Tuesday, March 18, 2008

The Problem with Attachment Theory

At its base, traditional attachment theory makes the following prediction – take identical children and separate them at birth. One is brought up in a loving family where it forms stable attachments to its adopted or real parents. The other is less fortunate. It is moved through a children’s home, to foster parents, back into care and has no chance to enjoy a proper upbringing. Child one goes to a good school, child two goes to lots of bad schools. According to attachment theory we should see marked differences between the children once they grow into adults. But we don’t.

Even if no one has been cruel enough to do the experiment mentioned above, there have been hundreds of studies of identical twins separated at birth. The researchers were trying to establish how much their personalities and behaviour were affected by nature and how much by nurture. The answer was that identical twins showed a correlation with each other of about 50%. But this figure was almost the same whether they were brought up together by their natural parents, brought up together by adopted parents or separated and brought up apart. Nurture seemed to play no part. Likewise, adopted children did not correlate to their adopted parents anymore than a stranger off the street would. There were some nurture effects while children were still growing up, but once they were adults, these disappeared.

The first person to try to make sense of these results (which were the opposite of what most researchers had expected), was Judith Rich Harris in her book The Nurture Assumption. She was not a psychologist, still less a geneticist and some felt her outsider-status gave them a license to ignore or insult her. Contrary to popular belief, Harris did believe in nurture, but from peers rather than parents. Since attachment theory is generic enough to handle peer-to-peer relationships, some attachment theorists responded to Harris’s criticisms of family-based nurturing effects by shifting their attention from the parlour to the school yard.

The trouble is that there is hardly any evidence for Harris’s peer pressure hypothesis and quite a lot against it. For a start, it is, at first sight, unlikely that parents could have practically no effect and other children such a lot. Work comparing children sent to nursery at six months and those kept at home has found some evidence that the former are more confident and aggressive, but this wears off before they grow up. The Chicago work on school places I referred to a few posts ago suggests it doesn’t matter which school you went to, although this was based on academic results rather than personality. More work is required and we are still hamstrung by having no reliable way to measure intelligence, but things are looking pretty grim for the peer-to-peer hypothesis. The family hypothesis is already dead, if not buried. If peer pressure goes the same way, as seems very likely, there will be nothing left for attachment theory to attach itself to as far as long term outcomes are concerned.

Of course, attachment theory still feels it has something to say about relationships. We are all happier in a stable unit than cast out on our own. Single people are sadder than married people, orphans are less happy than children with both parents. But I’m not sure that we need a special theory to tell us this or that attachment theory’s explanations rise much above the level of psychobabble.

Click here to read the first chapter of God's Philosophers: How the Medieval World Laid the Foundations of Modern Science absolutely free.

10 comments:

SteveG said...

I was hoping you could clarify something in this post...

In saying that the twin studies showed about 50% correlation with each other, are you saying that their 'sameness' (for lack of a better term) was 50%? If not, then can you explain exactly what is meant? and if yes, then would the statement that the had 50% non-correlation (non-sameness) also be true?

I'll refrain from any other comment until I am sure I understand what's being said here.
Thank!

James said...

As I understand it, identical twins share about 50% of personality traits (including religious and political predilections and probably intelligence if we could figure out how to measure it) compared to two random people (who will obviously have some similarities). They don’t share the other 50%. What’s important for AT is that this other 50% has no explanation at the moment. It is not the case that children share much at all with adopted parents or siblings (maybe up to 10% in some studies but in general well within expectation bias errors). What children share with their natural parents is about 25% which is what we’d expect from a purely genetic cause.

SteveG said...

James:

I mean this with all due respect (and I really DO respect you a great deal), but these two posts taken together are simply uniformed and clumsy.

It's very obvious to anyone even remotely familiar with AT beyond the Wikipedia level that you are simply speaking on an area of expertise with which you are largely unfamiliar.

I honestly don't know where to start with this.

If you have any interest, I'll give it some thought and try to see if I can explain where you've gone wrong in your understanding of AT (though I've done some little bit of that in my comments to no avail), but if you are convicted that it's rubbish and have said all you want to say, let me know and I'll beg off.

James said...

Steve, reading your comments, I felt you were only really asserting that stuff was tried and tested. I didn't get any information on (say), a specific prediction made by AT, how that prediction is measured and how genetic factors are accounted for. For instance, on breast feeding, what measurements have been taken to show that it has an effect beyond its dietary contribution. On parenting, what effects can AT predict and explain and what are the results of these tests. How does AT account for the lack of any correlation between adopted children and adopting parents?

Please be assured, I want to know. But I have not been able to find out this information anywhere, least of all, I'm afraid in your comments.

SteveG said...

That's fair.

I am a layman on this myself, whose simply researched and read a great deal on it myself, so I may not be able to satisfy what you are asking.

I get the sense though that we may be speaking past each other a bit, and that more generally we may be be speaking about AT in different contexts.

Let me see what I can offer, but it may take some time (work has suddenly heated up a bit the past day or two).

SteveG said...

James:
I had a bit of an epiphany as to why I think we may be talking past each other, and where I think you've gone wrong. :-)

You seem to be arguing that attachment theory suggests same/similar/correlated outcomes among different people. This is not what AT is saying though.

You say

At its base, traditional attachment theory makes the following prediction – take identical children and separate them at birth...According to attachment theory we should see marked differences between the children once they grow into adults."

...and later you focus on twin studies to see how their outcomes correlate to one another.

But I think this is a very faulty description of what AT is.

AT in fact is something very different. Here's a basic explanation from a paper I read recently on secure attachments and risk for psychatirc illness..

"Attachment theory holds that secure attachments, and the attuned infant-caregiver interactions that produce them, are crucial to healthy psychological development."

It's simply not focused on correlation in the way you are. Instead it's focused on impact to psychological health.

So I think we need to start by decoupling the types of results you seem to be expecting form AT from what it actually is looking at.

With this (psychological health) as context hopefully we can't a much differnt picture of AT.

In answer to your question on for what kind of prediction it might make and how to test that, etc.

From that same article I mention above, a prediction that children who are not securely attached might suffer from a higher degree of mental illness is made. And in fact that's exactly what is found...

When children with disorganized attachments [as determined by the classic strange situation test] were followed for 17 years, they also showed the greatest vulnerability to mental disorders later in life.

Back to the issue where I first interjected myself, that of the effect of divorce on outcomes.

AT would easily predict that a situation where attachment was damaged between both parents, the outcomes would be lower than in other situations. And by your own account, that is exactly what we see...even when compared to children who lose a parent to death (I explained the differences in terms of AT in the original comment).

With breastfeeding...AT would predict that those activities which more closely bond infants to caregivers will reap beneficial developmental affects.

The ‘vehicle’ by which this bonding delivers it’s benefits is dietary, but it’s the bonding activity that allows for the delivery and as AT predicts, it has a positive impact on development.

Likewise…the cry it out vs. comforting of infants. Again, the prediction is that the activities that promote bonding also promote development. That's been well established now.

You seem to suggest all of this is psychobabble, but when researchers are looking at how to treat or promote mental health, this has a real impact.

When researchers discover that Attachment Based Family Therapy has an 81% success rate in treating depressed adolescence This goes far beyond the realm of psychobabble…..in my humble opinion.

I don’t know if this helps put us on the same page or not, but if I want to be sure one point gets across…it’s that I think your evaluation of AT as rubbish based on judging it in terms of sameness between individuals (whether twins, or parents and children) is mistaken.

James said...

Hi Steve,

Thanks for your more detailed response. We may be talking past each other but I hope it is mutually useful to explore our positions on this important question.

***"Attachment theory holds that secure attachments, and the attuned infant-caregiver interactions that produce them, are crucial to healthy psychological development."

It's simply not focused on correlation in the way you are. Instead it's focused on impact to psychological health. ***

Well no. Your quote says “development” which would affect outcomes if it is true. Health is a different question altogether. Actually, I am happy to accept that healthy attachments may lead to healthier people purely because happy people do tend to be healthier. But I’ve always been careful to talk about outcomes when children grow up.

***From that same article I mention above, a prediction that children who are not securely attached might suffer from a higher degree of mental illness is made. And in fact that's exactly what is found... When children with disorganized attachments [as determined by the classic strange situation test] were followed for 17 years, they also showed the greatest vulnerability to mental disorders later in life. ***

But does this allow for genetics? How can you tell that the disorganised attachments are not caused by the mental problems of the parents being genetically shared by the children? In fact, your quotation has been disproved by adoption and twin studies which show no correlation between the adopting parents and their adopted children’s outcomes. Only between natural parents and their children do we see the correlation that this article reports. Therefore, the correlation must be genetic and not caused by AT. This is what our entire debate has been about. Can you see this? If I haven’t been clear then I’ll try and go over it again. But this is the fundamental mistake of AT – the fact that it assumes what it sets out to prove.

***AT would easily predict that a situation where attachment was damaged between both parents, the outcomes would be lower than in other situations. And by your own account, that is exactly what we see...even when compared to children who lose a parent to death (I explained the differences in terms of AT in the original comment).***

Again, you have not allowed for genetics. Linked below are two studies that do. What they both show is that all children of divorced parents (both natural and adopted) get unhappy and many drink too much and take drugs. AT may have something to say about this. However, the second study found that adopted children whose adopting parents divorce, on average, show none of the poor outcomes that natural children of divorcees show and none of the social disadvantage. What this means is that it is not the divorce itself which leads to poor outcomes, but the genetic inheritance of the children of parents who divorce. If the divorce itself was the cause then adopted and natural children would surely be affected in the same way. But they aren’t. The other study reaches a similar conclusion through a twin studies methodology.

http://newsinfo.iu.edu/news/page/normal/5960.html
http://www.apa.org/journals/features/dev364429.pdf

I think these studies are the key to our debate because they show that AT is wrong to say divorce directly affects outcomes. Let me know if you want me to go into more detail about them.

***With breastfeeding...AT would predict that those activities which more closely bond infants to caregivers will reap beneficial developmental affects. The ‘vehicle’ by which this bonding delivers it’s benefits is dietary, but it’s the bonding activity that allows for the delivery and as AT predicts, it has a positive impact on development.***

So how can you separate the bonding from the diet? What I want to see is a comparison between bottle-fed, breast-fed and expressed breast-fed babies. The last lot use breast milk from a bottle. I’ll bet my favourite shirt that there is no difference between the breast and expressed milk babies.

***Likewise…the cry it out vs. comforting of infants. Again, the prediction is that the activities that promote bonding also promote development. That's been well established now.***

No, this been disproved by twin studies. Any study that takes no account of genetics is useless. This is what I’ve been explaining at length but you seem to have a blind spot to it.

***When researchers discover that Attachment Based Family Therapy has an 81% success rate in treating depressed adolescence This goes far beyond the realm of psychobabble…..in my humble opinion.***

Compared to 47% for the group not treated at all. Anyway, I think you’ll find all “talking cures” claim success rates at about this level. So does homoeopathy and for much more crunchier conditions than teenage depression. So, incidently, do combined drugs and psychological treatments (which means, unfortunately, that we have no cure for depression beyond placebo effects).

Best wishes

James

SteveG said...

Hi Steve,

Thanks for your more detailed response. We may be talking past each other but I hope it is mutually useful to explore our positions on this important question.

Absolutely…at least on my end…you’ve definitely caused me to think this through from a different perspective.


But does this allow for genetics?

I think this is a big part of my struggle in this discussion. In my admittedly lay understanding of the topic, I have not seen a hint of why it wouldn’t. I’ve never thought of the two as being pitted against each other but as complimentary.

I don’t think (and hope I didn’t imply) that AT is an all explaining concept that leaves room for nothing else.

I guess I’d ask the question back…why wouldn’t it allow for genetics as a factor…even a dominant factor?

How can you tell that the disorganised attachments are not caused by the mental problems of the parents being genetically shared by the children?

I’ll ask back…how we can be sure that the mental problems of the parents are not attachment related themselves? How do we know they are being genetically shared?

In fact, your quotation has been disproved by adoption and twin studies which show no correlation between the adopting parents and their adopted children’s outcomes.

I am honestly curious…what outcomes are being measured in those studies?

Also, as you explained it, the twin studies showed 50% correlation, what is the 50% non-correlation due to? Can AT not coexist somewhere in the ‘unknown environmental factors?’ Or am I still missing what you are proposing?

Therefore, the correlation must be genetic and not caused by AT. This is what our entire debate has been about. Can you see this? If I haven’t been clear then I’ll try and go over it again. But this is the fundamental mistake of AT – the fact that it assumes what it sets out to prove.

But didn’t you do the same above? You assumed that the parent’s mental problems were genetically passed to them, and then genetically passed to the children? Other than the genes of the gaps analysis, what evidence is there for that? Are there specific genetic markers for each correlation?

Again, I am not arguing there is NO hereditary predisposition involved, but is it really as clear cut as you are suggesting?

But they aren’t. The other study reaches a similar conclusion through a twin studies methodology.

http://newsinfo.iu.edu/news/page/normal/5960.html
http://www.apa.org/journals/features/dev364429.pdf


But to my reading, these studies you’ve cited don’t go nearly as far as you do. They are filled with ‘this explains some of the effects of divorce’ and caveats about inadequacies of the studies involved, etc.

They are extremely interesting and seem reasonable, but they don’t seem to have the evidence to dismiss non-genetic (and in this case attachment) related effects…more importantly, they don’t seem to even try to say that (though in honesty, I haven’t yet read the longer one in its entirety).

I think these studies are the key to our debate because they show that AT is wrong to say divorce directly affects outcomes. Let me know if you want me to go into more detail about them.

Again…an article on how Divorce and children: Genes at the root of some problems doesn’t seem to be saying that AT is wrong here…it seems at best to give part of the answer. The content says nothing different. At best it says AT doesn’t have ALL the answers…but it doesn’t seem to say it has none of them. What am I still missing?

No, this been disproved by twin studies. Any study that takes no account of genetics is useless. This is what I’ve been explaining at length but you seem to have a blind spot to it.
This one leaves me scratching my head….

*It’s well established that elevated cortisol levels for an extended period slow brain growth in infants.
*It’s well established that cortisol levels in infants left to ‘cry it out’ during sleep (rather than being comforted to sleep) experience elevated cortisol levels due to the lack of comforting. They even know that the cortisol levels fail to go down when they fall asleep, but remain elevated throughout.

...unless the twin studies have specifically used twins to compare the long term effects of the two methods, it seems to be going too far to say that the studies disprove this effect. Perhaps in a culture where one or the other method is dominant, you will simply not be able to detect this.

Yes…I know…the twins have 50% correlation (which is to this point still very vague…what kinds of correlations…do they have any bearing on this specific issue), but I still can not see how that disproves the fact that children who experience the cry it out method of parenting actually have parts of their brains that are smaller.

The parts involved to my understanding are those that deal with emotional control rather than intelligence…would that show up in the twin study (i.e. more emotional maturity rather than a difference in intelligence)?

Compared to 47% for the group not treated at all. Anyway, I think you’ll find all “talking cures” claim success rates at about this level. So does homoeopathy and for much more crunchier conditions than teenage depression. So, incidently, do combined drugs and psychological treatments (which means, unfortunately, that we have no cure for depression beyond placebo effects).

I’ll have to dig up the research on this that I’ve read recently (I did a quick search but time is pressing so I'll have to try again later if you are interested), but this is not what I’ve seen them to be saying.

The latest studies I’ve seen are discovering that antidepressants have a very minuscule positive impact compared to placebo, but that therapy has a much higher rate than both placebo and antidepressants.

In fact the data you mention regarding combined drug and treatment outcomes is showing that the success is do mostly to the treatment aspect, and almost not at all the drugs.

That means that the outcome with drug + treatment is almost the same as only treatment, but the opposite is not true…drugs without treatment has results closer to doing nothing at all.

I have no information at all on homeopathy treatment so I won’t comment on it at all.
Cheers!

James said...

Steve, I'll get back to you when I can, but probably not for a week or so. Happy Easter.

SteveG said...

No problem...I understand it's tough to find the time.

Happy and Blessed Easter to you and yours as well!