RSS
Judith Rich Harris’ The Nurture Assumption is one of those books which I’ve been leaning on for over a decade, it’s so rich in novel and “counter-intuitive” facts which nevertheless just “make sense” in terms of how the world is as opposed to how we’re taught the world should be. The most important finding isn’t really about the importance of heritable variation in human behavior, though that’s not trivial, it’s that “shared [family] environment” matters so much less than you’d think. Quite a large fraction of the variation in outcomes is “non-shared environment,” which basically means we just don’t know, and so we can’t really control. In The Nurture Assumption it is posited that a large fraction of “non-shared environment” are peer group effects. One of the facts which the author uses to support this model is that the vast majority of people pick up their accents in their native language from their peers, not their parents. The telling is exception are children who are diagnosed as autistic. But I don’t know this literature in psychology well and so I’ve been repeating this fact for years without knowing the primary source. Now I think I’ve found it, Do children with autism acquire the phonology of their peers? An examination of group identification through the window of bilingualism:
Normal children whose parents have different native languages tend to develop an accent which is closer to their peers than to either parent. It was predicted that children with autism, because of their social deficits, might not acquire the accent of their peers, perhaps because of the lack of the normal drive to identify with peers. Bilingualism was used as a window into such social factors in language acquisition. Using audiotaped speech samples, the study found that in a sample of children with autism who were brought up in England and whose mothers were not English, 83.3% acquired their mother’s (non-English) accent. In contrast, among normally- developing siblings of children with autism who were brought up in England and whose mothers were not English, only 12.5% acquired their mother’s (non-English) accent. We suggest that such studies of unusual populations are of value in furthering our understanding of the larger population of children with autism, and the influences on normal social development.
The sample size was less than 100, and I’m very curious if this research has been followed up (using Google Scholar I didn’t stumble on anything very clear, but again, I’m not familiar with this discipline). My point in focusing on accent in the previous post is how we speak is often a social cue, rather than about language acquisition as such. Speaking of which, though it is common sense it is important to reiterate that children are basically savants when it comes to learning languages. Multi-lingualism is common in many parts of the developing world, and presumably it may have been in a more ethnically fragmented ancient world too. It’s a core human competency, not a signal of cognitive exceptionality.
English
Arabic
Chinese (Simplified)
Dutch
French
German
Hebrew
Hindi
Italian
Japanese
Korean
Persian
Polish
Portuguese
Russian
Spanish
Swedish
Turkish
Vietnamese
Children are not language savants. They are only good at learning sounds.
I am not aware of any evidence of any regression with age of ability to learn any other aspect of a second language. Adults learn faster per hour of studying, they just have better things do with their time, like hang out with expats. Here is a survey (ungated).
From studying abroad, I noticed that every one of my peers who had six hours of daily instruction and tried to socialize in the language learned remarkably fast. A few months was more than enough language development to have friendships, romantic relationships, etc.
That survey seems unclear on the point. On page 579 they note that:I can't think of too many disciplines where 8-year-old's performance surpasses that of adults within one year.
I am not aware of any evidence of any regression with age of ability to learn any other aspect of a second language. Adults learn faster per hour of studying, they just have better things do with their time, like hang out with expats. Here is a survey (ungated).Replies: @Hipster, @ziel
It would be best to compare college aged study abroad students after one academic year to kidergarten immigranta after one year.
From studying abroad, I noticed that every one of my peers who had six hours of daily instruction and tried to socialize in the language learned remarkably fast. A few months was more than enough language development to have friendships, romantic relationships, etc.
I am not aware of any evidence of any regression with age of ability to learn any other aspect of a second language. Adults learn faster per hour of studying, they just have better things do with their time, like hang out with expats. Here is a survey (ungated).Replies: @Hipster, @ziel
Children are not language savants.
That survey seems unclear on the point. On page 579 they note that:
I can’t think of too many disciplines where 8-year-old’s performance surpasses that of adults within one year.
If your accent says something about the type of people you hang out with, than a lot of people will think I hang out with guys who are Cosa Nostra connected.
I think an important lesson from this is, be very careful how you interpret these decompositions into genetic/shared/unshared environment. Lots of people go around saying stuff like, “The shared-environment effect for trait X is very small, so what you do doesn’t matter much to how your children turn out relative to trait X, unless you’re a child abuser or something.” That’s a fallacy, and a very common one. I think even Harris used to say that, I’m not sure; it’s been years since I’ve read anything by her. (I haven’t read the book, but I read the original paper on which it was based and various popular articles by and about her.)
This autism study illustrates the fallacy of that kind of statement. These statistical models all assume that we’re a homogeneous population, and that personal differences are just i.i.d. noise. But of course that’s not the case. Autism is an extreme example, but we have no idea what other heterogeneity might be relevant. Maybe something about how you relate to your children really does largely determine trait X; if very few parents do it, then it won’t show up in the shared-environment effect, which after all is just computed from averages. (For instance, a correlation is just an average of “co-deviations.”)
The statistical concept of genetic/shared/unshared is useful, but because it’s based on such a simplistic and – in the case of autism, at least – unrealistic model, it’s wrong to push it way beyond the facts as Harris (I think) and others do.
Of course, I'd say by Occam's Razor, they likely don't exist.Replies: @Aaron Gross
Harris does say that. But I’m not sure how necessary that caveat is, actually (it must be on some level, such as when severe injury or perhaps PTSD may be involved).
This might technically be true, but you’re thinking about it the wrong way. The 0 shared environment does, at the very least, set an upper bound on the extent that differences between different sets of parents could matter (especially now that we have much larger and wider studies that we did from back in Harris’s day). Even if there was some subset of families or parenting practices that had an effect, the 0 shared environment means that for the vast majority of us, it doesn’t apply to what we do. In other words, such people/cases – if they existed at all – must be rare (for the record, the shared environment effect is also 0 for mental disorders, so we did look).
Of course, I’d say by Occam’s Razor, they likely don’t exist.
Something that I have some doubts is if is correct to equate “shared environment” with “family environment” (look to all the stories – begining in the Bible – where “parents treating different sons in different way” is a important point of the plot).
Of course, I'd say by Occam's Razor, they likely don't exist.Replies: @Aaron Gross
Hi, JayMan, nice to see you again. Actually, I think Harris did say something like, “and possibly even the caveat about abuse is unnecessary.” But even with that caveat, she was always interpreting the behavioral genetics findings in a way that went far beyond the data, in fact far beyond what that kind of study even could show.
Regarding your upper bound: Well, the shared environment effect is generally not zero, of course. If it were, the upper bound on differences would be zero, too. But even if the shared environment effect is “negligible,” say the oft-quoted 5-10%, then the upper bound would be so high as to be irrelevant; the smaller the heterogeneous subpopulation, the higher the upper bound.
If, say, only a small percentage of parents play Mozart to their babies, and playing Mozart to babies strongly influences their adult IQ, completely overwhelming genetic and unshared environmental effects – this is a hypothetical example! – then that would be totally consistent with an overall shared environmental effect of only 5-10% on IQ. Of course I didn’t put numbers on that, because I’m too lazy, but I think this hypothetical example holds for the real-life fraction of Mozart-playing parents having a very strong shared environmental effect. You’re welcome to do the math if you like and correct me if I’m wrong; or, you’re welcome to agree with my intuition about that.
So it’s wrong to say, as people like Harris do, “Shared environment has a negligible effect, therefore playing Mozart to your babies probably won’t make much difference.” Of course even if this hypothetical example doesn’t hold for a group as large as Mozart-playing parents, there might be some parenting styles (not just pathological abuse) that do have an extremely strong effect, but that are more rare. What you can conclude from the findings that Harris cites is that taking the average over all parenting styles practiced today, there’s not much effect on average relative to genes and unshared environment. And similarly for all other aspects of shared environment.
You’re right of course that the parsimonious hypothesis (“Occam’s Razor”) is that the population is completely homogeneous. For instance, the parsimonious hypothesis is that autistic children pick up accents exactly the same way that normal children do. Parsimony is one way to decide between competing hypotheses that are otherwise satisfactory, and it might be a good way to choose a null hypothesis, but it’s not an excuse, in the absence of data, to impose a Procrustean model on reality.
If c^2 were consistently > 0, albeit small, that would actually be something interesting, and would open the door to all kinds of potential environmental effects (gene-environment interactions, for example). However, that's not what we see.
If nonshared environment may be through peer effects, and the lack of accent-adoption among autistic kids implies less peer effects, would this imply that nonshared environment estimates will be systematically lower for / shared-environment+heritability higher for autistic kids compared to non-autistic kids? If they aren’t, that would seem to be a blow to the peer effect theory.
Some quick googling didn’t turn up anything obvious, but perhaps someone here has seen or done such a comparison.
“Shared environment” isn’t just family environment; it is all environmental forces that children growing up together share. So it’s actually broader.
A commonly made point, but one that makes no sense when you think about it: sure, parents don’t treat each child exactly alike – children growing up together have far from identical environments. But they don’t need to be identical; they just need to be more similar than the environments of children who didn’t grow up together, which they are. None of the environmental variation between families makes children more similar.
I was curious about the autism finding so I did an informal test of my kids, one is diagnosed with high functioning autism, the other is not. I am from California but I’ve lived in Wisconsin for many years and my kids were born in Wisconsin. Here drinking fountains are called bubblers and in the many years of living here, I just cannot bring myself to use the term bubblers. I brought up images of drinking fountains and asked my non autistic kid to name them and she said ‘bubblers.’ Later, I brought up the same images and asked my son to name them and he said ‘drinking fountains.’ No one here calls them that and I assume he learned that term from me and feels no pressure to use the Wisconsin term ‘bubbler.’ Yes, a totally informal non scientific test but it was interesting that it seemed to work in our case. As an aside, I’ve noticed my son gets very tweaked when he listens to someone speak with a strong accent, seems to find it very distracting.
When my son was young it bugged me that all the other kids seemed to figure things out so easily while it was like pulling teeth for my son and I started to call the other kids ‘osmosis kids’. They just seemed to learn everything without any effort while my son needed explicit instruction for most everything. When my daughter was born I discovered what it was like to have an osmosis kid, myself, and was amazed how easily she learned, often without any instruction. Looks like that passive learning also applies to accent and language conventions which shouldn’t be surprising. Anyway, the good news is that unconventional learning can pay off: my son is doing very well in school as compared to many of those osmosis kids, he’s getting excellent grades while a lot of those osmosis kids are barely passing, probably feeling pressured by their peers to be cool, not smart.
“unconventional learning can pay off: my son is doing very well in school as compared to many of those osmosis kids”: if most children are osmosis children, then the school is teaching them all wrong.
Perhaps “osmosis learning” (i suppose that, in practice, this is learning by involuntary imitation, like learn to talk) is impossible to most school subjects: I suspect that only works to learn who to do something (you see other people doing something and your mirror neurons, or whatever, makes you imitate them); but how you, by involuntary imitation, will learn the History of the USA (or of Portugal)? Even mathematics (probably the only academic subject) that is about learning to do something) is probably to much complex to be learned by imitation only?
I wonder whether people in the future will still have regional accents. That is to say, as t.v., movies, and internet continue being easily accessible, will they have a homogenizing influence on the English language? Could media be like a “peer”?
I do think adults exagerrate how quickly small children learn languages. I live in an expat environment and constantly hear from parents how amazingly quickly their child is picking up German. “Within 6 months he was completely fluent!”, that sort of thing. Usually when I meet said child he or she still has an accent, makes grammatical mistakes, has gaps in vocabulary, etc. I suspect most 30 year olds could probably also learn how to function competently as a 6 year old in a foreign language within 6 months if they spent their time going to first grade in that language.
My observation is that the children of immigrants are far more likely to speak with bland TV accents than with the colorful local ones of their peers and peers’ parents.
Senator Brundlefly, I don’t think it will work that way. In my hometown, half the people use grammatically incorrect sentences. “They was…” etc. They know it is incorrect, but they want to sound like their friends.
No, the shared environment is actually exactly zero, as most evident from the largest studies. You see a >0 shared environment in studies of children and in (some) studies with small samples. There is no evidence from studies of heterogeneous populations that they exhibit a higher c^2 (shared environment).
If c^2 were consistently > 0, albeit small, that would actually be something interesting, and would open the door to all kinds of potential environmental effects (gene-environment interactions, for example). However, that’s not what we see.
I finally did read this book, second edition. In it the author concedes that shared family environment can be decisive if *the family is also the group.* This seems to mean either a family large enough to satisfy all social needs of all members, or a family which forbids contact with outsiders, or both. It also seems to require a dominant father or parent with unquestioned authority who basically forbids the kids to take on local values. An example might be Michael Jackson’s father, who raised his kid as parts of a family music business and who also practiced a religion strange and unappealing to almost all Americans (Jehovah’s Witness).
In this case the rule about shared family life is an empirical rule which applies to Americans along with the people of an unknown number of other nations, but not necessarily all. In America, between the media and the public schools, parents really cannot control their kids environment. (This is the big motive of homeschoolers). Furthermore, the parent who is able to do this must be exceptional both in motivation and capacity; I’ve known many parents who tried but just couldn’t win the battle.
In this sense, the message is “The weeny little things you do to mold your child, besides possibly being ineffectual and faddish as such (e.g, playing Mozart to and infant), will probably be overwhelmed by the local social environment” (especially because most parents are similar to a large class of other parents). And in this regard she does concede that choosing a certain place of residence is one thing that you CAN do on your own for your kids. (And lo! that is an obsessive concern of American parents, with school quality and zoning laws major factors in deciding housing prices, which in turn recently turned out to be a controlling influence on the American and world economy.)
From an anthropological or sociological point of view, as opposed to a psychological one,this general idea is non controversial. It’s only psychologists,educators, and parents, to my knowledge, who think that children can be molded by deliberate adult activity.