S63

Parasites, desiderata lists and the paradox of the organism

R. DAWKINS

Department of Zoology, University of Oxford

Key words: parasites, evolution, organism.

INTRODUCTION

Eavesdrop morning coffee at any major centre ofevolutionary theory today, and you will find 'para-site' to be one of the commonest words in thelanguage. Parasites are touted as prime movers in theevolution of sex, promising the final solution to thatproblem of problems, the puzzle that led G. C.Williams to proclaim in 1975 'a kind of crisis' athand in evolutionary biology (Hamilton, 1980;Tooby, 1982; Seger & Hamilton, 1988). Parasitesseem to offer a plausible justification for theotherwise futile effort females put into choosingamong posturing males (Hamilton & Zuk, 1982; butsee Read, 1990). Frequency-dependent selectionexerted by parasites is, according to one admittedlyminority view, largely responsible for the high levelsof diversity found in gene pools (Clarke, 1979). Onemight even extrapolate to a time when the entiremetazoan body could come to be seen as a giganticadaptation against microscopic pathogens.

I want to use parasites in an entirely different kindof evolutionary argument, almost a philosophicalargument. Parasites have a role in clarifying the verymeaning of that most basic unit in the hierarchy oflife, the organism itself (Dawkins, 1982, 1989).Parasites help us to think straight about the problemthat I shall call the 'Paradox of the Organism'.Much of the background to what I shall say isdiscussed at greater length, as part of a com-prehensive vision of life, in my book The ExtendedPhenotype. I shall not bother with further detailedcitations of this book.

To show the direction in which I am going, let meanticipate my answer to the question 'What is theorganism?' It will turn out that an individualorganism is an entity all of whose genes share thesame stochastic expectations of the distant future. Ishall begin by showing one thing that the organismis not, though it is usually thought to be, namely aunit of natural selection.

THE ORGANISM IS NOT AN 'OPTIMON1'

I have used the word 'optimon ' for that kind of unitin the hierarchy of life about which we may say:' Such and such an adaptation is for the benefit ofthat unit. ' For instance, a group selectionist might

argue - wrongly - that the species is an optimon,meaning that there are adaptations that are for thegood of the species. I have joined others in makingthe case that the optimon must be a self-replicatingentity, and that therefore even the organism is not atrue optimon. Genes, and to a lesser extent largerfragments of genomes, are true self-replicatingentities. Organisms, even asexually reproducingorganisms, are not.

Examine the logic of modern natural selectiontheory in sexual populations and observe that it is allabout changes in frequencies of copies of things. Inpractice these things are genes, in the sense ofMendelian units independently assorting in genepools, but any self-replicating entity would do inprinciple. In the perspective of evolutionary time,the genes inhabit not a pool but a river, flowingthrough time, a broad-fronted turbulent river downwhich the Mendelian particles zig-zag their way,changing partners at every generation, some in-creasing in frequency, others decreasing. Successfulgenes are those that become more frequent, un-successful ones those that become less frequent. Butindividual organisms do not have a frequency at all,or rather each individual has a frequency of one.This is one reason why genes, but not organisms, canbe optimons.

The other reason is more important, and it applieswhether reproduction is sexual or asexual. It is thatgenes, but not organisms, form replicating lineagesin which copying errors are passed in one directionalong the lineage. Even in an asexual lineage oforganisms, for instance a mother to daughter togrand-daughter lineage of asexually reproducingDaphnia, the organisms are not true replicas of theprevious generation. T o be sure, they resemble theprevious generation and may even be indistinguish-able from the previous generation. But identity is notthe same thing as heredity: heredity has causaldirectionality. Clonal organisms resemble each otherin the same sense as 100 copies of a book run off thesame printing press resemble each other. Maybe youcan't distinguish one copy from another, but theyhave not derived their characteristics one fromanother. All have derived their characteristics fromthe same parent, the same printer's block. In order tosimulate true heredity you would take one copy ofthe book and Xerox it. Then Xerox a copy of the

Parasitology (1990), 100, S63-S73 Printed in Great Britain

R. Dawkins S64

copy, then a copy of that copy and so on. If you made100 copies of a book by that method there would bea true hereditary succession among the copies:number 49 would be the daughter of number 48 andthe mother of number 50, and so on.

The operational test of whether we have trueheredity is to examine the fate of copying errors. Ina lineage of Daphnia, blemishes to genomes areinherited by subsequent members of the lineage;blemishes to bodies — lost limbs, for instance — arenot. In terms of the book analogy, what is passedfrom mother to daughter is not a body but theprinter's block that made the body. Each body isindependently run off a copy of the same printer'sblock, not run off any other body. Needless to say,this isn't a new point. I am simply reiterating theWeismannian dogma of the continuity of the germline and the non-inheritance of acquiredcharacteristics.

Organisms, then, are not replicators. So, what arethey ? The answer is that they are vehicles forreplicators, built by a cooperative of independentlyassorting replicators. As such, they are extremelyimpressive units of function, but they are, never-theless, not replicators. The organism is not anoptimon. We should strictly never say of anadaptation such as a wing or a behaviour pattern thatit is for the benefit of the organism. Nevertheless,organisms are such coherent units of function that itis extremely tempting to see organisms as units thatwork to maximize something on behalf of all theirgenes. Their existence as cohesive units of function,in the teeth of potential conflict among the truereplicators that they contain, constitutes a paradoxwhich I shall call the Paradox of the Organism.

THE PARADOX OF THE ORGANISM

The paradox of the organism is that it is not tornapart by its conflicting replicators but stays togetherand works as a purposeful entity, apparently onbehalf of all of them. Not only is it not torn apart; itfunctions as such a convincingly unified whole thatbiologists in general have not seen that there is aparadox at all! They have wrongly taken theorganism for granted as the unit about whichquestions of adaptation should be asked.

It is possible to do the mathematics of naturalselection completely forgetting that, as a matter offact, the genes are not swishing about in a liquid poolor river but bound up in solid chunks - organisms- and colossal chunks at that. The gene inhabitstwo time-scales, two worlds, corresponding to itstwo roles. In its evolutionary role the gene inhabitseternity, or at least geological time. Its companions inthe river of evolutionary time are other genes, andthe fact that in any one generation they inhabitindividual bodies can almost be forgotten.

But in its other role, its embryological role, the

world of the gene seems bounded by the skin of aparticular individual organism. Its companions arethe other genes that happen to make up the genomeof that organism. And it cooperates with those othergenes to produce a huge — by gene standards— machine, bristling with apparently purposefultechnical wizardry. To repeat, the paradox of theorganism is the paradox that organisms are suchmarvels of cooperative engineering; yet thecooperating genetic replicators that build them, intheir other role as denizens of deep time, would seemto have every reason not to cooperate, every reason tocheat on their temporary partners and steal a marchdown the next reach of the evolutionary river. Why isthe organism not torn apart by the conflictinginterests of the multitude of self-interested units thatit contains ?

The potential for tearing apart is ever present. It isstarkly demonstrated by the phenomenon of MeioticDrive, and rather more subtly made manifest in thechain of reasoning that leads to the doctrine of the'extended phenotype'. I shall take these in order.

MEIOTIC DRIVE

Meiotic drive is well known and I can be brief. Innormal meiosis, each member of an allelic pairenjoys an equal chance of getting into each gamete.Meiosis is a biological process like any other, and itcan come under genetic influence like any other. If agene happens to arise whose phenotypic effect is tobias meiosis, so as to give itself more than the usual50 % probability of getting into each gamete, thatgene will tend to spread through the gene pool, evenif, as is usually the case, its other effects aredeleterious. The phenomenon really exists, and it iscalled meiotic drive. The best known meiotic drivegene is the Segregation Distorter gene in Drosophila(Ciow, 1979). Now, one way of expressing theparadox of the organism is this. Why aren't all genessegregation distorters or worse? Why, instead, domost genes submit to the discipline of cooperatingwith one another in building a shared phenotype ?

THE EXTENDED PHENOTYPE

I usually demonstrate the extended phenotype ineasy stages, beginning with the demonstration thatan animal artifact, like a bird's nest, is a phenotypelike any other, under the control of the animal'sgenes in exactly the same sense as the shape of itsbeak is under the control of its genes. So, there canbe genes 'for' nest shape, nest colour and so on, inexactly the same sense as there are genes 'for' taillength or eye colour. I then go on to parasitesmanipulating their hosts, and demonstrate that themodification to the host, whether it is morphological,physiological or behavioural, can be seen as pheno-typic expression of the parasite's genes. The final

The paradox of the organism S65

stage in the argument, the idea of genetic ' action ata distance', again uses parasites, this time parasitessuch as cuckoos whose manipulation is achieved byremote control. By the same logic as before, themodification to the host's behaviour can be seen asphenotypic expression of the parasite's genes. The'central theorem of the extended phenotype'(Dawkins, 1982) is this:

An animal's phenotype tends to maximize the survival ofthe genes ' for' that phenotype, whether or not those geneshappen to be in the body of the particular animalperforming it.

Here I will leave out the artifacts and go straight tothe parasites. The literature on parasites manipu-lating their hosts has been reviewed several timesrecently (Holmes & Bethel, 1972; Ewald, 1980;Dobson, 1988; Keymer & Read, 1990; Moore &Gotelli, 1990; the medical and epidemiologicalimplications have been reviewed by Ewald,1980).

A typical and favourite example is the flukeDicrocoelium dendriticum, whose ungulate definitivehost needs to eat its ant intermediate host in order forthe worm to complete its life-cycle. This aptly named' brainworm' burrows into the suboesophagealganglion of the ant and, significantly, the ant'sbehaviour changes. Infected ants climb to the top ofgrass stems at a time of day when normal ants wouldretreat underground. There they clamp their jaws inthe stem and remain as if alseep, immobile andvulnerable to being grazed by ungulates. This story,like many others, has long been treated as a plausiblecase of parasites manipulating their hosts for theirown advantage (Wickler, 1976; Love, 1980). All thatI am adding to this familiar point is that it must bethe parasite's genes that are doing the manipulatingand that, if you examine what it ever means to talk ofgenes as having phenotypic expression, it followsthat the parasite's genes are having phenotypicexpression in the host's body.

I shall return to this logical argument aboutparasite genes having extended phenotypic effectsupon host bodies. Meanwhile I must concede that itdepends upon the assumption that the host really isbeing manipulated for the benefit of the parasite:that the host's altered phenotype really is anadaptation for the benefit of the parasite. It isn'talways easy to be sure whether a phenotype is anadaptation at all, or simply a byproduct, and thequestion of parasites affecting hosts is indeed one ofthe areas in which this is disputed. This topic needsa digression. It is also possible in some cases that thephenotype is an adaptation, not for the benefit of theparasite but for the benefit of the host in combatingthe parasite. I shall mention this at the end of thedigression.

ADAPTATION OR BORING BYPRODUCT?

In many cases, undoubtedly, parasites benefit bytheir intermediate hosts's being devoured by theirfinal host. Moreover, intermediate hosts are often, asa matter of fact, more likely to be eaten by finalhosts if they contain a parasite than if they don't. Itis tempting, therefore, to believe that this desirableoutcome has been engineered by the parasite. Butthere is always an alternative, killjoy explanation, the'boring byproduct' theory. Parasites are naturallyexpected to have a debilitating effect on hosts, andthis obviously could make them less nimble in flightfrom predators. So the apparent adaptation could bemerely a side-effect.

Even the boring byproduct theory allows thepossibility that natural selection has, indeed, workedon parasites to increase the extent to which they makehosts feel ill, because of the benefit to the parasite.But we are going to be impressed only in those caseswhere the intermediate host is made to do somethingthat we should not obviously expect tin the boringbyproduct theory. As in the case of any scientifictheory, its predictions impress us only to the extentthat they are counterintuitive. It is not a theory'sfault if its predictions happen to go along withcommon sense, but nevertheless it can't expect us tobe impressed ! It is no use, moreover, multiplying upthe number of different examples showing the samekind of thing, if all the examples predict the samecommonsensical principle.

What evidence would impress us ? There arebasically two kinds of effects on hosts that mightpersuade me that they are true parasitic adaptations.The first possibility is that, far from being made ill,the host might have some aspect of its life apparentlyimproved by the parasite. (Of course it is important tostress 'some aspects' and 'apparently', otherwise webeg the whole question of whether the parasite is aparasite at all.) Thus, instead of being made smallerby a parasite, hosts are sometimes made larger. Thisis often true in those cases known as 'parasiticcastration' (Reinhard, 1956; Baudoin, 1975; Moore& Gotelli, 1990). And Noble & Noble's (1976)textbook describes the following case:

Some species of carpenter ants infected with metacercariaeof the fluke Brachylecithutn mosquensis, are more obese thannoninfected ants and, unlike the latter, they do not concealthemselves but crawl on exposed surfaces where they areeasily found by birds that are the next hosts of the fluke.This behaviour seems to be a remarkable example of ananimal that sacrifices its life for its parasites. The'sacrifice', of course, is induced by the parasite.

Presumably the fluke benefits, not only by makingthe ants crawl in exposed locations where birds cansee them, but also by making the ants fatter and moretempting targets for the birds. The economicresources used by the ants to grow larger are not, of

R. Dawkins S66

course, provided by the parasite. The resources mustbe robbed from other uses in the ant colony to whichthey might have been put: perhaps larvae wenthungry as the infected ants were induced to eat moreand wax fat.

Turning to behaviour, sometimes parasites, farfrom making their hosts seedy and listless, seem topep up their activity levels. Moore & Gotelli (1990)note that acanthocephalans often increase the activitylevels of their intermediate hosts. An example of thisthat I have used before is from the work of Betheland Holmes on the behaviour of gammarids infectedwith Polymorphus paradoxus and P. marilis (Holmes& Bethel, 1972; Bethel & Holmes, 1973, 1977). Thetwo species of acanthocephalans are 'a iming' atdifferent definitive hosts, and they appear to changethe behaviour of their gammarid intermediate hosts,in opposite directions that increase their vulner-ability to predation by their respective definitivehosts. In neither case does this change in behaviourinclude an impairment in the activity level of thegammarids.

The same is true of many examples of insectvectors of blood parasites changing their behaviourwhen infected. Thus tsetse flies infected byTrypanosoma probe more frequently and feed morevoraciously than uninfected control flies (Jenni et al.1980). And bumblebee queens infected by thenematode Sphaerularia bombi 'flew almost incess-antly, stopping to dig, then moving on. In theprocess, they deposited nematode larvae' (Moore &Gotelli, 1990). Giles (1983) and Milinski (1985)reported that sticklebacks infected with the tape-worm Schistocephalus solidus were less fearful in thepresence of predators than uninfected controls.

From our point of view, parasites that make hostslarger rather than smaller, or more active and lessfearful, are interesting because it is harder to writethe effect off as a boring byproduct. The same is truewhere the host is induced to do something un-expected by commonsense, something bizarre. Oneof my favourite illustrations concerns nematomorphlarvae who need to break out of their insect hosts andreturn to water:

...a major difficulty in the parasite's life is the return towater. It is, therefore, of particular interest that the parasiteappears to affect the behaviour of its hosts, and' encourages' it to return to water. The mechanism bywhich this is achieved is obscure, but there are sufficientisolated reports to certify that the parasite does influenceits hosts, and often suicidally for the host...One of themore dramatic reports describes an infected bee flying overa pool and, when about six feet over it, diving straight intothe water. Immediately on impact the gordian worm burstout and swam into the water, the maimed bee being left todie (Croll, 1966).

Much as I enjoy this anecdote, I can't help worryingabout it. How was the observer (not identified in the

book) able to see something so small as a gordianworm bursting out of a bee ?

The second type of case in which the boringbyproduct theory seems less plausible is where theparasite achieves a detailed fit to some complicatedaspect of its host's physiology, something toostatistically improbable to have come about bychance. A beautiful example, quoted by Keymer &Read (1990) concerns a fungus called Moniliniavaccinii-corymbosi, the 'Mummy-Berry fungus',which infects blueberries and is economically quiteimportant (Batra & Batra, 1985). The fungus inducesthe blueberry plant to grow fake flowers. Insects visitthese fake flowers, pick up the fungus's asexualreproductive conidia and transport them to realflowers. The conidia infect the host flowers andinduce them to produce not normal blueberries butseedless, inedible 'mummy-berries' . The fungusoverwinters in the mummy-berries. The fake flowersinduced by the fungus are impressive in their detail,which is why I bring the example up here. Theybecome reflective in ultraviolet light which isattractive to insects, make a smell, and secrete nectar.It seems hard to write off such a cluster of apparentlyspecific adaptations as a boring side-effect of parasiticinfection. The case for detailed parasite manipulationseems strong.

My favourite example along these lines is onereported by Fisher in 1963. As it happens, this isanother case where the host is induced to becomelarger.

Tribolium larvae infected with Nosema sp. attain larger sizethan uninfected controls. Infected larvae undergo as manyas six supernumerary molts and most commonly die asgiant larvae, weighing two times as much as nonpara-sitized controls (Fisher, 1963).

Fisher went on to describe experiments on Blaberuscockroaches.

From these experiments it was concluded that the parasiteNosema produced a substance with juvenile hormone ac-tivity and sufficed to replace or augment that produced bythe corpora allata of the host.

Cheng remarked in his textbook (1973):

If the situation in Tribolium is similar to that in Blaberus,and there is no reason to believe that it is not, then theenhanced growth in parasitized Tribolium can beattributed, directly or indirectly, to the contribution of ajuvenile hormonelike material by Nosema.

Presumably a giant larva is a more bountiful sourceof food than an adult half its size. So, the parasitewould seem to have something to gain by themanipulation. Now to the point that this example ismaking. The juvenile hormone molecule seems fartoo improbable a specific juxtaposition of atoms fora protozoan to hit upon as a boring, accidental by-product of something else. The case seems strong

The paradox of the organism S67

that it has evolved as a specific adaptation tomanipulate the physiology of the host.

Another example, though here the functionalsignificance from the parasite's point of view is lessclear, is the effect of the tapeworm Spirometramansonoides on rats. Under experimental conditionsan infected rat may grow to nearly eight times itsnormal weight. It seems to be due to the secretion,by the parasite, of a growth-hormone-like substance.Here it has been suggested that the tapeworm mayhave ' borrowed' a mammalian, even specificallyhuman, gene and is using it to manufacture amammalian hormone (Phares, 1987). Whether this isso, or whether it has independently evolved theability to synthesize the hormone, we can hardlywrite off the parasite's effect on the host as a boring,accidental side-effect.

But in any case, as Darwinians, the word accidentshould give us pause. Do not all adaptations start offas accidents, random mutations ? What is thedifference between an accidental side-effect thathappens to benefit the parasite (which I am calling anon-adaptation) and a favoured mutation (which Iam calling an adaptation) ? The solution to thisconundrum lies in the distinction between singlestep and cumulative selection (Dawkins, 1986). Asingle step of natural selection cannot produce anadaptation sufficiently complex to impress us asbeing obviously an adaptation. All the undisputedadaptations, from the aerodynamics of a swift's wingto the acoustics of an owl's ear, impress us becausethey fit their purpose in many different respectswhich could not have come about in a singlemutational step. They have evolved their efficiencycumulatively, in series, over many steps of accidentalmutation followed by non-accidental selection. Atleast some alleged cases of parasites manipulatingtheir hosts seem to meet the criterion: it really ishard to imagine that the synthesis of insect juvenilehormone by the protozoan Nosema is anything otherthan a highly specific Darwinian adaptation thatmust have taken many cumulative steps of naturalselection to perfect.

The same seems to me true of some of thebehavioural effects of symphylic caterpillars in ants'nests. Often they bristle, literally, with equipmentfor manipulating their protectors. The caterpillar ofThisbe irenea has a sound-producing organ in itshead which apparently has no purpose other thansummoning ants (De Vries, 1988). Near the insect'srear end is a pair of telescopic spouts which exudeseductive nectar. On its shoulders stand another pairof nozzles secreting not food but a volatile substance,which has a dramatic impact upon the ants' be-haviour. An ant coming under the influence of thischemical leaps clear into the air. Its mandibles gapeand it turns aggressive, far readier than usual to biteor sting any moving object except the caterpillar

responsible for drugging it. Moreover, an ant underthe sway of a manipulating caterpillar eventuallyenters a state called 'binding', in which it becomesinseparable from its caterpillar for a period of manydays. Once again, the complex of elaborate organs inthe caterpillar, and the complex of specific effects onthe ant's behaviour, seem to leave no doubt thatnatural selection, for the benefit of the parasite, hasbeen at work, cumulatively over many generations.

Before leaving the digression on boringbyproducts as an alternative to parasitic adaptation,I must also deal with the third possibility, hostadaptation. Raised temperature in response to in-fection is just the kind of effect that might tempt oneto ingenious stories about parasites manipulatinghosts. No doubt a feverish brain does all sorts ofthings that make its unfortunate possessor morelikely to be eaten by predators, and in some cases thismight benefit the parasite. There seems good reasonto believe, however, that fever is an adaptation byhosts to create unfavourable conditions for parasites;see, for example, Boorstein & Ewald (1987) whosepaper has the added attraction of including anexperimental investigation of protozoan-inducedfever in grasshoppers.

I must also mention that, since The ExtendedPhenotype was published, one of its main examplesof parasite manipulation has been claimed to be ahost adaptation. Snails are frequently induced bytrematode parasites to grow larger, and specificallyin some cases to develop thicker shells. Followingthe logic above, this seems temptingly to argue astrong case for parasite manipulation. A smallersnail, or a thinner shell, could have been put down togeneral debilitation, a boring side-effect. But athicker shell must surely mean that something moreinteresting is going on. The parasite, I suggested, isselfishly prolonging the life of the individual snail inwhich it lives, at the expense of the snail's re-production.

The theory makes use of the idea of an economictrade-off. From the snail's point of view, I suggested,there is an optimal shell thickness, a compromisebetween the protective benefits conferred by a thickshell and the economic savings in calcium and otherresources conferred by a thin one. If snails wouldreally benefit from thicker shells they would growthem without being prompted by flukes. I regardedthe compromise as one between reproduction andindividual survival. A thicker shell would favoursurvival of the individual snail but would exacteconomic costs that would reduce its reproductivesuccess. The parasite has the same interest as thesnail in the snail's survival, but it has no interest inthe snail's reproducing itself. From the parasite'spoint of view the optimum shell thickness is shiftedtowards favouring individual snail survival at theexpense of snail reproduction.

R. Dawkins S68

Minchella (1985), however, put forward analternative theory. His paper discussed the phenom-enon of parasite-induced gigantism and parasiticcastration in general, and listed three hypothesesabout it. Firstly, the boring byproduct theory:

The traditional view of gigantism has been that it is a side-effect or non-adaptive consequence of parasite-inducedcessation of reproduction ... The energy which was to beused for reproduction becomes available to the host forincreased growth.

Secondly, he took the parasite adaptation theory,which he attributed to Baudoin (1975) and to me. Heremarked that both these first two hypotheses regardthe host as 'an unresponsive partner whose actionsare dictated by the actions of the parasite'. He wassceptical and, as a third option, put forward his owntheory. Gigantism, parasitic castration, and pro-longed life at the expense of reproduction generally,are, Minchella suggested, adaptations on the part ofthe host. The host temporarily switches resourcesinto fostering its own longevity, at the temporaryexpense of its reproduction, in an attempt to outlivethe parasite and return to reproduce at a later date.

Minchella and three colleagues attempted to testbetween the three hypotheses, using the specific caseof trematode effects upon snail shells (Minchellaet al. 1985). Rather surprisingly, since these resultsappeared in the same year as Minchella's hypothesiswas published, they found against that very hy-pothesis : it turned out that the snails cannot outlivethe parasites anyway. They concluded in favour ofthe parasite manipulation hypothesis of Baudoin andmyself.

This is gratifying, though I take passing exceptionto the label 'prudent parasite' that they attach to theparasite manipulation hypothesis. They say that it isequivalent to Slobodkin's (1961) 'prudent predator'theory. Slobodkin's theory notoriously was based onan uncritical group-selectionism that was egregiouseven by the standard of ecologists. Predators weresupposed to restrain their hunting in the interests ofconserving stocks for the future. This does not quitehave to be group selectionist: we could assume thateach individual predator has exclusive hunting rightsover a geographical area, or in some other waysecures privileged access to the benefits of its ownprudence. But there was no indication thatSlobodkin was aware of this need, and 'prudentpredators' have rightly had a bad name among theDarwinian cognoscenti ever since.

The parasite manipulation hypothesis is alsoshored up by indirect evidence of its underlyingeconomic assumption of trade-offs between snailshell thickness and reproductive rates. Hamilton(1980) looked at 5 snail species coexisting in thePotomac river, Physa heterostropha, Bithyniatentaculata, Helisoma trivolvis, Mudalia carinata andGoniobasis virginica. She measured the ability of the

shells of each of these species to withstand attacks byduck and crayfish predators, and correlated it withreproductive rate measured as the intrinsic rate ofincrease. She found a negative association acrossspecies between effectiveness of armour and re-productive rate. Although it is always risky to drawconclusions about trade-offs from across-speciescomparative studies (the 'other things being equal'assumptions are so fragile), her data are undeniablyexactly what I should wish them to be. Herconclusion is aptly summed up in her title:' Reproduction or shell armour - a trade-off in fresh-water gastropods'.

For the purpose of developing my argumentfurther, any example of parasite manipulation wouldserve. I'll use the snail shell thickness example on theassumption that it really is an authentic case ofparasite manipulation. If you do not find thisexample plausible, simply substitute any case whereyou are prepared to accept that a parasite has evolvedto manipulate a host's phenotype for the parasite'sbenefit.

EXTENDED GENETICS

Let's assume, then, that flukes really are manipu-lating snails into growing larger shells, and translatethe story into the language of the extended pheno-type. Essentially, this means taking a gene's-eye-view.

Any Darwinian adaptation comes about throughthe non-random survival of genes in a gene pool.There must have been genetic variation in tendencyto produce the adaptation of interest: some animalshad it, others didn't, because of differences in theirgenes. The ones that had it were more successful,and so more copies of the relevant genes were passedinto future gene pools. This process continued untilthe phenotypes associated with these genes becamethe norm. In this case, since it is a fluke adaptationwe are talking about the genes are in the fluke genepool. But the phenotypic effects that we are talkingabout are manifested in the snail's body. It is thesnail's shell, not any part of the fluke's body, that isthe phenotype of interest.

Of course the fluke genes work proximally viaphenotypic effects on the fluke's body. But this doesnot nullify the conclusion that these fluke genes arealso exerting phenotypic effects on the snail's body.Any conventionally recognized phenotypic effect of agene is the end of a cascade of earlier embryologicaleffects, in this case all within the same body. A gene' for' waltzing behaviour in mice presumably begins,like most other functional genes, by influencing thesynthesis of a protein. This affects something elsewhich affects something else which, further downthe causal cascade, distorts the embryonic devel-opment of the balance organs in the inner ear. Thisis the penultimate step in the cascade, and it leads to

The paradox of the organism S69

the 'ultimate', behavioural effect that we actuallyobserve, namely waltzing behaviour.

But I pu t ' ultimate' in inverted commas advisedly.There is nothing to stop us recognizing any stage inthe cascade as ' t he ' phenotypic effect of interest.Instead of assaying phenotype behaviourally, ageneticist could dissect the ears of specimens andstudy the genetics of abnormal inner ear anatomy.Or a biochemical geneticist could detect the ab-normal protein product of the 'waltzing' gene, anduse that as its phenotypic marker. Similarly, there isno necessary reason why waltzing behaviour shouldbe regarded as the last step in the cascade. Waltzingmice might be especially vulnerable to predation bycats. The gene might therefore be labelled, not the'waltzing gene' nor the 'abnormal inner ear gene'but the 'vulnerable to cats' gene. Which label wechoose will depend upon which step in the cascade ofcausal effects seems to be the most salient in the lifeof the animal, or the most relevant to the interests ofthe geneticist.

Similarly, fluke genes must exert their effect onsnail shells via a cascade of prior effects within thefluke body, but this does not prevent us from seeingthe cascade as extending on outside the fluke's body.To say that fluke genes have phenotypic effects onsnail bodies is to make exactly the same kind oflogical extension as we are accustomed to makinganyway. It is just a little startling because it is anunfamiliar idea. Nevertheless, the logical correctnessof the principle is inescapable.

I imagine three geneticists examining the samedata and analysing it in three ways. The first is afluke geneticist. He observes that some flukes secretea chemical and others do not. This chemical affectsthe physiology of the host snail, and induces it tobuild a thicker shell, diverting into this projectresources that it would have preferred to use forother purposes such as reproduction. The flukegeneticist studies this phenotype in successivegenerations of flukes, and observes that it breedstrue, or at least has a hereditary component. For thefluke geneticist, the snail is just part of the en-vironment, like the pond in which it lives.

The second geneticist is a snail geneticist. Heobserves that some snails have thicker shells thanothers and he looks to see if the characteristic breedstrue over generations of snails. He finds that it doesnot; or at least that there is also a substantialenvironmental component to the variation which heidentifies as associated with the presence of flukes.As far as the snail geneticist is concerned, the flukesconstitute environmental noise, comparable to thenoise that might be introduced by differing calciumlevels in the water of different ponds.

The third geneticist — and the one whose approachI commend - is an extended geneticist. She looks atvariation in snail shell thickness down pedigrees ofsnails (with different flukes) and down pedigrees of

flukes (in different snails). The phenotype that she isstudying seems unambiguously a snail phenotype.But it turns out to vary under the influence of bothfluke genes and snail genes. The two sources ofgenetic influence interact with one another in just thesame kind of way as we conventionally see differentgenes from the same genome interacting with oneanother. Most phenotypes are influenced by a largenumber of interacting genes, often pushing indifferent directions. The final phenotype is acompromise between these shared genetic influences,and these interact too, with non-genetic'environmental' influences.

In the present case the fluke geneticist admitscomplicated interactions and compromises amongfluke genes, but consigns the snail genes to'environmental noise'. The snail geneticistacknowledges complicated interactions andcompromises among snail genes, but relegates thefluke genes to environmental noise. The extendedgeneticist achieves a more penetrating analysis. Shesees interactions among fluke genes, interactionsamong snail genes, and interactions between flukeand snail genes, all bearing upon the same, sharedphenotype. All these factors, of course, also interactwith environmental variables. For the extendedgeneticist the category 'environmental noise' issmaller. It includes the calcium content of the pondbut does not include the genes of the fluke or thegenes of the snail. These have been taken out of theenvironmental residue, into the fold of geneticallyexplained variation.

Later in her extended genetic analysis, the ex-tended geneticist might investigate the possibility ofhijacking even more sources of variation out of thecategory of 'environmental noise' and into thecategory of 'extended genetic influence'. Perhapseven the calcium content of the pond is under thegenetic influence of some animal or plant, andnatural selection is working on these genes becausetheir phenotypic effects feed back and influence theirown welfare.

There is, indeed, no obvious limit to the distanceover which extended genetic influences may travel. Abeaver dam is an undoubted adaptation. Exactly howit benefits beavers is not clearly understood, but itpresumably has something to do with the lake that itcreates. Part of the story is probably that the lakeprovides a secure and convenient route fortransporting logs. Whatever the exact advantages,beaver lakes are certainly adaptive phenotypes thathave evolved through the natural selection of beavergenes. There must have been genes 'for ' lakes ofvarious kinds, or natural selection would have hadnothing to work on. Presumably there were genesfor lake size, genes for lake shape, genes for lakedepth, and so on. The fact that all these genes musthave worked via influences on the building behaviourof beavers is irrelevant to the logical point being

R. Dawkins S70

made. In just the same way as it is legitimate to speakof genes for beaver behaviour, even though thisphenotype is a product of some prior effect, say onneuroanatomical wiring, so it is in principle legit-imate to speak of genes for round lakes versuselongated lakes. The point of introducing beaverdams as extended phenotypes is that they can be veryextended indeed - several square miles!

Returning to parasites, some of the most strikingexamples of extended phenotypes are provided bythe galls that plants are induced by insects to make.I have previously quoted Mayr on the subjectbecause, as long ago as 1963 galls moved him to uselanguage highly congenial to the extended phenotypethesis. A recent paper on the subject makes explicituse of extended phenotype terminology (Weis,Walton & Crego, 1988):

The evolution of the plant-gall interaction is complicatedby the fact that galls are phenotypic entities that developunder the influence of both plant and insect genotypes.From the plant's perspective, the gall is a developmentalabnormality, induced by an environmental stimulus, i.e.the insect. When viewed from the insect's perspective, thegall is a phenotypic extension. Selection may act upon theinsect to alter this extended phenotype, and thus gallformation is an adaptation of the insect.

Galls also provide an excellent example of anadaptation too complicated to be written off as aboring pathological byproduct:

Unlike microbe-induced plant tumours with theirunstructured cell proliferation, insect galls are' harmoniously organized entities with an orderly arrange-ment of cell layers and determinate growth which resultsin structures with particular size and shapes'. New,localized developmental gradients emanating from the gallmaker take over plant cell differentiation. Concentric tissuelayers, with nutritive and protective function, differentiatearound the insect. The information for this developmentalprocess undoubtedly lies in both the insect and plantgenomes (Weis, Walton & Crego, 1988).

Meiotic drive and the extended phenotype, then,from their different directions, serve to emphasizethe dilemma that I am calling the paradox of theorganism. The phenomenon of meiotic drivereminds us that the organism is a cooperative ofentities that — one might have thought — would beconstantly at civil war. It is a highly successful unitof pacification of fundamentally warring particles.How does it do it? Or, to phrase the question at thegene level where it should be phrased, why do thegenes cooperate so spectacularly ? Why don't they allbehave like segregation distorters ?

Meanwhile the doctrine of the extended phenotypereminds us that there is no necessary reason whyorganisms should exist in the first place. Genessurvive in the gene pool by virtue of their phenotypiceffects. But those phenotypic effects do not necess-arily have to be bundled up into discrete organismal

vehicles. The phenotypic effects by which a genelevers itself into future gene pools comprise all ofthat gene's effects on the world. It is only acontingent fact, not a necessary one, that so manyphenotypic effects of genes are, as matter of fact, tiedtogether in the gigantic and ingenious clusters thatwe call individual organisms. The paradox of theorganism remains, and the time has finally come toshow how parasites help us to solve it.

BLURRING THE BOUNDARIES

Thinking about parasites can help us to solve theparadox of the organism because parasites blur thepractical dividing line between one organism andanother, between themselves and their hosts. Byexamining this practical blurring between organismswe can sharpen up our theoretical perception of whatit means to be an organism at all.

Smith (1979) used the vivid simile of the CheshireCat's grin to dramatize the blurring of boundariesbetween parasite and host.

In non-living habitats, an organism either exists or it doesnot. In the cell habitat, an invading organism can progress-ively lose pieces of itself, slowly blending into the generalbackground, its former existence betrayed only by somerelic. Indeed, one is reminded of Alice in Wonderland'sencounter with the Cheshire Cat. As she watched it, 'itvanished quite slowly, beginning with the tail, and endingwith the grin, which remained some time after the rest ofit had gone'.

Smith was talking about intracellular symbionts,including putatively ancient ones such as mito-chondria. Here the blurring is so complete that thehistorical separation of the parties may become allbut undetectable. But there is a kind of blurringbetween parasite and host that can arise even thoughtheir tissues may not physically blend. There is ateleonomic blending — a blending of functionalinterests — which may or may not be accompanied bya blurring of physical boundaries.

OVERLAPPING AND NON-OVERLAPPING

DESIDERATA LISTS

Think about the conflict of interest between aparasite's genes and the genes of its host, and notethat the conflict is only partial. If you make a listof those outcomes in the future that would benefitthe host's genes, and those that would benefit theparasite's genes, they diverge in many ways, but thereis a partial overlap between them. Both sets of genes'want ' the host to go on living, at least for a while,because the host's body is the present vehicle forboth sets of genes. But they may differ in how longthey want it to go on living. The parasite's genes mayrequire the host to be eaten by a definitive host, whilethe host's genes would prefer this not to happen. Insuch a case, selection on host genes is pushing the

The paradox of the organism S71

shared phenotype towards increasing longevity,while selection on parasite genes is pushing ittowards decreasing longevity. In other cases, such asthe snails discussed above, the host's genes may beselected to divert resources towards reproduction atthe expense of survival, so tending to reducelongevity, while the parasite's genes are pushing it tosurvive longer at the expense of reproduction.

But those are details, specifics. My general point atthe moment is that there is a partial overlap ininterests between the two sets of genes. We could, inprinciple, make a list of events in the future that bothsets of genes would 'want'. There will also, ofcourse, be a list of events in the future that only thehost genes want, and another list that only theparasite genes will want. Let us call these two the'host desiderata list' and the 'parasite desideratalist', and we'll call the overlap between the two liststhe 'shared desiderata list'. For completeness weshould note that in theory the shared list includesfuture outcomes that we take for granted like thesun's continuing to shine, but we are specificallyconcerned with outcomes over which the genes canhave some phenotypic influence. The doctrine of theextended phenotype has taught us not to ignore theparasite desiderata list where the host phenotype isconcerned, for the parasite genes can have aninfluence over the shared phenotype just as the hostgenes can. Now, what kinds of factors will affect thesize of the shared list in comparison to the other twolists ?

I believe that it is possible to give a powerfulgeneral answer to this question. The answer is thatparasite genes and host genes will agree on the samedesiderata list to the extent that they use the sameroute into future generations. What are the routes ofa gene into future generations ? Sperms, eggs, spores,runners, suckers, airborne capsules of DNA, what-ever vehicles are used to transport DNA into thefuture, it is here that we must concentrate ourmeasurement of overlap or divergence. Suppose aparasite reproduces by spores which it inserts intothe eggs of its host, and the spores then grow up toform a new generation in the body of the host's child.In this case there will be a substantial overlapbetween the parasite and host desiderata lists. Bothsets of genes want the host to reproduce successfully;both will probably agree on what is the desirablebalance in the trade-off between longevity andreproduction; both will agree that the host shouldnot be eaten by predators, should be sexuallyattractive, should build a good nest and be a goodparent. Both sets of genes will probably agree onwhat is the best song for the host to sing, what is thebest colour for its tail, and so on.

At the other extreme, consider a parasite thatcauses its host's body to burst, shedding parasitespores into the wind in a puff of dust. In this case thetwo desiderata lists will scarcely overlap at all.

Presumably both lists will mention keeping the hostalive for a while, but the parasite list wants it aliveonly enough to build up a good bulk of spores.Thereafter the lists diverge markedly. The parasitegenes care nothing for the reproduction of the host;they have no interest in making it sexually attractive,no interest in its proficiency as a nest-builder or as aparent.

Between these two extremes will be a spectrum ofparasite/host relationships with varying overlaps offuture interests. My point is that the key variableaffecting the degree of overlap in interests is theextent to which the parasite and host share the samegametes (or spores, etc). Where parasite and hostshare gametes, the shared desiderata list will berelatively large, and the parasite will evolve tobecome benign. Where parasite reproduction is viaan entirely separate channel from host reproduction,the shared desiderata list will be relatively short, andthe parasite will be malignant.

My hypothetical parasite with spores travellinginside host eggs is not quite the most extreme alongthe spectrum towards benignness. The shared de-siderata list here is undoubtedly large, but therecould be a slight divergence of interests. Over thehost's sex ratio, for instance. Host genes desire,according to Fisherian logic, equal investment insons and daughters. Parasite genes, in this hy-pothetical example, have an interest in the host'ssuccessfully rearing daughters; but these genes haveno interest in the host's rearing sons since they donot stand to be passed on in host sperms. Thesehypothetical parasite genes are in the same positionas mammalian mitochondrial or cytoplasmic DNA(Eberhard, 1980; Cosmides & Tooby, 1981;Charnov, 1982).

The largest possible shared desiderata list will befound in those cases where the parasite DNA notonly shares gametes with host DNA but is actuallyspliced into the host chromosome.

And so we are led full circle, back to the paradoxof the organism. The paradox can now be restated interms of desiderata lists. We take it for granted thatall the genes in the genome of an organism share thesame desiderata list, but this is a contingent, not anecessary fact. Why does it so often turn out to betrue? Why don't the separate genes, or subsets ofthem, go their separate ways with separate desideratalists ? The solution to the paradox now almost statesitself, so clear has the parasite spectrum madematters. An organism's own nuclear genes share thesame desiderata list for exactly the same reason asparasite genes and host genes do when they sharegametes. The important fact about an organism'sown nuclear genes is that they all share the samegametes. It is for this reason, and this reason alone,that they stand to gain from the same set of outcomesin the future. It is for this reason alone that they all'agree' over what is the optimum state of every

R. Dawkins S72

aspect of the phenotype, all agree on the correct winglength, leg colour, clutch size, growth rate, and soon.

Those cases where nuclear genes don't share thesame gametic expectations are precisely the caseswhere they 'rebel' and don't pull together with therest - segregation distorters and their ilk. Even morerebellious would be nuclear genes that found awholly non-gametic avenue into the future. Supposethat a human nuclear gene discovered the trick ofsplicing itself directly into chromosomes of anotherindividual — a trick that is now known to be quitecommon in prokaryotic DNA. Now natural selectionmight favour such a gene if it mutated in such a wayas to circumvent the orderly, communal, gameticroute into the next generation and move sidewaysinstead. It might, for instance, cause strategicallyplaced cells to tickle the lining of the nose, inducinga sneeze. The speculation has to be completed, ofcourse, by the assumption that the rebel DNA,blown out in a sneeze, is then breathed in by anotherhuman where it splices itself into the new victim'sgenome. Whether this example is wholly speculative,or whether viruses and their kind are in fact derivedfrom rebellious host DNA does not matter toomuch. The example serves, in any case, to underlinethe central point: that the only thing that really bindsall of an organism's 'own' genes in a commonenterprise is the fact that they all share the samegametic route into the future.

Of course, sexual reproduction ensures that thegenes of an organism share the same genetic routeinto the future only in a probabilistic sense. Everysperm or egg produced by an individual in normalsexual reproduction is unique, so the genes in thediploid genotype do not share future expectations inan absolute sense. But so long as meiosis is a fairlottery - and it looks significantly as though elaboratesteps have been taken to ensure that it is - everygene in an individual's genome has the samestochastic expectations of the future. This is whythey cooperate with one another, and this is theexplanation of the definition of the organism withwhich I began: 'An individual organism is an entityall of whose genes share the same stochasticexpectations of the distant future'.

By this definition, parasites and hosts whosestochastic future expectations were identical wouldevolve closer and closer together, with more andmore blurred edges, until they would fuse andbecome the same individual. The reason why fluketissue and snail tissue are still not fused is that theirdesiderata lists are not totally overlapping. It is notthat there is something qualitatively fluky about oneset of genes and something qualitatively snaily aboutthe other set. Either of both sets of genes could havebeen put together, originally, by fusion of earliersymbiotic unions. What keeps snails and flukes apartin evolution is their divergent future interests, and

this is because they do not share reproductivepropagules. What keeps host genes together withhost genes — and parasite genes together with parasitegenes - is that they do share future interests.Parasites, then, have led us to the solution to theparadox of the organism. The genes in an organismshare desiderata lists. And this is simply becausethey submit to the same meiotic lottery and possessthe same stochastic gametic destiny.

I am grateful to Anne Keymer and Andrew Read forinviting me to the conference at which this paper wasdelivered, and to Helena Cronin for the advice andassistance at every stage.

REFERENCESBATRA, L. H. & BATRA, s. w . T. (1985). Floral mimicry

induced by mummy-berry fungus exploits hostpollinators as vectors. Science 228, 1011-12.

BAUDOIN, M. (1975). Host castration as a parasiticstrategy. Evolution 29, 335-52.

BETHEL, w. M. & HOLMES, J. c. (1973). Altered evasivebehaviour and responses to light in amphipodsharbouring acanthocephalan cystacanths. Journal ofParasitology 59, 945-56.

BETHEL, w. M. & HOLMES, j . c. (1977). Increasedvulnerability of amphipods to predation owing toaltered behaviour induced by larval acanthocephalans.Canadian Journal of Zoology 55, 110-15.

BOORSTEIN, S. M. & EWALD, P. W. (1987). Costs andbenefits of behavioural fever in Melanopus sanguinipesinfected by Nosema acridophagus. Physiological Zoology60, 586-95.

CHARNOV, E. L. (1982). The Theory of Sex Allocation.Princeton, N.J . : Princeton University Press.

CHENG, T. c. (1973). General Parasitology. New York:Academic Press.

CLARKE, B. c. (1979). T h e evolution of genetic diversity.Proceedings of the Royal Society of London, B 205,453-74.

CROLL, N. A. (1966). Ecology of Parasites. Cambridge,Mass: Harvard University Press.

CROW, j . F. (1979). Genes that violate Mendel's rules.Scientific American 240, 104-13.

COSMIDES, L. M. & TOOBY, j . (1981). Cytoplasmicinheritance and intragenomic conflict. Journal ofTheoretical Biology 89 , 83-129.

DAWKINS, R. (1982). The Extended Phenotype. Oxford:W. H. Freeman.

DAWKINS, R. (1989). The Selfish Gene, 2nd Edn. Oxford:Oxford University Press.

DAWKINS, R. (1986). The Blind Watchmaker. Harlow:Longman.

DE VRIES, P. J. (1988). T h e larval ant-organs of Thisbeirenea (Lepidoptera: Riodinidae) and their effectsupon attending ants. Zoological Journal of the LinneanSociety 94, 379-93.

DOBSON, A. P. (1988). T h e population biology of parasite-induced changes in host behavior. Quarterly Reviezvof Biology 63, 139-65.

EBERHARD, w. G. (1980). Evolutionary consequences ofintracellular organelle competition. Quarterly Reviewof Biology 55, 231-49.

The paradox of the organism S73

EVVALD, P. w. (1980). Evolutionary biology and thetreatment of signs and symptoms of infectious disease.Journal of Theoretical Biology 86, 169-76.

FISHER, F. M. (1963). Production of host endocrinesubstances by parasites. Annals of the New YorkAcademy of Science 113, 63-73.

GILES, N. (1983). Behavioural effects of the parasiteSchistocephalus solidus (Cestoda) on an intermediatehost, the three-spined stickleback, Gasterosteusaculeatus. Animal Behaviour 31, 1192-4.

HAMILTON, s. (1980). Reproduction or shell armor - atrade-off in freshwater gastropods. Bulletin ofAmerican Malacological Union, 46th Annual Meeting,American Malacological Union, Louisville, p. 71(Abstract).

HAMILTON, w. D. (1980). Sex versus non-sex versusparasite. Oikos 35, 282-90.

HAMILTON, w. D. & ZUK, M. (1982). Heritable true fitnessand bright birds: a role for parasites? Science 218,384-7.

HOLMES, j . c. & BETHEL, w. M. (1972). Modification ofintermediate host behaviour by parasites. InBehavioural Aspects of Parasite Transmission (ed.Canning, E. U. & Wright, C. A.), pp. 123-149.London: Academic Press.

JENNI, L., MOLYNEUX, D. H., LIVESEY, J. L. & GALUN, R.(1980). Feeding behaviour of tsetse flies infected withsalivarian trypanosomes. Nature, London 283, 383-5.

KEYMER, A. & READ, A. (1990). Behavioural ecology: theimpact of parasitism. In Parasitism - Co-existence orConflict (ed. Toft, C. A. & Aeschlimann, A.). Oxford:Oxford University Press (in the Press).

LOVE, M. (1980). The alien strategy. Natural History 89,30-2.

MAYR, E. (1963). Animal Species and Evolution.Cambridge, Mass: Harvard University Press.

MILINSKI, M. (1985). Risk of predation of parasitizedsticklebacks (Gasterosteus aculeatus L.) undercompetition for food. Behaviour 93, 203-26.

MINCHELLA, D. j . (1985). Host life-history variation inresponse to parasitism. Parasitology 90, 205-16.

MINCHELLA, D. J., LEATHERS, B. K., BROWN, K. M. &

MCNAIR, j . N. (1985). Host and parasitecounteradaptations: an example from a freshwatersnail. American Naturalist 126, 843-54.

MOORE, j . & GOTELLI, N. j . (1990). A phylogeneticperspective on the evolution of host behaviours. InParasitism and Host Behaviour (ed. Barnard, C. J. &Behnke, J. M.). London: Taylor & Francis (in thePress).

NOBLE, E. R. & NOBLE, G. A. (1976). Parasitology : theBiology of Animal Parasites, 4th Edn. Philadelphia:Lee & Febiger.

PHARES, c. K. (1987). Plerocercoid growth factor: ahomologue of human growth hormone. ParasitologyToday 3, 346-9.

READ, A. F. (1990). Parasites and the evolution of hostsexual behaviour. In Parasitism and Host Behaviour(ed. Barnard, C. J. & Behnke, J. M.). London: Taylor& Francis (in the Press).

REINHARD, E. G. (1956). Parasitic castration of Crustacea.Experimental Parasitology 5, 79-107.

SEGER, j . & HAMILTON, w. D. (1988). Parasites and sex. InThe Evolution of Sex (ed. Michod, R. E. & Levin,B. R.), pp. 176-93. Sunderland, Massachusetts:Sinauer.

SLOBODKIN, L. B. (1961). Growth and Regulation ofAnimal Populations. New York: Holt, Rinehart &Winston.

SMITH, D. c. (1979). From extracellular to intracellular:the establishment of a symbiosis. Proceedings of theRoyal Society of London, B 204, 115-30.

TOOBY, j . (1982). Pathogens, polymorphism, and theevolution of sex. Journal of Theoretical Biology 97,557-76.

WEIS, A. E., WALTON, R. & CREGO, c. L. (1988). Reactiveplant tissue sites and the population biology of gallmakers. Annual Review of Entomology 33, 467-86.

WICKLER, w. (1976). Evolution-oriented ethology, kinselection and altruistic parasites. Zeitschrift furTierpsychologie 42, 206-14.

WILLIAMS, G. c. (1975). Sex and Evolution. Prince-ton, N.J.: Princeton University Press.