In 1983, after reviewing hundreds of studies on birth order -- the product of 35 years of research by psychologists all over the world -- Swiss researchers Cécile Ernst and Jules Angst concluded that birth order has little or no effect on adult personality. "Whenever studies of representative adult samples were carried out with an unobjectionable method," they said, "birth order differences were nil" (1983, p. 186). In other words, firstborns and laterborns were indistinguishable in personality. However, Ernst and Angst observed that one kind of study did find significant differences between firstborns and laterborns: the kind in which the subjects' personalities were judged by their parents or their siblings.

Ernst and Angst's observations were usually correct and this was no exception. The distinction they made between those two kinds of studies has been confirmed by subsequent research. Studies using standard methods for assessing personality find no effects of birth order or occasional effects of negligible size that fail to hold up from one study to the next (e.g., Beer & Horn, 2000; Freese et al., 1999; Hauser et al., 1997; Jefferson et al., 1998). On the other hand, "all-in-the-family" studies -- in which subjects might be asked to compare themselves to their siblings or are asked questions that tap their feelings about their parents -- generally produce significant and consistent birth order effects (e.g., Paulhus et al., 1999; Salmon & Daly, 1998).

In other words, birth order effects are dependent on context. If the subject is asked questions that evoke thoughts, feelings, and memories associated with the childhood home and family, birth order effects -- personality differences between firstborns and laterborns, or between firstborns, lastborns, and middle children -- are usually found. If the testing conditions do not evoke the family context, birth order effects are seldom found. What is the explanation of this discrepancy and what are its implications for theories of birth order?

All theories of birth order rest on an unstated assumption: that the experiences children have at home during childhood have permanent effects on their personality and behavior. Some birth order theorists stress the fact that firstborns and laterborns have different relationships with their parents; others point to the sibling relationship itself. In either case, the assumption is made that children develop patterns of behavior in interacting with their parents and/or siblings that subsequently affect the way they behave outside of the family, not just in childhood but in adulthood as well.

That assumption rests, in turn, on a more basic assumption: that learned behaviors transfer or generalize from one context or situation to another, at least during childhood. Moreover, this transfer or generalization is assumed to occur unconsciously and involuntarily. It's not that children think, "This person reminds me of my mother so I'll behave with her the way I behave with my mother" -- it's that they behave that way automatically.

Examining assumptions is always a good idea and that's what I'm going to do here. I'll begin by looking at the evidence for the general proposition that young children generalize: that what they learn in one context or situation affects their behavior in other contexts or situations. Then I'll get more specific and examine the evidence that the particular strategies used by firstborns and laterborns in dealing with their parents and siblings are used in other social contexts. At the end I'll discuss what appears to be an exception to the rule but turns out not to be.

Birth order effects are presumed to have their roots in childhood. No one, as far as I know, has proposed that these effects suddenly appear in adulthood -- a time of life when people see much less of their parents and siblings. The important evidence, therefore, must come from studies of children. The evidence that I will describe here points clearly to one conclusion: that children are not born with a tendency to transfer behavior learned in one context to other contexts (Harris, 2000a). There is no built-in bias in favor of repeating previously learned behaviors simply because they have already been learned. If there is a bias, it is in the other direction: to proceed on the expectation that, because the circumstances are different, the rules for how to behave are going to be different, too.

The theory of development I proposed in The Nurture Assumption can account for the difference in results produced by the two different kinds of birth order studies. To put it more strongly, the discrepant results are a confirmation of the theory's predictions. According to the theory, children acquire certain patterns of behavior in dealing with their parents and siblings. But these patterns of behavior, and the feelings that accompany them, are linked to the context of the childhood home and family. Children learn other ways of behaving outside the home, in the presence of people who are not their close relatives.

Textbooks in introductory psychology usually define generalization something like this: "The tendency for a response learned in one situation to be elicited by other, similar situations" (Hall, 1983, p. 692). John B. Watson's classic experiment with Little Albert is often used to illustrate the concept. As Watson and his research assistant Rosalie Rayner -- later his second wife -- described the experiment in 1920, they first conditioned the 11-month-old infant to fear a white rat by making a loud noise behind his back whenever he reached toward the rat. After seven pairings of the rat and noise, Watson and Rayner reported, the child became afraid of the rat and also showed transfer (generalization) of his fear to other animals and objects -- a rabbit, a dog, a fur coat, and a wad of cotton wool.

But a later researcher, Ben Harris (no relation), looked at the various reports Watson and Rayner gave of the Little Albert experiment and decided that psychologists had been misled:

Even directly conditioned fears showed little or no transfer when Albert was confronted with the same animals in a different room:

Albert's "slight reaction" to the rabbit, for example, consisted merely of his turning his face away from the animal.

I am not going to claim that generalization doesn't exist or never occurs. My point is that generalization to different situations isn't automatic. Generalization has to be learned.

As demonstrated by the experiments of Carolyn Rovee-Collier and her colleagues, a young baby's initial inclination is not to generalize at all. Rovee-Collier (1993) begins her experiments by teaching a baby a simple trick. She ties one end of a ribbon to the baby's ankle and the other end to a mobile suspended over his head. When the baby kicks his foot, the mobile jiggles. Most babies are delighted to discover they can make the mobile jiggle; they show their enthusiasm by greatly increasing their rate of kicking the beribboned foot. Moreover, they will remember this trick several days later: if they're shown the mobile again, they will again kick the appropriate foot. But they will do this only if nothing has been changed. If the blue doodads hanging from the mobile have been replaced with red doodads, or a different liner has been draped around the crib, or the crib has been rolled from the bedroom to the den, the baby will gaze cluelessly at the mobile, as though he had never seen such a thing in his life.

But it's very easy to teach the baby to generalize: all you have to do is to train him with a "series of different mobiles or in a variety of contexts" (Rovee-Collier, 1993, p. 133). When the learning setup is varied in this way, the baby learns that the rule "kick your foot to make the mobile jiggle" holds true under a variety of conditions. Babies learn this easily but they don't come into the world already knowing it, even though it seems obvious to us that a trick that made a mobile jiggle in the bedroom should also make it jiggle in the den.

Teaching generalization isn't always so easy, however. It is very difficult to teach school-age children to generalize problem-solving skills they learn in the classroom. If they learn a method that works on one problem they almost invariably fail to use it to solve another problem, even if the two problems are very similar. According to learning theorist Douglas Detterman,

Detterman's conclusion, after reviewing 90 years of research on transfer of training, was: "People do not spontaneously transfer even for situations that seem very similar" (1993, p. 18).

Very young children do not spontaneously transfer even when the situations seem identical. Karen Adolph (1997) studied the development of infant locomotion. She discovered that when babies first learn to crawl they will attempt to crawl down an inclined plane (a carpet-covered walkway) that is much too steep for them to manage: "In their first week of crawling, most infants plunged headlong down impossibly steep hills" (p. 65), landing ignominiously on a cushioned surface at the bottom. Over a period of several months, repeated experiences with the experimental setup taught them which slopes they could negotiate and which were too steep. But then Adolph's babies learned to walk. Incredibly, there was no transfer at all from crawling to walking -- in the first weeks of walking, the babies attempted to toddle down the same steep slopes that had caused them to plunge headlong as crawlers, with the same results!

In a later experiment, Adolph (2000) discovered that infants also fail to use spatial knowledge acquired while sitting up (in the pre-crawling stage) when they begin to crawl. The knowledge acquired while sitting up is linked to the sitting-up position; babies do not use it when they are in the crawling position. Even though these results fit my predictions, I must confess that I find them amazing.

The studies of Rovee-Collier and Adolph show that babies come into the world with a built-in bias against generalizing. Generalization is something babies have to learn how to do, not something they do automatically. The way they learn to do it is by finding out that a response that worked in one situation will also work in other situations. We don't have to learn how to work the light switch in every new room we enter because we've encountered a variety of light switches in a variety of different rooms and they've all worked pretty much the same way. As a result of our past experience with light switches, we feel pretty confident that the switch on the wall next to the door will make the lights go on.

Thus, when a response has been made in several different situations and has produced the desired results in all of them, the response is likely to be repeated in new situations. That's the first condition under which generalization occurs. The second condition is when the "new" situation is so similar to the one in which the response was learned that the subject doesn't perceive it to be a new situation -- it's just the same situation again. In Rovee-Collier's experiments, the babies who were shown a different mobile -- a mobile with red doodads instead of blue ones -- stared at it cluelessly, not realizing that the foot-kicking trick would work with this mobile, too. But what if you replaced the doodads, not with red ones, but with doodads in a slightly different shade of blue? If you varied the shade of blue, you'd find that the closer it was to the original shade, the more likely it was that the baby would kick his foot. When the shade of blue is very similar to the original, he can't tell the difference. As the color becomes less and less similar to the original, he becomes less and less sure that it's the same mobile.

When two stimuli or two situations are very, very similar, a response learned in one will be replayed in the other. But do we really want to call this "generalization"? Isn't it simply a repetition of previously learned behavior?

A final complication is that similarity is in the eye of the beholder. As the author of an up-to-date psychology textbook put it, "You can't predict to what degree an individual will generalize from one stimulus to the next unless you understand something about the individual's mental concepts" (Gray, 1999, p. 138). If the two stimuli evoke different mental concepts, generalization may not occur even if the stimuli are objectively very similar. On the other hand, if the two stimuli evoke the same mental concept -- which might mean only that the individual can't tell them apart -- the individual is likely to respond in the same way to both.

Two stimuli will evoke the same mental concept when they've been found, through experience, to be equivalent. A baby soon learns that Mommy is Mommy, even if she is wearing different clothes or has changed her hairstyle. But if Mommy happens to have an identical twin who looks just like her but isn't as nice, he will quickly learn to tell them apart and will assign different mental concepts to them.

I've been talking about simple responses such as kicking a foot, but complex patterns of social behavior also fail to generalize. Even in toddlerhood, children tend to behave differently in different social contexts -- so differently that, in parent-teacher conferences, parents often find it hard to believe that the teacher is talking about their child (see cartoon). One researcher, baffled by the fact that parents' judgments of their child's temperament agreed so poorly with reports from the child's teacher at the day-care center, arrived with obvious reluctance at this conclusion: "There exists the possibility that the toddler's actual behavior differs systematically in the home and day-care settings" (Goldsmith, 1996, p. 230).

Behavior that differs in different contexts is called "context-specific." The clearest example of context-specific behavior is offered by language. In the United States, many children -- the offspring of immigrants, for the most part -- speak one language at home and a different language outside the home. I described such a child in The Nurture Assumption (pp. 64-70, 253-257); I called him Joseph. Joseph and his parents moved from Poland to rural Missouri when he was seven and a half. When he came to this country he spoke hardly a word of English but he picked up the language rapidly (Winitz et al., 1995). Within a year, Joseph was speaking English almost as well as his American classmates; within another year (like most children who immigrate before puberty) he was speaking it without an accent -- even though he continued to speak Polish at home with his parents. He could switch back and forth between his two languages as easily as I switch between two programs on my computer. Joseph's language behavior outside his home -- his unaccented English -- was not a blend or a compromise between what he learned at home and what he later learned outside the home: it consisted entirely of what he learned outside the home.

Psycholinguists call the ability to toggle between two languages "code-switching." According to my theory of development (Harris, 1995, 1998), code-switching is a characteristic of learned behavior in general, not just language. Many of the social behaviors children learn at home are inappropriate outside of the home. Children are alert to signals that their social context has changed and that different behaviors are called for; they are quick to acquire new patterns of behavior appropriate to their new setting. They can switch between behavioral programs as easily as Joseph switched between language programs, and for the same reason. An experiment done many years ago showed that children who follow rules such as "Do not lie" or "Do not cheat" in one context may ignore them in another. Children who conscientiously follow family rules at home, for example, may cheat in games on the playground or on tests in the classroom; children who are honest in school may be dishonest at home (Hartshorne & May, 1930/1971).

But sometimes a previously learned pattern of behavior does turn out to be useful in a new setting. To use the example of language again, most American children speak English both at home and outside the home. Those who learned the language at home don't have to learn it all over again in the nursery school or day-care center: they only have to be given convincing evidence that speaking English is the appropriate thing to do in their new setting. Some are convinced more easily than others. Many toddlers are reluctant, when they first begin to talk, to use their new words with people outside their immediate family.

Other behaviors, too, can be learned at home, tried out tentatively in a new setting, and retained if they work there. A child who learns to read at home, or to play a musical instrument, doesn't have to learn these things all over again at school. Both at home and at school, it is conventional to say "please" and "thank you," to eat with a spoon or fork, to refrain from grabbing things that are being used by someone else, to protest when one's own belongings are grabbed, and so on. Furthermore, individual children might find, through experience, that something they learned to do in one context is also useful elsewhere. A child who amuses her parents with her clever quips and funny stories might find that cracking jokes also makes her popular among her peers. When children generalize a learned social behavior from one context to another, they aren't repeating it blindly, like a wind-up toy. They have learned through experience that certain behaviors yield comparable results in a variety of contexts.

I like to use the example of language for two reasons: first, because the evidence is readily available and so clear that no statistical tests are needed; and second, because language behavior -- whether a child speaks English or Polish, for example -- is purely environmental, entirely learned. Genes do not predispose a child to speak one language rather than another.

But genes do influence most other aspects of a child's behavior, and these genetic influences are the source of a good deal of the confusion about generalization, because genetic influences on behavior cut across contexts. Your genes influence your behavior wherever you go. Behavioral genetic studies have shown that about half the variance in personality characteristics such as agreeableness, aggressiveness, and conscientiousness can be attributed to genetic differences among individuals (see "Why Can't Birth Order Account for the Differences Between Siblings?" on this website). Children whose genetic makeup predisposes them to have an agreeable personality will tend to be agreeable in every social context. Children whose genetic makeup predisposes them to be conscientious will be better at resisting the temptation to cheat, both at home and in school. Inherited physical attributes also play a role. Children who have the good luck to be born with a pretty face or an appealing smile are likely to experience social success in every social context and to behave accordingly.

These similarities of behavior across contexts are easy to mistake for generalization. Developmental psychologists often fall into this trap. For example, researchers who study infant attachments find correlations (weak but sometimes statistically significant) between the security of a toddler's attachment to his mother and his later success in other social arenas -- friendships with peers, for instance. Ignoring the precept that correlation doesn't imply causation, some attachment researchers (e.g., Sroufe et al., 1999) interpret the later social successes as consequences of what the child learned earlier with his mother: because he got along well with his mother, the child expects that he will also get along well with his peers.

But the attachment theorists' assumption that the child's peer relationships are affected by what he learned from his interactions with his mother is contradicted by other evidence. For example, babies whose mothers suffer from postpartum depression tend to exhibit a distinctive pattern of behavior in interactions with their depressed mothers: they behave in a subdued manner, with somber face and muted movements. Researchers have found, however, that the babies' subdued behavior is "specific to their interactions with their depressed mothers" (Pelaez-Nogueras et al., 1994, p. 358). With caregivers who are not depressed, these infants behave normally. Human babies who are smart enough to realize that, just because Mommy is depressed, it doesn't mean that everyone is depressed, are also smart enough to realize that, just because Mommy loves me, it doesn't mean that everyone will love me.

I predict that if behavioral genetic methods are used to control for genetic differences among children, the correlations reported by attachment researchers will disappear. It will be found that the weak correlations between secure attachments to mother and the success of the child's other relationships are due entirely to genetic factors. Through their effects on the child's physical and psychological characteristics, genes have an indirect influence on all the child's relationships.

Very few aspects of social behavior are unaffected by inherited characteristics -- that's why I like to look at language and accent, which are free of genetic complications. When you eliminate the contribution made by the genes, you get a clearer picture of how the environment affects children's behavior. Outside his home, Joseph spoke English without a Polish accent. The language he used with his teachers and friends was unaffected by what he had previously learned from his parents.

Much of the theorizing in developmental psychology is based on the idea that children transfer learned behavior from one social setting to another. Those who question this idea fight an uphill battle. When Greta Fein and Mary Fryer (1995a) reported that children who are taught sophisticated games of make-believe by their mothers do not use these games when playing with their peers -- in fact, they are no more sophisticated in their play with peers than children whose mothers are devoid of imagination -- other developmentalists attacked their conclusions (see Fein & Fryer, 1995b). When Michael Lamb and Alison Nash (1989) pointed out that there was "little empirical support" for the notion that children who are securely attached to their mothers have better relationships with their peers, they were ignored (e.g., by Sroufe et al., 1999). When I stood up in public and said that behaviors learned at home do not transfer to outside-the-home contexts (Harris, 1999), a chorus of developmentalists stood up and said "Yes they do!" and they pointed to evidence from intervention studies.

The intervention studies the developmentalists had in mind are those in which parents are taught more effective methods for managing their obstreperous children -- methods that employ reinforcement for good behavior and time-outs for bad behavior, instead of yelling and hitting. In many of these studies, the interventionists actually succeed in changing the parents' behavior, at least temporarily. When this happens, the children's behavior at home often improves as well. But the developmentalists claimed that teaching parents better ways of managing their obstreperous children also improves the children's behavior in school. If this were the case, it would be a clear contradiction of the predictions generated by my theory.

But it is not the case. I came to that conclusion after a long, hard look at the published research. Rex Forehand, a leading researcher in the field of parent-training interventions, came to the same conclusion. In a review of twenty years of research in this field, Wierson and Forehand (1994) reported that training can improve the parents' behavior toward the child and the child's responsiveness to the parents. "However," Wierson and Forehand admitted, "research has been unable to show that the child's behavior is modified at school" (p. 148). Likewise, a school-based intervention can improve the child's behavior in school but will have no effect on troublesome behavior at home (Barkley et al., 2000; Grossman et al., 1997). Although some researchers claim to have shown transfer of the effects of an intervention from one context to another (e.g., Cowan & Cowan, 2002; Dishion & Andrews, 1995; Forgatch & DeGarmo, 1999), I have examined these studies and found them to have serious flaws (see Harris, 2000b; Harris, 2002).

The developmentalists believed that the evidence from intervention studies was the best evidence against my theory -- the best evidence that what children learn at home transfers to outside-the-home settings. Instead, intervention studies provide strong evidence in favor of my theory. Improving children's behavior at home, by modifying their parents' child-rearing style, does not improve their behavior at school. Likewise, improving behavior at school does not improve it at home. It is possible to improve children's behavior in both places, but to do that we need home-based interventions and school-based interventions.

Much of the work in the intervention field is based on the theories of Gerald Patterson (1982). Patterson postulated that antisocial behavior outside the home has its origins in the home. Some parents, he said, use ineffective methods of discipline that have the effect of reinforcing defiant or aggressive behavior. As a result, their children behave even more defiantly or aggressively -- not just at home, but in the classroom, the playground, and the streets. The bad behavior acquired at home spills out, and pretty soon these children are a thorn in the side of society.

Developmentalists find this story so persuasive that they are undeterred when their studies fail to support it. Thomas Dishion, for example, is one of Patterson's followers. In 1994, Dishion and his colleagues reported the results of a study of children's "coercive" behavior -- the type of behavior (hostile, uncooperative, or bossy) that is supposed to result from inadvertent reinforcement by the parents. The researchers recorded instances of coercive behavior under two conditions: when children were interacting with their parents in the laboratory, and when they were interacting with their peers on the playground. Behavior in the two social contexts, the researchers found, was "only weakly correlated": the correlation was .19 (Dishion et al., 1994, p. 260). If you tried to predict how children would act with their peers on the basis of how they acted with their parents, you would be wrong most of the time. Contrary to Patterson's predictions, there were children who behaved in an obnoxious manner with their peers even though they were reasonably well-behaved with their parents.

But the correlation between behavior in the two contexts wasn't zero -- it was .19 -- and that does require an explanation. I provided one in The Nurture Assumption (p. 62): I explained that the weak correlation between behavior with parents and behavior with peers could be due entirely to genetic influences on behavior, which were not measured in Dishion's study. A genetic predisposition to be disagreeable or aggressive would presumably increase the chances that a child will be coercive both with parents and with peers. That explanation raises another question: if there are sizable genetic influences on disagreeableness and aggressiveness, and if these genetic influences cut across contexts, why is the correlation between coercive behavior with peers and with parents so low?

There are two reasons. First, since children have different experiences in different social contexts, even those who are predisposed to be disagreeable or aggressive might not behave that way everywhere they go. Consider, for example, a child who was born with an aggressive temperament but who is small for his age and puny in build. This child has learned from experience that he cannot dominate his peers, so he has given up trying. His parents, on the other hand, are intimidated by his outbursts of temper, so he continues to behave in a coercive manner with them. Now consider another child, also born with an aggressive temperament, but who is big and strong for his age. He has been successful in his attempts to dominate his peers and behaves in a bullying manner towards them. But his parents also happen to be big and strong and aggressive (heritable traits run in families), and they have little tolerance for disobedience. He doesn't dare to defy them.

The second reason for the low correlation between behavior in different social contexts is that different genes might be involved: the genetic influences that predispose children to be coercive with their peers might not be identical to the genetic influences that predispose them to be coercive with their parents. Schmitz and her colleagues (1996) studied activity level in 7-year-olds, obtaining measurements of how physically active the children were in two settings: at school (based on ratings by their teachers) and in the laboratory (where they were tested individually by trained observers). These settings differ in two important ways: the school is a familiar environment, the laboratory is not; and at school, but not in the laboratory, the children were observed in the company of their peers.

The teachers' and testers' ratings of the children's activity level were only weakly correlated -- as it happens, the correlation was .19 (just a coincidence -- don't read too much into it!). But this study, unlike the study of coercive behavior, used behavioral genetic techniques, and the researchers were able to estimate the degree to which genetic and environmental influences contributed to activity level in the two settings. They found that the correlation between the two settings was due entirely to genetic influences on the children's behavior. But genetic influences were also found that didn't contribute to the correlation -- genetic influences that were specific to the school or to the laboratory. One child might be very active on the playground but sit still in the lab, whereas another might fidget in the lab but avoid taking part in active games on the playground.

The fact that activity level in a given setting is highly heritable doesn't mean that a child will be equally active in every setting. The differences between behavior in two social contexts -- like the similarities -- can be due to genetic, as well as environmental, influences. Unless researchers use methods that permit them to estimate the magnitude of these influences, their results are uninterpretable -- to put it bluntly, worthless. The explanations they give of their results are based on assumptions that can neither be tested nor supported by their data.

At school, children are surrounded by their agemates. At home, children spend their time with people who differ from them in age, with consequent disparities in size, strength, and knowledge. Because age is a major determinant of power and influence among children, birth order is important at home.

Birth order affects the way children behave with their siblings and parents. The child with older siblings has to learn how to get along with them because they are bigger and stronger than he is. The child with younger siblings learns lessons in responsibility, leadership, and, we hope, patience. The child in the middle is landed with both jobs. These children will be treated differently by their parents: firstborns are given more responsibility, laterborns more affection (see "Why Can't Birth Order Explain the Differences Between Siblings?" on this website).

At home, in the context of the family they grew up in, firstborns and laterborns develop different patterns of behavior. But from the standpoint of evolutionary psychology, it wouldn't make sense to expect these patterns of behavior to generalize to social contexts outside the family.

One of the principles of evolutionary psychology is that humans and other animals behave differently toward kin and non-kin (Pinker, 1997). The ability to make this distinction has been reported for species ranging from wasps to salamanders to woodpeckers to gorillas (Pfennig & Sherman, 1995). An animal might be friendly toward a member of its species that is perceived to be a relative, but attack one -- perhaps even eat it -- if it is perceived to be a nonrelative. It might mate with a nonrelative but reject the overtures of a relative. Primates are good at making such distinctions. Why should we expect human children to behave in the same way toward family members and people outside the family?

To enable them to deal with a complex social life, evolution has equipped children with massive brains. As Steven Pinker (1998) put it in his Foreword to The Nurture Assumption, "Relationships with parents, with siblings, with peers, and with strangers could not be more different, and the trillion-synapse human brain is hardly short of the computational power it would take to keep each one in a separate mental account." According to my theory of development, that's exactly what children do: keep each of their relationships in a separate mental account. They don't generalize from one to another because they don't have to -- they have the brainpower to learn new rules and develop new expectations for each new relationship -- and because it wouldn't be to their advantage to do so.

Evolution designed childhood as preparation for the future. Children have to prepare themselves to make their way in the larger world, the world outside the family. In that world, family baggage is more likely to be a hindrance than a help. Laterborns who have learned to yield the right-of-way to their older siblings would be handicapped if they went through life with the expectation that they will always have to yield to others. Firstborns who have learned that they can dominate their younger siblings could find themselves in big trouble if they tried to act the same way with their peers. But the evidence is clear: in the company of their peers, laterborns are not more likely to yield, and firstborns are not more likely to try to dominate, even at nursery-school age (Abramovitch et al., 1986). Children who fight like cats and dogs with their siblings are not more likely to have stormy relationships with their friends (Stocker & Dunn, 1990). And the peer relationships of only children are perfectly normal -- their lack of experience with siblings does not disadvantage them in any way in their dealings with their peers (Falbo & Polit, 1986).

As I said at the beginning, standard studies of adult personality generally show no birth order effects at all (or effects of negligible size that fail to be replicated in other studies), whereas studies that involve judgments of or by family members generally do show birth order effects. Ernst and Angst (1983) put their faith in the studies that used standard methods and offered several hypotheses (pp. 170-171) to explain why the all-in-the-family method might produce spurious results. They pointed out, for example, that judgments made by family members, such as parents' descriptions of their children's personalities, necessarily involve comparing people who differ in age: the firstborn in the family is always going to be older than the laterborn. People change as they grow older, and they change in ways that correspond to the popular stereotypes associated with birth order: they become more responsible, more conscientious, less rebellious, less willing to take risks. If you compare an older child to a younger one, or an adult to an adolescent, on the average you will judge the older one to be more responsible, less rebellious, and so on. Though the age difference between siblings will eventually become unimportant, parents' ideas about their children, and siblings' ideas about each other, may be based (at least in part) on the days when they all lived under the same roof. (See "Why Do People Believe in Birth Order Effects?", forthcoming on this website.)

This explanation might account for some of the birth order effects obtained in all-in-the-family studies, but I think another of Ernst and Angst's explanations is more important: they proposed that the reported behavioral differences between firstborns and laterborns might be specific to the family context. This explanation is related to the first one, because the behavioral differences between siblings are due in part to the age differences between them. But it differs from the first explanation because it pertains specifically to how people act in the presence of their parents and siblings, and acknowledges that they act differently in other social contexts. It doesn't rely on people's memories of how their children or siblings behaved in childhood -- it doesn't have to, because they still behave that way when they get together. The age differences between siblings, differences in temperament and talents, and differences in the way they are treated by their parents, all contribute to the way people behave in the presence of these family members. These patterns of behavior, though context-specific, are persistent: with the right technique, they can be evoked and assessed at any time in the lifespan.

Thus, personality judgments made by parents and siblings are not wrong: it's just that they are necessarily based on how people behave in the presence of their parents and siblings and therefore do not give an accurate picture of how these people behave in other contexts.

Consider, for example, a 1999 study by Delroy Paulhus and his colleagues. The researchers asked their subjects to compare themselves with their siblings, and their siblings with each other, by indicating which member of their family was the academic achiever and which was the most rebellious. The data revealed a significant birth order effect: firstborns were most often nominated (or nominated themselves) as the family achiever; laterborns were most often named as the family rebel.

Undoubtedly, many people believe that their oldest brother or sister is the family achiever and the youngest in their family is the rebel. The question is, are these family stereotypes accurate? This is an easy question to answer because there are voluminous amounts of objective data on school grades, graduation rates, and the like. The answer is no, these stereotypes are not accurate. Studies that controlled for family size and social class have demonstrated that firstborns do not, on average, make better grades than laterborns; nor are firstborns more likely to go to college. Laterborns are not more likely to rebel in childhood by underachieving in elementary school, or to rebel in adolescence by dropping out of high school (Blake, 1989; Ernst & Angst, 1983; McCall, 1992). Ernst and Angst were right when they concluded that the all-in-the-family method yields spurious results.

This doesn't mean, however, that we can't make direct comparisons between siblings. In fact, if it's done right, making comparisons between siblings is the most accurate and sensitive way to look for birth order effects. In most birth order studies, each subject comes from a different family, so the researchers have to compare firstborns from one set of families with laterborns from a different set. Differences between firstborns and laterborns due to unmeasured differences between these two sets of families are hard to distinguish from differences due to birth order. By comparing siblings from the same family, researchers can focus on the differences within families and filter out all the differences between families.

This method was recently used to settle a long-standing question in psychology: Are there birth order effects on intelligence? Joseph Rodgers and his colleagues (2000) settled the matter by doing a definitive study: they made direct comparisons between the IQ scores of siblings in the same family. They found no birth order effects on IQ -- none at all. Contrary to popular belief, firstborns are not one whit smarter, on the average, than their younger brothers and sisters.

Both of these studies -- the IQ study and the one in which people were asked to name the family achiever and family rebel -- made use of within-family comparisons. But the methods used in these two studies differed in important ways. To illustrate the differences, I'll describe a study that used both methods at once.

The researchers, Kirby Deater-Deckard and Robert Plomin (1999), were interested in aggressive behavior, and they studied pairs of young siblings -- biological and adoptive -- from the same family. The children's aggressiveness was rated both by their parents and by their teachers; these ratings were made five times over a period of several years, when the children were between the ages of 7 and 12. On the average, parents judged the older of the two siblings to be more aggressive than the younger one. According to the teachers' ratings, however, there was no significant difference between the two.

These conflicting results are informative. Teachers' and parents' ratings differ in two ways, as the researchers pointed out. First, "Teachers are able to compare each child to a broader reference group of children," whereas most parents lack this broader reference group and tend instead to compare their children with each other. Second, teachers can observe the behavior of a given child in "an environment that provides opportunities for peer interaction," whereas "parents usually only observe children's behavior at home and with other family members" (Deater-Deckard & Plomin, 1999, p. 145).

Parents base their judgments largely on how their children behave at home -- which means, in the case of aggressiveness, largely on how they behave with each other. Because the older sibling can inflict more damage than the younger one, the younger is more motivated to avoid a physical confrontation and, if one occurs, more likely to end up getting hurt. So the older sibling is seen by the parents as being more aggressive. And, because most parents lack a broader reference group to use as a standard of comparison and tend to base their judgments on how a given child compares with his or her siblings, parents tend to exaggerate the differences between their children (Simonoff et al., 1998; Spinath & Angleitner, 1998). Thus, a child may be judged highly aggressive by her parents if, in her behavior at home, she is more aggressive than her little brother or sister.

Teachers, on the other hand, observe the child in the company of others the same age, where the fact that she is the biggest one at home gains her no advantage. Knowing how children of that age typically behave, teachers have a more accurate standard of comparison and can judge whether a given child is more or less aggressive than average. Teachers' judgments are like IQ tests in that respect: they are not affected by differences due to age. The IQ score of a child or adolescent is computed by comparing his or her performance on an IQ test with that of others of the same age. The fact that older children have more knowledge and cognitive sophistication than younger ones is corrected for by the scoring system. That's why Rodgers and his colleagues were able to compare the IQs of siblings in the same family: the measure they used automatically controlled for the age differences between the siblings.

Implicitly or explicitly, parents' judgments of their children involve making comparisons between them. In contrast, in the case of the IQ scores and the teachers' ratings of aggressiveness, it was the researchers doing the comparing. For each pair of siblings in their study, Deater-Deckard and Plomin calculated the average teachers' rating of aggressiveness for the older child and compared it to the average teachers' rating for the younger one (each child was rated five times by five different teachers). The averaged ratings for older and younger siblings were virtually identical. Though firstborns might be more aggressive at home, in the classroom and on the playground they are no more aggressive than laterborns.

In order to find out how children behave in the classroom and the playground, researchers either have to observe them in those contexts or ask their teachers or their peers. Once they have this information, they are free to make comparisons between siblings. As long as it is only the researchers doing the comparing, and as long as the measures they use aren't much affected by the age differences between siblings, making direct comparisons between siblings is an excellent way to study the effects of birth order.

This method has also been used with adults. Researchers have used standard self-report questionnaires to look for birth order effects on adult personality (Hauser et al., 1997) and social attitudes (Freese et al., 1999). Comparing the responses of siblings from the same family, they found no significant differences between firstborns and laterborns.

There is one type of birth order effect that can be detected in the world outside the family; it involves "problem" teenage behaviors such as tobacco and drug use and early initiation of sexual activity. Ernst and Angst (1983) found that tobacco and drug use were more prevalent among laterborns. Rodgers and Rowe (1988) found that younger siblings become sexually active at an earlier age than their older brothers and sisters.

Such findings can be attributed either to direct sibling influence (the older sibling provides the younger one with drugs or information, or the younger sibling imitates the older one's behavior) or to the influence of the peer group to which both siblings might belong (younger siblings often hang around with their older siblings' friends). In either case, the younger sibling is introduced at an earlier age to the temptations of teenage life. Because earlier initiation of substance use is associated with a higher risk of continued use (Grant and Dawson, 1997), the early start can explain why substance use is more prevalent, all through adolescence, among laterborns.

The hypothesis that early initiation of problem teenage behaviors results from exposure to older peers is supported by evidence from studies of kids who go through the physical changes of puberty at an earlier age. The same problem behaviors -- early substance use and early sexual activity -- are associated with early physical maturation in girls (Stattin & Magnusson, 1990; Wilson et al., 1994) and have been explained in the same way: early maturers are more likely to hang around with kids who are older than themselves.

Thus, laterborns may get into trouble in adolescence if they hang around with their older siblings and their older siblings' friends. This birth order effect appears to be limited to problem teenage behavior -- there's no evidence of lingering effects on personality. Nor is there any evidence that behaviors acquired at home are being transferred to outside-the-home settings -- the activities in question almost invariably get started outside the home. Though these activities no doubt contribute to the notion that laterborns are rebels, they are rebels only in the sense that they do the same things as their older siblings, but at an earlier age.

They don't, however, drop out of school at an earlier age (Blake, 1989). Dropping out of school isn't something that happens all at once, in adolescence -- drop-outs have generally been performing poorly in school for years. Since laterborns don't perform more poorly in elementary school than firstborns (McCall, 1992), they aren't at increased risk of dropping out. Why doesn't the influence of their older siblings' peer group cause them to perform poorly in elementary school? Because doing poorly in school isn't age-related, the way problem teenage behavior is. Teenagers are much more likely than younger kids to use drugs or have sex, but they're not much more likely to do poorly in school. The older sibling may be a better student or a worse one than the younger sibling, which means that the influence of the older sibling's peer group may be either good or bad.

Birth order affects the way we behave with our siblings and our parents, and the way we feel about them. It may also affect the age at which adolescents begin to do things their parents do not want them to do. But birth order has little or no effect on adult personality, when personality is measured by standard, well-substantiated methods in studies that include all the necessary controls. Effects of negligible size are occasionally reported in such studies, but they have generally not been replicated. Thus, birth order probably has no effect at all on adult personality.

The reason birth order does not affect adult personality is that children discover at an early age that many of the behaviors they acquired at home are irrelevant or counterproductive in other social contexts. Fortunately, they are not compelled to retain these behaviors -- they are free to acquire new patterns of social behavior, better suited to their outside-the-family contexts.

Childhood is preparation for the future. As Erik Erikson observed many years ago, the child has "no workable future within the womb of his family" (1963, p. 259). If children are destined to leave that family behind, why should they retain behaviors that were useful within the family but are not useful outside of it?

I thank Stephen L. Black, William T. Dickens, Charles S. Harris, Steven Pinker, and Robert Plomin for their helpful comments on earlier versions of this essay.