velocity.

© 1999 Macmillan Magazines Ltd

bations or changes in flow velocity frompumping, and through in situ precipitationof supersaturated mineral phases. Colloidsare removed from the aqueous phase bydeposition on stationary macroparticles; it isthe efficiency of deposition that regulates thefacilitation of contaminant transport6.

Kersting and colleagues’ analysis1 of col-loids isolated from pumped ground waterclearly identified the source of the Pu as asingle underground nuclear test site, the testwell and detonation site being 1.3 km apart.The elimination of other detonation cavitiesand contamination as the source of Pumakes it evident that the Pu has been trans-ported through the ground water by someprocess.

As delineated by Ryan and Elimelech5,however, three conditions must be met fordefensible evidence that colloids have trans-ported contaminants: first, colloids must bepresent; second, contaminants must associ-ate with them; and third, the colloid–conta-minant combination must move throughthe aquifer. The results of Kersting et al. qual-itatively meet the first two conditions. But, asthe authors point out, the possibility of sam-pling artefacts meant that they could notquantify some of the parameters needed forsupporting the detailed assessment of col-loid-facilitated Pu transport in their study.That is, the third condition has not been rig-orously met. Nevertheless, their work clearlyshows that a low-solubility contaminanttravelled some way from the source, perhapsat or near the local groundwater flow velocity(1 – 80 m yr11).

Has the Gordian knot been cut? No, I donot think so, although a good slice has beentaken out of it. The fundamental difficultyremains the gap between field observationsand expectations based on bench-scale exper-iments and theory. For example, according toclassical filtration theory, colloid transportshould be relatively limited (tens of metres7 orless under typical subsurface conditions).Advances in incorporating8 macroparticlechemical heterogeneity and site blockage intomodels have helped to narrow the gap, but it

nonetheless remains wide.The trouble with most field studies is that

the systems are difficult to manipulate, andare often too large and heterogeneous tocharacterize accurately. There is a great needto develop meso-scale experimental systems(several metres to more than ten metres) forthe careful evaluation of the effects of systemheterogeneity on colloid transport and thetesting of methods for scaling up from thebench to the field.

Given that the work by Kersting et al.shows that Pu may be transported consider-able distances through groundwater sys-tems, can one conclude that the colloidaltransport of actinides provides a significantexposure pathway from nuclear testing andwaste sites? Not really. The very properties ofcompounds that make them good candi-dates for colloid-facilitated transport — lowsolubility and high particle reactivity —limit the amount of contaminant that can betransported: colloids are both the means andthe bottleneck. But we need to know more.T. H. Huxley9 had it that “It is the customaryfate of new truths to begin as heresies and toend as superstitions [dogma]”. In its evolu-tion from heresy to dogma, colloid-facilitat-ed contaminant transport has become a per-ceived truth, widely recognized, but rarelyunderstood in detail.Bruce D. Honeyman is in the EnvironmentalScience and Engineering Division, ColoradoSchool of Mines, Golden, Colorado 80302, USA.e-mail: bhoneyma@mines.edu 1. Kersting, A. B. et al. Nature 397, 56–59 (1999).

2. Buffle, J. & Leppard, G. G. Environ. Sci. Technol. 9, 2169–2175

(1995).

3. McCarthy, J. F. & Zachara, J. M. Environ. Sci. Technol. 23,

496–502 (1989).

4. Honeyman, B. D. & Santschi, P. S. J. Mar. Res. 47, 951–992

(1989).

5. Ryan, J. N & Elimelech, M. Colloids Surf. A 107, 1–56

(1996).

6. Roy, S. B. & Dzombak, D. A. Environ. Sci. Technol. 31, 656–664

(1997).

7. McDowell-Boyer, L. M., Hunt, J. R. & Sitar, N. Wat. Resour. Res.

22, 1901–1921 (1986).

8. Song, L. & Elimelech, M. J. Colloid Interf. Sci. 167, 301–313

(1994).

9. Huxley, T. H. in Darwiniana: Essays by Thomas H. Huxley

(Appleton, New York, 1896).

news and views

24 NATURE | VOL 397 | 7 JANUARY 1999 | www.nature.com

Colloid +contaminant

(mobile)

Water Water

a b

Macroparticle Macroparticle

Figure 2 Comparison of generalized two- and three-phase groundwater systems. a, Two-phase;b, three-phase. The third phase in b is a colloid or microparticle, shown here with contaminantmolecules sorbed to it, thus making them mobile. Colloidal material is usually chemically similar tothe stationary macroparticle phase.

100 YEARS AGOThe native arithmetic of Murray Island,Torres Strait, is described by the Rev.A. E. Hunt in the latest Journal (NewSeries, vol. I Nos. 1 and 2) of theAnthropological Institute. The only nativenumerals are netat (one) and neis (two).Higher numbers would be describedeither by reduplication, as neis netat,literally, two-one for three; neis-i-neis, ortwo-two for four, &c., or by reference tosome part of the body. By the lattermethod a total of thirty-one could becounted. The counting commenced at thelittle finger of the left-hand, thencecounting the digits, wrist, elbow, armpit,shoulder, hollow above the clavicle, thoraxand thence in reverse order down theright arm, ending with little finger or righthand. This gives twenty-one. The toes arethen resorted to, and these give tenmore. Beyond this number the term gaire(many) would be used. English numeralsare now in general use in the Islands.From Nature 5 January 1899.

50 YEARS AGOFertilized mouse ova have been cultivatedin vitro, and their development filmed, byFriedrich-Freska and Kuhl, who used asmedium a clot of guinea pig plasma andmouse embryo extract containingsegments of Fallopian tube. Like Chang, Ihave been working on ovum culture witha view to transplantation of ova. Changhas used rabbits, with serum as a culturemedium; I have chosen to use mice, asmore readily available, and because (likemost domestic animals) they have nakedeggs. Seeking a medium readily preparedin large quantities, I have tried salinehen-egg extracts. The procedure adoptedrevealed an unanticipated differencebetween the viabilities of two-cell andlater tubal stages; eight-cell ova survivedand developed in culture, whereas two-cell ova did not. ... A few becameblastocysts either completely separatedfrom the zona, or spherical and still half-enclosed in the split and distendedmembrane. … A physiological differencebetween two-cell and eight-cell stagesseems clearly established; the reason forthe difference remains obscure. Theresult recalls a similar one of Chang’s,namely, survival of morulae, but not two-cell ova, when transplanted into therabbit uterus, though both survived whencultivated in serum.From Nature 1 January 1949.