« More on outsourcing jobs.... | Gene Expression Front Page | New London »
August 07, 2003


It is a familiar idea in evolutionary biology that sexual selection may produce some trait despite its being disadvantageous to general fitness. The peacock’s tail is a classic example.

The Handicap Principle goes a step further by claiming that some traits are selected precisely because they are disadvantageous. The HP was first put forward by the Israeli zoologist Amotz Zahavi in the 1970s. For a long time it was resisted by other biologists, but since about 1990 it has been more widely accepted.

I recently read Amotz and Avishag Zahavi’s book The Handicap Principle: a Missing Piece of Darwin’s Puzzle (1997). This is the most interesting book on evolutionary biology I have read for a long time. The Zahavis apply the HP not only to sexual selection but to a wide range of other animal and human traits. Some of their ideas are far-fetched, but always stimulating.

The Zahavis argue that the HP comes into play whenever an animal trait conveys information to other animals (of the same or different species), where it is in the interests of both sender and receiver that the information should be honest, but where it would be advantageous to other animals to ‘cheat’ by sending false information. The only way to deter cheating is for the ‘signal’ to be so expensive to produce that it is not worth while to cheat. Since animal signals do exist, and in many cases it would be advantageous to cheat if this could be done without cost, it seems to follow that signals must have evolved to be expensive to produce, and this is just another way of stating the Handicap Principle.

It is one thing to show that the HP must apply, but another to show how ‘handicaps’ evolve. The reason why biologists resisted the idea was that the Zahavis offered no mathematical models to show how a gene promoting a handicap could be selected, and other biologists who tried to devise such models found that they did not work. The problem is that such a gene starts by reducing fitness, and it is difficult to see how it can get a sufficient advantage from ‘honesty’ to offset this until it has already become common in the population. At this point it becomes advantageous for the ‘receivers’ (e.g. females, in the case of mating behaviour) to evolve a strong preference for it. This would explain how a handicap can be maintained by selection, but not how it evolves in the first place.

This seems to be still a weak point in the Zahavis’ arguments, and they are very vague about the process by which handicaps evolve, and the conditions under which this can occur. Mathematical biologists have however done a lot to clarify this, notably Alan Grafen in two papers in 1990. As Grafen explains it, the key condition for a handicap to evolve is that the marginal cost of producing a signal (e.g. the peacock growing a longer tail) must be greater for a low-quality animal than for a high-quality one, so that the low-quality animal stands to lose more by trying to ‘cheat’ than it gains. It is plausible that this condition often applies. For analogy, suppose that men can obtain sexual partners by lifting heavier weights than their rivals. It may be that with sufficient training and practice any man can lift the heaviest weight available, but it will still be easier for the biggest and strongest men to reach the position of being able to lift any given weight. The contest will therefore be a reliable test of strength.

The Handicap Principle can also be applied to cultural and economic phenomena such as conspicuous expenditure, initiation rites and tests, extreme sports and contests, artistic skill, and fashions in dress. Here we do not suppose that each cultural trait has been produced by biological selection. It sufficient that humans should have a general-purpose ability to detect cheating. If the only honest signal is a costly signal, then culture will spontaneously select signals which impose costs on the signallers. This principle seems capable of explaining a great deal in human behaviour that otherwise seems perverse and wasteful. Of course, in one way this is a rather pessimistic conclusion. But it does also offer the hope that by understanding the roots of behaviour we may consciously change it.


Posted by David B at 04:13 AM

Is there a reason to assume that the trait comes first and that female preference for it comes later? Seems like it might be easier for something like this to evolve if you assume that males have a random distribution of some feature and that the first 'event' is a mutation on the female side that results in females preferring males at one extreme end of the spectrum (e.g. those with the showiest tails in the case of peacocks - though initially the tails would scarcely be 'showy' at all...)

Posted by: bbartlog at August 7, 2003 06:34 AM

This is why women like men who waste their money.

Posted by: Gordon Gekko at August 7, 2003 09:46 AM

I've always assumed that peacocks are telling peahens, "Yo, ladies, check out my tail! Obviously, I have excellent all-around genes to drag a ridiculous tail like this around and not have been eaten by wolves already. If you mate with me, your daughters will be ultra-healthy. Some of your sons will get eaten by wolves because they can't get away fast enough, but the survivors will be score big with the next generation of ladies."

Posted by: Steve Sailer at August 7, 2003 02:08 PM

"The Zahavis argue that the HP comes into play whenever an animal trait conveys information to other animals (of the same or different species), where it is in the interests of both sender and receiver that the information should be honest, but where it would be advantageous to other animals to ‘cheat’ by sending false information. The only way to deter cheating is for the ‘signal’ to be so expensive to produce that it is not worth while to cheat. Since animal signals do exist, and in many cases it would be advantageous to cheat if this could be done without cost, it seems to follow that signals must have evolved to be expensive to produce, and this is just another way of stating the Handicap Principle."

I did not understand this explanation. How can information be honest? How does a third animal enter the picture?

Posted by: Inventor at August 7, 2003 02:37 PM

I fail do see how a preference for handicaps enters into sexual selection--or rather, how it differs at all. It is true that sexual selection does not always fall in line with what's needed to survive, but women don't look at men in wheelchairs and think, "I want to have his children." If you're merely talking survival handicaps, then how is this "handicap selection" any different than sexual selection?

Posted by: PhlegmAsiv at August 7, 2003 03:08 PM

I should have ended by saying 'read the book', because the Zahavis are more persuasive than I am!

I sympathise with bbartlog's point - this is the easiest way for selection to get started. I'm not sure if the Z's would agree - like I said, they are vague about the start of the process.

Phlegmasiv - (a) the concept of HP overlaps with that of sexual selection. Think intersecting circles in a Venn diagram. (b) the handicap has to be of such a kind that it demonstrates its owner's fitness. Wheelchairs are therefore not a good example.

Inventor - the point about honesty is that if a signal is cost-free, anyone can produce it, and if it is advantageous to do so, they will. Anyone can say 'I'm rich', but the only way to prove it is to splash your money about.

Sorry I'm not explaining this very well - just got back from a journey and feeling exhausted. I can understand people being sceptical about the HP, because it is counterintuitive, but even the biologists who started out critical about it, like Dawkins and Maynard Smith, now accept that it can work (see Selfish Gene 2nd edition).

Posted by: David B at August 8, 2003 12:27 PM

From the synopsis on this site, it seems that the authors present HP as a refinement of sexual selection, and I am not buying it. There are plenty of instances of signals and of cheaters where cheating does not reduce the value of the signal. For instance, plenty of animals mimic threatening creatures. The poisonous coral snake has several non-venomous mimics. Likewise, plenty of species select for behaviours that cheat on other signals like chest size for instance where a bird might ruffle its feathers during courtship to make it look like it has a larger chest.

When I think of a handicap principle and the peacock's tail, I think of the saying that a gazelle need not outrun the lions as long as it outruns its sisters.

Males are reproductively very redundant and expendable compared to females. A peahen has a natural handicap when it comes to predation because she has to watch out for junior. Having a hobbled male around to occupy the predators while she and her chick sneak away improves the odds of her survival and the survival of her chick. Ironically, it can also improve the odds that the male's genes survive when the chick grows up to reproduce.

Let's suppose that well-nourished, disease-free, parasite-free birds grow larger and more symmetrical so they have larger, more symmetrical chests and longer, more symmetrical tails. Consider two hens in different colonies that each have perverse sexual fetishes: one likes a man with a big brawny chest and the other likes a good piece of tail. As a result, they both end up picking better-nourished, better-groomed, healtier males as partners and eventually their female descendants come to dominate each colony.

Eventually the males start cheating. Because the fetishes respond to secondary characteristics, the males adapt by putting their energy into those secondary characteristics and not necessarily into the features that improve survival. In one colony, small weak males end up with long symmetric tails and in the other colony, small weak males learn to puff out their chests and ruffle their chest feathers.

Now we come to the good part. Suppose there is a year or two of extreme predation as happens at the crests of predator population cycles. To make the numbers easy, suppose each colony has 100 males, 100 females and 100 chicks. Further assume that if the female survives, so does her chick.

The male chest rufflers do not really have a handicap when it comes to escaping predators whereas the long tailed males have a very real handicap. Suppose that during extreme predation 100 adults are killed.

In the long-tailed colony, the predators are drawn to the appetisingly large appearing males and find them easy to catch. In this colony, the predators get 98 males and two females. In the chest ruffler colony, the predators get 50 males and 50 females.

At the start of the next season, the long-tailed colony has 98+49=147 females and 2+49=51 males assuming a 50-50 split between the sexes in the chicks. The chest ruffler colony has 50+25=75 females and 50+25=75 males at the start of the next season. Even though almost the entire population of adult long-tailed males is wiped out, they managed to get an addional 48 of their young to survive and the population of adults only decreases by 2 compared to a decline of 50 chest rufflers.

Consider what happens with a second year of extreme predation. In the long-tailed colony, the predators kill all 51 adult males and 49 females because they run out of slow males to kill. In the chest ruffler colony, they kill 50 males and 50 females again.

At the start of the next season, the long-tailed colony has 98+49=147 females and 0+49=49 males. In contrast, the chest rufflers have 25+13=38 females and 25+12=37 males.

Even if the predation begins to decline a little, the chest ruffler colony is wiped out before the end of the third season whereas the long-tailed population is hardly touched.

Because males are so redundant and expendable, handicapping the males actually improves the survival of the genes and this has nothing to do with the accuracy or the cost of signals. It has only to do with the higher female investment in reproduction.

Similarly, male cardinals are bright red and have behavioural 'flaws' that cause them to seek the attention of predators.

Posted by: Bob at August 8, 2003 04:26 PM

I should also add that if there were a third colony where females evolved to respond to the primary characteristics of nourishment, grooming and health, the predators would likewise wipe them out in three seasons.

Posted by: Bob at August 8, 2003 04:34 PM

Bob - I'll have to think about your chest/tail example quite hard! Meanwhile, if you haven't already read them, see the notes at the end of The Selfish Gene, 2nd edition, which has a good discussion of the HP.

As to the coral snake example, surely this is a case of Batesian mimicry? I think the general view is that Batesian mimicry only works so long as the (harmless) mimics are comparatively rare. If they become too common, the warning signal of the coral snake's marking is devalued, and it would have to evolve something else.

Posted by: David B at August 9, 2003 12:43 PM


I think the model you are proposing offers an interesting opportunity to make a point about evolution. Instead of a single fitness measure, consider two fitness measures: a measure of what is good for the organism and a measure of what is good for the genes.

A handicap would represent a situation where the measure of fitness for the organism is sub-optimal or even negative while the measure of fitness for the genes is maximized. In the example I gave above, the long tails impede the ability of male peacocks to survive predation, but by increasing the survival rate of the males' chicks, the trait optimizes survival of the males' genes.

In the end, evolution is about gene survival and gene fitness. We humans are no more than our chromosomes' strategy for reproduction. As a reproductive strategy we have been successful in the short term, but we do not yet have the provenance and track record of cock roaches. Who knows whether we will even prove to be a good strategy in the long term?

Evolution has no purpose and no compassion. It does not care what is good for you or for me or for the male peacock or for any individual. It's just a dumb automatic process for keeping score, and the only score that counts is whether the genes propagate to the next generation.

Posted by: Bob Badour at August 9, 2003 07:12 PM


Okay, too much mimicry leads to diminishing returns for both the mimic and the threatening creature. The thesis you presented in the synopsis is the importance of truthfulness and the high cost of the signal necessary to guarantee honesty.

In the example of the peacocks, consider the three hypothetical colonies: long-tails, chest-rufflers, and adoneans.

In the long-tail colony females respond sexually to a secondary characteristic and males cheat.

In the chest-ruffler colony females respond sexually to a secondary characteristic and males cheat.

In the adonean colony females respond sexually to primary characteristics and the males are all prime specimens of strength, beauty and health.

In the example of extreme predation I gave earlier, the only honest group, the adoneans, get wiped out. In fact, if the strength and speed of the adonean peacocks helps them evade predation compared to the peahens, the colony might get wiped out sooner than the chest-rufflers. Once all the females are gone, that's it. There is no more colony.

It is actually the male handicap that improves survival for the genes and for the females, and the handicap does not need to communicate anything. There are plenty of species where the male is killed and consumed during the act of reproduction to provide a source of energy and protein. (Talk about one man, one vote, one time... ;-) We males are just massively redundant. We are full of physical and behavioural handicaps that kill us off young.

A hundred females can make do with a single male, and it doesn't matter whether there are 100 males or 1000 males if there is only one female.

Posted by: Bob Badour at August 9, 2003 07:50 PM

GC: Obviously a handicap must increase 'fitness' in the technical sense if it is to evolve by NS. I used the phrase 'disadvantageous to general fitness' to try to distinguish this from 'fitness' in the technical sense, but it would have been better if I used some other word, some as 'quality'. Again, see the discussion in Dawkins (2nd edition).

Bob: the whole question of 'honest signals' and mimicry is very interesting, and I may come back to it in another note.

On your model of 'evolution of a handicap', I think you are suggesting that a male handicap (such as the peacock's long tail) may be favoured by selection because males (unintentionally!) decoy predators away from the females.

On a purely factual point, don't peacocks and other male birds with elaborate feathers usually molt them as soon as they have finished mating? If so, this rather weakens your theory, as the males will not act as 'decoys' while the females are brooding the eggs and raising the chicks, which is when the 'decoy' would be most useful.

As to the details of the model, you have simplified things by assuming that different types of bird form separate 'colonies', so that there is no competition between the variants within a colony. I don't think this is reasonable. If a migrant short-tailed male entered a 'long-tailed' colony, or if a short-tailed mutant emerged within the colony, it would have a huge selective advantage from being less vulnerable to predation. To maintain the prevalence of the long-tailed form despite this selective disadvantage, you have to appeal to some other principle of sexual selection, e.g. Fisher's runaway process, or the HP itself.

I recognise that any model of evolutionary processes has to make simplifying assumptions. I do it all the time myself! The danger is that you may 'simplify away' the points that really need explaining.

Posted by: David B at August 10, 2003 04:41 AM

By 'this selective disadvantage' of course I mean the disadvantage of the long-tailed form relative the the short-tailed migrant or mutant.

I should also add that I don't accept the assumption that a 'low quality' male can grow a long tail as easily as a 'high quality' male. To return to my weightlifting analogy, maybe a 100-pound weakling can grow big muscles if he trains hard enough, but someone who starts out with a muscular physique will grow even bigger muscles with the same amount of training.

Posted by: David B at August 10, 2003 04:51 AM

I separated the colonies for clarity, and there is nothing to say that the peacock did not evolve its long tail in a geographically isolated colony.

You have to look at evolution as a battle of the sexes. The females evolve long-tailed males to increase their own survival. Males would evolve to molt the long tails in response to increase their own survival. The tug of war back and forth will eventually evolve long-tailed males who molt neither too soon nor too late after the rut. Molting too soon will reduce the survival of the chicks and molting too late will reduce the male's chance of reproducing multiple seasons.

A short-tailed male would not have a reproductive advantage among females with the long-tail fetish. The short-tailed male would have to come along at the same time a female comes along with a perverse fetish for short tails, and even then his descendants might get wiped out during another period of extreme predation. Of course, for all we know this is happening right now and in a million years peacocks will all have short tails.

The weightlifting example is misleading. There are a number of phenotypes that affect a weightlifter's ability: hormone production, height (bone length) and insertion point. Muscle size is determined more by muscle belly length.

It is possible for a tall weightlifter with long muscle bellies and unfortunate insertion points to have huge muscles and not be all that strong, and for a short weightlifter with short muscle bellies and fortunate insertion points to have small muscles and massive strength.

Consider two genetically identical birds with the exception of one genotype difference that causes one of the birds to put more energy into growing a tail. If both birds have identical developmental nutrition, the bird who puts more energy into tail production will be smaller with a longer tail. It is possible that even if the long-tail received less nutrition that it would still have a longer tail, but be even smaller still.

If females are choosing mates based on a tail fetish, they will pick the longest tailed male even if it is smaller.

By the same analogy, if human females select mates on the basis of muscle size alone, human males will cheat by evolving longer muscle bellies even if that does not improve their strength.

Posted by: Bob at August 10, 2003 10:07 AM

I was wrong about the peacock - they don't molt their tails immediately. On the other hand, some birds with extreme plumage - like the African widow bird with the very long tail feathers - do molt them as soon as the mating season is over.

I can't agree that the geographical separation of the long tailed and short tailed variants is just a minor detail of the model. You are simply assuming that populations with distinct traits and preferences have somehow become established. But a large part of the problem is to explain how this happens in the first place, given that there are males with different traits and females with different preferences, all in the same population.

I agree that if the entire female population of an area has a strong preference for long tails, then a short tailed immigrant or mutant may be selected against, even if it has a major advantage w.r.t. predation. But we should also expect female migrants or mutants with a preference for short tails (or no preference at all) to emerge, and these would give an advantage to males with shorter tails. I still argue that a disadvantageous trait cannot be maintained without something akin to Fisherian selection.

But I must admit that the idea of male traits as a 'decoy' to predators is ingenious. I just don't think it would stand up in court!

Posted by: David B at August 10, 2003 02:36 PM

re: Coral snake mimics, and other mimicry where the signal is "poison!":

These signals are not zero cost. The bright colors make the mimics an easy target for any predator that can tell the difference between the mimic and the real thing, and also for predators recently arrived from a different biome that don't know the signal. And if the mimic becomes too common compared to the actually poisonous creature, the predators may learn to simply take a chance. (It's not likely with something as deadly as coral snakes, but most objects of such mimicry will only make the predator ill, like monarch butterflies and brightly-striped hornets.)

Posted by: markm at August 10, 2003 03:59 PM

It occurs to me that a 'decoy' effect of the kind proposed by Bob could work, even in a mixed population, if the benefit of the self-sacrificial 'decoy' male is received only by the females who have mated with him. The benefit could then be regarded as a form of paternal investment. But this doesn't strike me as a very realistic scenario. It is much more likely that the long-tailed male, by diverting predators away from females, would confer a benefit equally on all females in the district, whether he had mated with them or not, and whether or not they had a preference for long-tailed males. So it would not tend to increase the frequency of such a preference.

markm: good point about coral snakes, etc. There is a similar argument in the Zahavis' book.

Posted by: David B at August 11, 2003 02:44 AM

Sorry to keep changing my tune about peacocks. Originally I thought the males shed their tails after mating (vague recollection of a TV nature prog), but then I checked and found a statement that peacocks only molt once a year, in the Spring.

However, I have just Googled on peacocks, and found various peacock-fancier's sites which imply that the males do lose their tails after mating. E.g., from 'The Wild Peacock Page': "Peacocks molt and lose their beautiful plumage usually in August. We have to practically run after the birds as they drop the feathers, before they get ruined by the rain. A male can drop all his long feathers in one day".

This does sound like a post-mating adaptation, but if anyone knows different, please correct me.

Posted by: David B at August 11, 2003 11:27 AM