YT YouTube2Text

YouTube2Text → Transcripts → How DNA Makes Us Who We Are | Robert Plomin | Talks at Google

How DNA Makes Us Who We Are | Robert Plomin | Talks at Google — Transcript

by Talks at Google · 9,453 words · 1,399 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00[Music]
0:09Thank You Jana and thank you all for
0:11coming on your lunch hour it's great to
0:16talk to a group like this I usually talk
0:18to academic groups or people in
0:19education so I'll be curious to see what
0:21you make of all of this but I'm I've
0:24been informed I'm not here to tell you
0:26things that are relevant to what you do
0:28at Google it's more about general
0:30interest in genetics and behavior when I
0:34started in graduate school in the early
0:3670s you might find it hard to believe it
0:38was dangerous to even talk about
0:40genetics psychology was so dominated by
0:42environmentalism the view that you are
0:45what you learned that it was dangerous
0:48professionally and sometimes personally
0:49even to mention genetics we were the
0:52textbooks of the time this in the 1970s
0:54we tell you if you look up schizophrenia
0:56it would tell you that schizophrenia is
0:58caused by what your mother did to you in
0:59the first three years of life and since
1:02then a mountain of evidence has
1:04convinced most like most scientists that
1:06genetics is a major factor in most
1:09aspects of behavior and today I'd like
1:11to just because I want to leave time for
1:13questions I'm gonna focus on one aspect
1:16of behavior school achievement in part
1:19because this is a talk I gave at New
1:21Scientist live on Saturday so what do
1:24you think is responsible for how well
1:28you did at school so it's easy to think
1:32about tangible environmental factors
1:34like your parents your teachers your
1:37schools but what I'd like to convince
1:39you is that something intangible DNA is
1:43the major factor responsible for
1:45differences between children in their
1:48performance at school the environments
1:50important to but the way it works is
1:52very different from the way we thought
1:54it worked before we took genetics into
1:56account
1:57previous studies of the environment are
1:59confounded because things that run in
2:01families could run in families for
2:03reasons of nature genetics as well as
2:06nurture but if you ignore the nature and
2:07assume everything's nurture you make
2:09some very wrong conclusions about the
2:12environmental in
2:13as on behavior and this is all a lot
2:16more relevant and important now with the
2:20DNA revolution so I think that'll
2:22probably interest you even more and I'll
2:24try and get through the first part of my
2:27talk quickly in order to spend more time
2:28on the DNA revolution so this is my book
2:31blueprint I'm very sorry that penguin
2:33didn't get it here in time but I bet you
2:35it'll come right after my talk sounds
2:37law but penguin is usually very good at
2:40these things
2:41the first half of the book is about
2:42genetics basically how do we know
2:45genetics is so important and what have
2:47we learned about the environment by
2:49studying it in genetically sensitive
2:51designs and then also some of the
2:53implications of it and as I say instead
2:55of psychopathology like psychiatric
2:58disorders or personality I'm going to
3:00just focus on one aspect of this huge
3:02area school achievement then the second
3:05part of the talk is about in the book is
3:07about the DNA revolution identifying
3:09specific DNA differences that account
3:12for differences in behavior and that's
3:14what's going to change everything so I
3:16think it's very important that it's the
3:18reason I wrote the book now is to say we
3:22really have to discuss some of these
3:23things because they're happening now so
3:25just to make sure on the same page you
3:28know it's a genetics heritability these
3:31are difficult concepts for people to
3:33grasp but when you ask people how
3:35heritable is eye color that is how much
3:37is it due to inherited DNA differences
3:39I've done these surveys and people on
3:41average would say it's about 90%
3:43heritable that means it's almost
3:44entirely due to inherited DNA
3:47differences but when you turn to complex
3:49traits like wait how irritable do you
3:52think weight is and the pop if you do
3:55population surveys probably not of
3:57groups as smart as this one but people
4:00would say about 20% 30% they think there
4:02might be some genetic influence but it's
4:04mostly environmental but in fact it's
4:0670% heritable across a wide range of
4:08studies including DNA studies and that's
4:11really important because you know it
4:13means of the differences between people
4:15in this room you all seem to be skinny
4:16so for me the differences are
4:20substantial but the major factor is an
4:23DNA differences which is really
4:26important in our society with fat
4:27blaming and you know fat shaming the
4:29sorts of stuff that goes on you know to
4:31recognize people like me have a lot it
4:33find a lot easier to put on weight than
4:34you skinny people and a lot harder to
4:36lose weight so when you turn to school
4:39achievement though people underestimate
4:41genetic influence even more they think
4:44it's about maybe a little bit of genetic
4:45influence but not much because
4:47achievement is is the means the
4:50etymology of it is by dint of effort as
4:53opposed to ability which people say oh
4:56that's genetic but in fact school
4:58achievement tested school achievement in
5:00English in schools in England is about
5:0360% heritable meaning 60% of the
5:07differences between kids in their tested
5:09school achievement is due to inherited
5:11DNA differences and you can see how
5:13important it is to know that because we
5:15tend to assume it's all environmental
5:17it's all due to what how good a school
5:18you went to how much your parents push
5:21you on this but parents don't have
5:23nearly as much control as they think
5:25they do so that's one of the messages I
5:27want to get to the most misunderstood
5:30concept in genetics is heritability
5:33these six syllables and so it's
5:36important just to spend a minute talking
5:38about it we're only talking about what
5:40makes people different why some of us
5:42are heavier than others why some of us
5:44do better at school than others of our
5:47three billion base pairs of DNA in the
5:49double helix of DNA 99 more than 99% of
5:52those DNA bases are the same for
5:54everybody but 1% differs and we're asked
5:58that the 90% is what makes us human
6:01but the 1% is what makes us different
6:03genetically so what we're asking is to
6:06what extent does the 1% of DNA that
6:08differs between us make us different in
6:11traits like school achievement we're
6:14only talking about describing what is
6:16rather than what could be so we study
6:20like twins and adoptees and DNA in a
6:23representative English sample but that
6:26only tells us about what is now say in
6:29terms of weight given our genetic
6:31differences and our environmental
6:32differences now how much do genetic
6:35differences make a difference
6:36if you change the environment if you
6:38study a poppy different population in a
6:40different time that would change but
6:42that's a good descriptive statistic that
6:45is sensitive to the environment in which
6:47you study it and we're only describing
6:49the normal range of variation I mean
6:51sure it's an issue you all face that you
6:53can only talk about you can't generalize
6:56beyond the sample that you've described
6:59so these samples that we study are may
7:01be representative of ninety-five percent
7:03of the population but they don't include
7:05say families where parents are abusing
7:07the kids because those people don't
7:08participate in studies nor does it
7:10include the genetic extremes of single
7:13gene mutations that are very severe and
7:16debilitating but they're very very rare
7:18one in a hundred thousand one in two
7:20hundred thousand so if you have a sample
7:22of a few tens of thousands of people you
7:24don't have those in the sample and
7:26lastly I want to emphasize we're talking
7:28about genetic influences these are
7:31nudges probabilistic propensity x'
7:34they're not innate they're not immutable
7:38and they're not deterministic so I have
7:42to whip through this but then those are
7:45really important points and people you
7:47scratch the surface what people know
7:49about genetics and they somehow think
7:50you're talking about innate immutable
7:52influences which is the case for single
7:55gene disorders which is how most of us
7:57learn about genetics from Mendel you
7:59know there are thousands of single gene
8:00disorders as I say they're very rare but
8:03if you have the gene on chromosome four
8:05the form of a gene called an allele for
8:07Huntington's disease you will die from
8:09Huntington's it's hardwired
8:11deterministic it doesn't matter what
8:13your environment is or anything else the
8:16promise when we talk about complex
8:18traits not just psychological traits but
8:20most of a medical burden in society is
8:22caused by common disorders and they're
8:26not caused by single genes they're
8:27influenced genetically but by many many
8:29genes a small effect that makes them
8:31probabilistic rather than deterministic
8:34and that's a really difficult jump for
8:36people to make so any one of the methods
8:38we use for a hundred years to study the
8:41extent to which genetic influences is
8:43important is a sort of biological
8:44experiment where there are two types of
8:46twins I'm sure you all know 1% of all
8:49births are twins
8:50one third of those are identical twins
8:52called mono zygotic it's a single zygote
8:56that is a fertilized egg that in for
8:58reasons unknown in the first few days of
9:00life divides in two and those are clones
9:03of one another they're genetically
9:04identical if you sequence their DNA so
9:07they're 100% alike genetically the rest
9:10of the twins are like any brother and
9:12sister who happened to be born at the
9:14same time because their mother had two
9:15eggs in the womb at the time and were
9:19fertilized so they share 50% of their
9:22genes they're 50% similar so you'd
9:24predict that any traits a school
9:26achievement that is influenced
9:28genetically you'd have to predict
9:30identical twins will be more similar
9:32than fraternal twins and you can use the
9:35extent to which that's true to estimate
9:37heritability and so when I came to
9:41England in 1994 I began the twins early
9:43development study which is the world's
9:45largest study of development of twins we
9:47started with about 15,000 pairs of twins
9:50and we studied them 14 times through
9:52young adulthood most recently at 22
9:54years of age and about 10,000 pairs
9:59continue to participate importantly as
10:01I'll mention later we also have DNA on
10:03them so they led the way in terms of
10:05some of the DNA analyses I talked about
10:07I was interested in studying things that
10:09hadn't been studied much before like
10:11language development in the early child
10:13years or behavior problems which
10:15developed very early on and are highly
10:17heritable in early childhood like
10:19attention deficit problems but when they
10:22got to school because I was interested
10:24in cognitive development I really wanted
10:26to study kind of the business end of
10:29cognitive ability and that's school
10:31achievement so this summarizes 15 years
10:33of work in the twins early development
10:36study these are the heritability
10:38estimates based on identical and
10:41non-identical twin correlations across
10:44all the key stages which he stated one
10:46at seven going to nine to twelve sixteen
10:49is GCSE 18 is a-levels and you can see
10:54that the average heritability exceeds
10:5660% at all ages including the very first
11:00grade
11:01you know a lot of people would have
11:03expected that in first grade it's more
11:05your performance is more function of
11:06what your parents did but it's actually
11:09just as much genetic as a levels or GCSE
11:12and in the behavioral sciences
11:14explaining 5% of the variance is a very
11:17big deal
11:18explaining 60% of variable variance
11:21meaning differences between people is
11:23sort is off the scale so it's a huge
11:26effect and oops my end got misplaced
11:31it's not just the twin method there's
11:34many other methods that are used if the
11:36twin methods like a biological
11:37experiment the adoption method is like a
11:40social experiment so family members
11:42share genes as well as environment
11:45nature and nurture for law we've known
11:48forever everything runs in families
11:49parents who do well at school have kids
11:52who do well at school for decades people
11:54said no problem that's just nurture
11:55those parents provide a better
11:57environment for their kids but when you
12:00start thinking about genetics to say
12:01well but they share 50% of their genes
12:02could it be nature not just nurture the
12:06adoption design separates that by
12:08studying parents who are related
12:10genetically to their kids
12:11these are birth parents who relinquish
12:13their kids for adoption at birth and
12:15then so they share nature but not
12:17nurture and then adoptive parents who
12:21adopt those kids early in life and share
12:23nurture but not nature so the book also
12:27describes a study I've done for 40 years
12:30at the a called the Colorado doctor
12:32Colorado adoption project which is about
12:35a 250 families adopted families where we
12:37had data on the birth parents the
12:39adoptive parents and the adopted kids
12:41longitudinally through adulthood from
12:43infancy to adulthood and this summarizes
12:4620 years of research the we also had
12:51matched non adoptive families who share
12:53genes and environment with their kids so
12:56general learning ability which we call G
12:59to avoid the word intelligence which is
13:01just like a red flag to a bull for
13:04people but there is this construct of
13:07general cognitive ability which is one
13:08of the better measured traits so just
13:12don't have a knee-jerk reaction against
13:14it
13:15but what's been known for a long time as
13:17it runs in families and that parents and
13:20their children increasingly resemble
13:22their each other as the kids grow up so
13:25that shows you it the parent-offspring
13:27correlation when parents share genes an
13:29environment of their kids goes up to
13:30about 0.3 ish by late adolescence so is
13:35it nature or nurture for a very long
13:37time it was assumed to be nurture and
13:39it's not an unreasonable hypothesis but
13:42if it's nurture that is the parents are
13:44giving the kids the environment they
13:46need to develop cognitively you'd have
13:48to predict that the adoptive parents
13:50ought to be just about as similar to
13:51their adopted kids and the correlations
13:54are actually zero so that suggests it's
13:59not nurture in the sense we've thought
14:02about it as systemic effects of the
14:04parents say well then is it really
14:06nature could it really be that these
14:08birth parents who don't share
14:09environment but shared genes are
14:12correlated as much as parents who Bri
14:14are the kids and the answer is yes so
14:16it's a powerful demonstration of the
14:19importance of nature and the
14:21unimportance of nurture as we've defined
14:23it in terms of systemic effects of the
14:26family environment the environment is
14:29important because heritability is not a
14:31hundred percent it's more like fifty
14:33percent on average but it's a very
14:36different sort of environment than we
14:38anyone ever thought about from Freud
14:40onwards so the first part of the book
14:42and I'm whipping to the conclusions here
14:45because I want to get on to the
14:47molecular genetics it talks about how do
14:50we know genetics is important and just
14:52how important are they but then some of
14:55the most important findings are about
14:56the environment when you control for
14:58genetics so the second big finding is
15:00this one I just alluded to that the
15:03environmental effects are important but
15:05they're not nature and in the sense of
15:07systemic effects of the environment
15:09adoptive children you know a third of
15:11adoptive families adopt a second child
15:13and they're genetically unrelated to
15:15each other they correlate zero with
15:19siblings who grow up together and share
15:21genes they correlate substantially you
15:23know 0.24 personality and 0.3
15:27for say cognitive development so for 30
15:32years people been trying to find out
15:33what are these mysterious factors that
15:35make two kids in a family different from
15:37one another and it could be lots of
15:39things accidents you know or illnesses
15:41or different peers or the parents
15:45treating them differently
15:46but after 30 years of research no
15:48systematic factors have been found and
15:50so I've come to the conclusion called
15:53the dark hypothesis that the effects are
15:55essentially idiosyncratic stochastic
15:59random in a word chance so they're
16:02unsystematic it could be like Bill
16:05Clinton in his biography talks about why
16:08did he go into politics and he says it's
16:09because when he was grow at 16 he shook
16:12JFK's hand see that would be a good
16:15example of this I mean it's not a
16:16systematic variable you could measure
16:18very well and maybe you know knowing
16:20Bill's history here you can't bet a lot
16:22on its veracity I suppose but that's the
16:25sort of thing you get these
16:26idiosyncratic experiences like why are
16:29you doing what you're doing I know in my
16:30case it's all these chants sorts of
16:33events just little nudges in one
16:36direction or another that snowball so
16:40that's what we think it's about and the
16:42other third finding is what we that's we
16:44call non-shared environment it's not
16:46environment shared by kids growing up in
16:48the same family going to the same
16:49schools the third finding is called the
16:52nature of nurture it's the idea that
16:53what looks like systemic effects of the
16:55environment are actually mediated
16:57genetically so you know correlations
17:01don't imply causation right everyone
17:03knows that but if you see a correlation
17:04like I do once a week in the papers
17:07parents do this and the kids are like
17:09that so parents who read a lot the kids
17:11have kids who read better when they go
17:12to school it's so hard to resist an
17:15environmental interpretation but
17:17correlations don't imply causation and
17:20if you start saying what about genetics
17:21it'll drive you mad because you know
17:24they share 50% of their genes it could
17:26just be parents who read have kids who
17:28read but more increasingly it's sort of
17:31the nature of nurture that the
17:33correlation goes in the opposite
17:34direction from the way we think it goes
17:36parents are responding to differences in
17:38their kids we
17:39I really see as a grandparent with six
17:41kids one grandchild does what your
17:43thought grandchildren are supposed to do
17:44she'd let you read her all day long she
17:47loves words and reading but then the
17:50first one I had she didn't want to read
17:54she wanted to you know kick a ball
17:56around she wanted to be active it almost
17:58would have been abusive I said no you're
17:59a grandchild I'm a grandparent you sit
18:01there and I read to you that's what
18:03you're supposed to do but it doesn't you
18:05know we're responding to differences in
18:06the kids and that's as it should be and
18:09also in education we should be
18:11recognizing that kids differ try to
18:13minimize the weaknesses maximize the
18:15strengths which I'll try to get on -
18:17oops so if you I know I went through
18:19this quickly but if you look at these
18:21three things together they lead to the
18:23title of the book which I agree is
18:24provocative and misleading but what it
18:27what I'm trying to say is that DNA is
18:29the major systemic force making us who
18:32we are as individuals it's systemic in
18:35the sense I try to emphasize that
18:36because the environment is important but
18:38it's not systemic and so what I'm saying
18:41to tie these together is if you were
18:43cloned and your clone was reared
18:46obviously in a different woman
18:47prenatally grew up in a different family
18:49with different parents went to a
18:51different school had different friends
18:52had a different job that that clone
18:55would be very similar to who you are now
18:58not just in school achievement but in
19:00personality and psychopathology in fact
19:04that clone would be as similar as
19:05identical twins reared together so a
19:09school achievement they correlate about
19:100.7 identical twins reared together
19:13being reared apart doesn't make you any
19:15less similar so you can you see that
19:20that's talking about the importance of
19:21nature the unimportance of the systemic
19:24family environment and it's not just the
19:25hypothetical experiment I don't know if
19:27you know who this is
19:29nope yeah you'll know in just a minute
19:34this is Bobby who grew up in Long Island
19:35and a very wealthy family he went to
19:38university in upstate New York and on
19:40the first day everyone's calling Bobby
19:41Eddie girls were coming up and giving me
19:43a hug and kissing Oh Eddie it's so good
19:45to see so you thinking psychology
19:47experiment looking for the cameras
19:48but then he met Eddie and they quickly
19:52worked out they had the same birth date
19:53they were adopted from the same adoption
19:55agency in New York and the publicity
19:59that came from that led to the rare
20:01circumstance of a third identical
20:03triplet because I told you that the
20:05zygote separates for reasons unknown
20:07sometimes one of those I goat separates
20:09again so these are clones of one another
20:11they have the same DNA sequence and the
20:14film this is a documentary film that won
20:17the Sundance Award last year called
20:19three identical strangers and it's
20:20available on streaming and I really
20:22recommend it as a dramatic illustration
20:23of the points I'm trying to make but it
20:26is just an anecdote and an illustration
20:28and just watch the first half the second
20:31half is a very bad story because did you
20:35think why weren't they separated and
20:37didn't know about each other's existence
20:39why was one put in a lower-class home
20:42one in a middle-class home when an
20:43upper-class home a Nutty psychiatrists
20:47who was actually out to prove nurtures
20:50important he thought it was going to be
20:51a definitive experiment identical
20:53triplets put one in a lower class room
20:56one in middle-class filming one an
20:57upper-class family they already had
20:58adopted another kid from the agency so
21:00they knew what the parenting styles were
21:01alike so they made them as different as
21:03possible and he was he was a Freudian
21:06you know this is in the 50s just assume
21:09the environment it's all about nurture
21:10but then when the results started coming
21:12out he actually buried them so they're
21:14under lock and key at Yale Medical
21:16School till twenty sixty six never been
21:19published but this film exposes and is
21:21probably at least half a dozen if not
21:23more identical twins who have been
21:25separated as a result of this he was a
21:28psychiatrist to that adoption agency
21:31right so it is just an anecdote but
21:34behind it is systematic data on twins
21:36reared apart I did a study in Sweden of
21:38over a hundred pairs of identical twins
21:40reared apart as well as other adoption
21:42designs like biological parents and
21:44their adopted away kids adoptive
21:46siblings biological siblings adopted
21:48apart and increasingly DNA you can use
21:51DNA itself in unrelated individuals to
21:54estimate heritability but it would take
21:56too long to explain that now but it's
21:57kind of the hot new
21:58thing so they all converge on the
22:01conclusions that I mentioned so I would
22:04have been well just mention implications
22:08of that research and the main thing to
22:09emphasize there's no necessary policy
22:11implications this is an old-fashioned
22:13view that policy depends on values as
22:16well as knowledge and I'm increasingly
22:18cynical about this that better decisions
22:20are made with knowledge than without
22:21more often I think if your data your
22:24research agrees with the values of the
22:28government they'll use it but it doesn't
22:30actually inform the policy very much but
22:33here's the one thing that got quite a
22:34bit of attention in terms of schools if
22:36you see what I've been talking about can
22:38you see how this makes sense schools
22:39matter but they don't make a difference
22:41so schools matter a lot they kids have
22:45to learn basic skills of literacy and
22:47numeracy and inculturation but they
22:50don't make a difference kids going to
22:51the same school aren't any more similar
22:54as if then if they had gone to different
22:56schools and part of that's because 60%
23:03of the differences are genetic anyway
23:05and the environmental effects are not
23:08systemic so one quick fact that's
23:11important for you guys because you look
23:12like you're at risk of child bearing age
23:15and you know how in England it's we have
23:18this crazy selective system for
23:20secondary schools parents spend hundreds
23:22of thousands of pounds to get their kid
23:23into a better school better school why
23:26mostly based on Ofsted ratings in the
23:28schools people don't ask about the
23:31effect size of the Ofsted rating so they
23:33cost about 10 grand each you know it's
23:34the really good ratings of schools
23:36atmosphere teachers support bullying the
23:39whole Schmeer but then they are the
23:43primary difference between schools in
23:45the league tables so the question is how
23:49much of the variance how many of much
23:50are the differences between kids and
23:52their GCSE scores is accounted for by
23:55Ofsted ratings of school quality so some
23:58kids go to schools where they had very
24:00high ratings and some go to schools were
24:02there very low ratings so heritability
24:04counts for 60% Ofsted ratings account
24:08for 4%
24:09which is not a noticeable difference so
24:12you get these mean differences but
24:14you've got to ask about the effect size
24:16and the effect size is very small if you
24:19correct for socioeconomic status because
24:21kids aren't randomly assigned to schools
24:23it goes down to one percent that's not a
24:26difference you can even detect you know
24:28with you know your experience you need
24:31very large samples and statistics to
24:33detect it and then finally what looks
24:36like systemic effects of schools are
24:39often genetic effects in disguise so one
24:42quick example of this is that you you
24:45probably know that there's a big GCSE
24:47difference between kids and selective
24:49schools and non selective schools it's a
24:51whole grade difference that's a
24:53correlation between going to a selective
24:56school and how well you do on GCSEs but
25:00it's hard to avoid interpreting that
25:03environmentally the selective schools
25:04have more resources you know better
25:06playing field probably maybe better
25:08teachers but it isn't they're selecting
25:14for genetically influenced traits of
25:15school achievement and ability and it's
25:17a self-fulfilling prophecy if you select
25:20the kids who do the best at school and
25:22have the greatest ability they're gonna
25:24do the best at school and have the
25:25greatest ability if you're correct you
25:27merely just correct GCSE scores for what
25:30the schools select on earlier
25:32achievement and ability there's no
25:35difference in performance there's no
25:36added value of selective schools now a
25:39lot of parents will say but in their
25:42cups at least well I'm not just sending
25:44them there to get better or school
25:45achievement you know they get better
25:47contacts half of the judges in the UK
25:49are from the 7% of selective schools so
25:53maybe there is an access difference or
25:55whatever but if you're really just
25:57thinking about achievement the evidence
25:59is that it doesn't matter it doesn't
26:01make a difference now it might matter
26:04there might be nicer places to be I'm
26:06not even convinced of that there's
26:08greater self-harming at these selective
26:09schools my grandson I got a lot of
26:11pressure because my son wanted to send
26:13his kid to one of these selective
26:14schools nearly 30,000 a year when he had
26:17a perfectly good
26:19comprehensive and what was I gonna say
26:26well I just I resisted that because I
26:30don't believe it makes it oh yeah and a
26:31kid like him
26:32Tristram no less is a star in his school
26:36the comprehensive elementary school but
26:38when these kids get to these high
26:39pressure selective schools they're no
26:41longer a star they're lucky to be kind
26:43of average and so and there's a lot of
26:47pressure and they kick kids out if they
26:49don't do well on the pretest and stuff I
26:51mean it's a crazy stressful system that
26:54I think destroys learning so and the
26:57enjoyment of learning so I think we need
26:58to think about education in a different
27:01way and thinking about it genetically
27:02most of what's going on is kids are
27:04selecting and modifying and creating
27:06environments in part correlated with
27:08their genetic propensities so kids in
27:10the same classroom can experience
27:12different environments based on their
27:14genetic propensities they just ask more
27:16questions
27:17you know they follow up on stuff and I
27:21think we need to move from this passive
27:23model of imposed instruction which is
27:28the you know instruction is from Latin
27:30into our to shove in which is the way we
27:33think of it we're shoving in this
27:34national curriculum in their head
27:35instead we need to move to an active
27:37model of shaped environments in which
27:39kids select environments that are
27:40conducive that are fitting with their
27:43propensity
27:44you know not everyone this might be
27:46heresy here but not everyone needs to
27:48know advanced math for example so we
27:51need to think about what kids are good
27:52at and maximize their strengths and
27:54minimize their weakness and mostly have
27:56them learn to learn because it's well
27:57known I mean everybody says I don't know
28:00if it's true but that the kids in
28:02elementary school today are going to be
28:03doing jobs that don't exist now they've
28:06got to learn to learn it's not a matter
28:08of specific skills that they learn they
28:10need to learn to learn and mostly they
28:12need to enjoy learning but our test
28:14obsessed high-stakes testing culture is
28:18really destroying any sort of enjoyment
28:21of learning
28:22okay I'll get off the soapbox and I
28:25would be happy in my that was 35 years
28:28of my research and I would have been
28:29happy if it ended there but then along
28:31came the DNA revolution and as I say I
28:34think it's going to change everything
28:35it'll allow us to predict from DNA alone
28:37problems and promise from birth because
28:41your DNA doesn't change throughout life
28:43and that allows us to move towards
28:45prevention so rather than waiting until
28:47problems occur like waiting till kids go
28:49to school and then they fail at reading
28:51you can predict reading problems and if
28:54you can predict them you can prevent
28:56them and Prevention's got to be a better
28:58way to go that's what the way all of
28:59medicine is moving towards prevention
29:01don't wait till people have heart
29:03attacks because we're not very good at
29:05fixing things like that or obesity or
29:07alcoholism or schizophrenia let's
29:09predict who's got the problems find out
29:11what those processes are and intervene
29:13to prevent the problems and then it
29:16really is transforming science already
29:18almost all large studies now are
29:20including DNA and it will transform
29:22society parenting as well as education
29:25I'm writing my next book on the genetics
29:27of parenting and then how we understand
29:31ourselves at the end of the book has my
29:34the first world's first polygenic
29:36profile for psychological traits and
29:38it's for me and so I describe what does
29:40it mean that I'm at the 94th percentile
29:42for BMI people body mass index people
29:45say oh then you're just gonna give up
29:46say you're a genetic fatty but I know
29:48it's not the case for all of these
29:50problems it's motivating to say ok I've
29:53got to work harder at it I've got to
29:55change my environment to make it less
29:58easy for me to eat junk food for example
30:01so it's important to know about the DNA
30:04revolution and that's what I'd like to
30:06talk about but I don't have to worry
30:08about getting techy a little bit with
30:09you guys I suppose the first step is to
30:11get DNA if you've done any
30:13direct-to-consumer testing ancestry.com
30:1523andme
30:16you spit in a tube you can get DNA from
30:19any cell in your body because a
30:21remarkable thing you may not have
30:23appreciated before you start life as a
30:24single cell half the DNA from your
30:26mother
30:27have from your father that unique set of
30:29DNA is the same DNA in every cell in
30:31your body trillions of cells so you can
30:34get
30:34DNA from any cell in your body saliva
30:37actually doesn't have cells but it does
30:39have cells that are left off from inside
30:42your mouth that's partly what saliva is
30:43doing so all you need is one cell you
30:46know if you drink from a cup mi5 can get
30:49your DNA because you leave a cell or so
30:51on the lip of the cup so once you get
30:54DNA then you genotype the DNA and
30:57there's as I say only 1% of our DNA
31:00differs the most common type of DNA
31:03difference is called a snip single
31:05nucleotide polymorphism polymorphism is
31:09just a difference and a nucleotide it
31:11refers to the basis of DNA so you know
31:14the DNA code is written in a four-letter
31:16code of a C's T's and G's so what this
31:18shows is your two chromosomes and if you
31:21can see the little letters you'll see
31:22that those two chromosomes have exactly
31:25the same nucleotide bases except for one
31:28and that would be true for us we 99%
31:31similar 1% differs and what we'll do is
31:34so instead of having a C some people
31:37have a T those are called alleles
31:39alternate forms of DNA and once we get
31:42that difference we can just simply
31:44correlate it with traits like academic
31:47achievement
31:48they call it association in genetics but
31:50it's just a correlation between whether
31:53you have one zero one or two C's for
31:56example in that and here's the first one
31:58that was discovered using these new
32:00techniques that I'll describe it was
32:02published in science and it was
32:05discovered using this a theoretical
32:07approach that I'll describe in just a
32:08minute but I just wanted to show you
32:10what a snip looks like really this is a
32:12snip where we humans all used to have
32:14tea tea at this one particular locus
32:17spot on the chromosome then some guy had
32:20a mutation DNA is incredibly reliable in
32:23its replication but when you've got
32:25three billion base pairs of DNA every
32:27time your cells divide you're
32:28duplicating that once in a great while
32:30you get a difference so this guy had an
32:32A instead of a T he could have had a C
32:34or a G but he had an A turns out it
32:37rapidly spread through human populations
32:39especially in Europe word originally
32:41developed because it helps you store
32:45body fat
32:46and back in the Stone Age that was a
32:48very good thing but you can imagine now
32:51in a fast food nation it's not such a
32:53good thing to be efficient at storing
32:55fat and so now 40% of the population has
32:58an a allele and if you have 2a alleles
33:00as I do you hit your 6 pounds heavier on
33:04average than people who have no a
33:06alleles and then people with 1a alleles
33:09in between so that's what we mean by an
33:10association it's literally a correlation
33:12between the 0 1 or 2
33:14a alleles and the trait in this case
33:17body mass index okay now the thing
33:20that's changed everything is a
33:21technological advance called a snip chip
33:24so this is a DNA array the size of a
33:27postage stamp that consists of
33:29synthesized short fragments of DNA
33:32that's around a particular snip and what
33:37the method is the same method we've
33:38always used to detect snips you denature
33:41you raise the temperature of DNA and it
33:43separates and DNA doesn't like to be
33:46separated it wants to find his mate you
33:47chop it up you put a fluorescent tag on
33:50those little fragments of DNA and you
33:52wash them over this plate that has these
33:56probes millions of probes for snips so
34:00if you put this cartoon at the right
34:02hand corner is showing you if you have
34:04the right snip it will hybridize because
34:07DNA wants to hybridize but if you have
34:10the wrong allele there it won't be able
34:12to hybridize so what you're left with
34:14then is fluorescence indicating whether
34:18or not you have that particular allele
34:21and if you have a strong signal you can
34:24see in this array on the lower left some
34:27of the bright dots are brighter than
34:29others if it's dark it means you didn't
34:31have any if it's medium it means you had
34:34one and if it's very bright you had two
34:37of that allele and so you can study
34:40millions of these snips this way and
34:43very cheaply it started out with
34:44thousands of pounds for this and now
34:46it's the real costs are more like 40
34:48pounds to do a whole millions of snips
34:52on this one little chip and it's very
34:54reliable so it's an amazing
34:55technological events that's really
34:57transform the life signs
34:58so now instead of correlating one snip
35:01with the trait you can correlate
35:02millions of snips with the trait and
35:04it's a theoretical you don't just look
35:07at a few candidate genes like serotonin
35:09because you think it's important in
35:10depression which was a dead end
35:12instead you can take an a theoretical
35:14approach just saying let's look at
35:15millions of snips across the genome and
35:17see if any of them are associated with
35:19the trait that's called genome-wide
35:22Association and this is the thing that's
35:26changed my career in the last year there
35:28was a genome-wide Association study
35:30published of educational attainment with
35:32over a million people and the reason
35:34that's important is that they could
35:36detect very tiny differences early on
35:38these studies weren't successful because
35:41they weren't powered to detect small
35:42effects everyone thought heritability is
35:45going to be due to a few genes a big
35:47effect but we never found those instead
35:50what you find in this study and
35:51throughout the life sciences is that
35:53above that line is genome-wide
35:55significance corrected for a million
35:58multiple testings and each dot is a snip
36:01and snip Schlow DH to gether on a
36:02chromosome are correlated with each
36:04other therefore they should also be
36:06correlated with the trait so these Peaks
36:09suggest the most significant results but
36:11the point is thousands of these snips
36:14are significantly associated with a
36:17trait in this case educational
36:19attainment which is merely years of
36:20education and this is what's found
36:22throughout the life sciences there are
36:24no big effects that snip I showed you
36:26for body mass index people thought well
36:28it's just a tiny effect 1% of the
36:30variance turns out it's one of the
36:32bigger effects most of the effects are
36:34very much smaller so that's been a
36:37startling revelation it means you need
36:39huge samples to detect these effects but
36:42how are you going to use them then if
36:43there's so many tiny effects if you want
36:45to study brain beat on gene brain
36:47behavior pathways good luck because
36:49these are such tiny effects it's hard to
36:52see them but I'm interested in
36:54predicting behavior and you can do that
36:55by adding up these snips just like you
36:58do add up your add ups keep items on a
37:00scale and you have to get them in the
37:02right direction and you weight them by
37:04the effect size of their Association
37:07just because
37:08a snip like these account for more than
37:11some of these snips that are less
37:13associated but it's just simply adding
37:15them up and that's a polygenic score and
37:17this is what's transforming everything
37:20because they're 100% reliable they're
37:23unbiased they're cheap but unlike any
37:26other predictor we have you can predict
37:27just as well from birth as you can from
37:30later in life and most of what we know
37:32about prevention in medicine as well as
37:34in psychiatry even think of obesity its
37:37earlier Prevention's that work best but
37:40we can't predict early but we can now
37:42with DNA and it's one of the few
37:45correlations that do imply causation in
37:47the limited sense that there's no
37:49reverse causation you inherit your DNA
37:53sequence and nothing changes that and
37:55before you ask me about epigenetics or
37:56gene expression yeah these snips have to
37:58be expressed but if they're associated
38:00with a trait like school achievement
38:02they were expressed we don't have to
38:04know anything about the pathways in
38:06between and nothing in behavior
38:09environment or the brain changes your
38:11DNA sequence so this is really happening
38:14people worried about this ten years ago
38:16but then direct-to-consumer companies
38:18like 23andme ancestry.com came along and
38:21twenty five million people have voted
38:23with their checkbook to do this to pay
38:24for it themselves mostly what you get
38:27are single gene disorders and most
38:30people do it for ancestry because it is
38:32fascinating you might think you know
38:33your ancestry but you don't we're all
38:35mongrels and we come from diverse parts
38:38of the world and we have a lot of
38:40relatives out there we didn't know about
38:43some great stories and I just see that
38:45BBC documentary last month about Dib
38:48Ling's donor inseminated siblings who
38:51don't find out till they're 18 there's
38:53one father who's inseminated 65 kids so
38:57they're actually half siblings who
38:58didn't know about each other so it's a
39:01Wild West out there and
39:04what they're not doing yet is these
39:06polygenic scores but that's the big
39:08thing now they're all struggling to do
39:09it a new company came up this week
39:11that's trying to sell polygenic scores
39:1423andme won't do it because they were
39:16burned by FDA for reasons we can go into
39:18so they allow if you do 23andme with one
39:20push of a button you download all your
39:22genotypes and these other companies with
39:25one push of a button you upload them and
39:27then they give you these polygenic
39:29scores and so the big news is it's not
39:32announced yet but it's really cool that
39:33the NHS the government has this given 80
39:36million pounds to make genotyping free
39:39on the NHS so when you go to the
39:41hospital beginning next year and they
39:44take blood you'll be asked do you want
39:45to do this and how much information do
39:48you want and when they've done this in
39:50Finland and Estonia they're
39:52oversubscribed right away 85% of the
39:54people want to do it because in a way
39:56it's so much better than
39:57direct-to-consumer companies in terms of
39:59data confidentiality for example they're
40:02not going to sell it on to Pharma as
40:0423andme does anonymously and also they
40:08won't if you do 23andme and you're in
40:10the unlucky 1% of the population who has
40:14two alleles for this recessive trait for
40:17Alzheimer's you could find out you're at
40:18a 60% risk for having Alzheimer's and
40:23and what do you get from them a link
40:25saying you might want to find out more
40:27about Alzheimer's but there's nothing
40:28you can do about it so if you had it
40:31with the NHS and you have nice the
40:33National Institute for clinical
40:34excellence they could decide you can't
40:36do everything for everybody there's some
40:37things you can't do anything about
40:38so probably the standard procedure will
40:41be do you want to just know the stuff
40:43that NHS thinks you ought to know about
40:46certainly heart attacks because you can
40:48predict those from early in life and you
40:50can prevent heart attacks alcoholism you
40:54know it's not that good apologetics core
40:55but you cannot become alcoholic unless
40:57you drink a lot of alcohol for a long
40:59time if you drink as much as your
41:01friends they might not be at risk for
41:03alcoholism but you might so there are
41:05things like that that you might really
41:06want to know about and this is 160
41:08million or behind this if it works well
41:10with the first 5 million people so the
41:13only thing is you have to agree to make
41:14your NHS electronic records available
41:17for
41:17anonymized because the idea is to get
41:20bigger and bigger samples to detect
41:22smaller and smaller effects so here are
41:24the big one big ones in psychiatry a lot
41:29of work has gone into severe mental
41:30disorders this is bipolar and major
41:34depressive and depression and with
41:36schizophrenia you can explain seven
41:38predicts seven percent of the liability
41:40toward schizophrenia now it's maybe 50
41:42percent heritable so it's a long way to
41:44go but this is all just in the last few
41:46years once we recognize that the biggest
41:48effects are very small but the star and
41:51the thing that's really been amazing to
41:53me is that we use the educational
41:57attainment genome-wide Association study
41:59to create polygenic scores for kids in
42:02my ted's project the twins project so
42:047,000 kids we create their score for
42:07educational attainment which means years
42:08of schooling completed well they're 16
42:11so they haven't completed school right
42:12but we I was interested in saying how
42:15much of the variance will we predict in
42:16tested school achievement and the answer
42:19is we predict 15% of the variance in
42:21school achievement more than you predict
42:23for the target rate of educational
42:25attainment in adults which is 10% and
42:27this is the strongest prediction yet in
42:30the behavioral sciences explaining 15%
42:33of the variance school achievement 60%
42:36heritable so we've got a long way to go
42:38I have no doubt in a few years we'll be
42:39at 30% it's a technical problem to get
42:42to 60% we're going to need whole genome
42:44sequencing because the snips we use are
42:46great we've got millions of them but
42:48there's a lot of the genome
42:50we're not tagging with them so that's
42:52the next big thing whole genome
42:53sequencing but right now 15 percent of
42:56the variance is a lot of variance to
42:58explain you think of Ofsted ratings of
43:00school quality explaining four percent
43:02of the variance this is better
43:03prediction than you can make from income
43:05and that you know SES sort of variables
43:08to the family and I just wanted to show
43:11you this a bit more detail if you take
43:13this correlation about 0.4 for the
43:16sample now we take the sample now and
43:19divide the 7,000 kids into equal groups
43:21of a hundred deciles
43:22and you can see that there's a linear
43:24relationship the y-axis is GCSE scores
43:28you know the lower the higher the
43:30polygenic score the higher the GCSE
43:32scores and but there's a big difference
43:34at the extremes even though we're only
43:36explaining 15% of the variance the
43:38average grade at the bottom decile is a
43:40c' and the average grade of the upper
43:42decile is an a-minus but it's only 15%
43:45of the variance so an important point to
43:47emphasize here is I don't know if you
43:49can see the yellow dots this is a
43:50boxplot so it means the kids in the
43:52lowest decile 75% have a grade of C or
43:55lower but 25% have a higher grade and
43:58some of them have grades in the A's it's
44:00not a perfect prediction conversely the
44:03kids in the highest group 75% have an
44:05a-minus or greater but 25% have a lower
44:08grade but that's what will always be the
44:10case because heritability isn't 100% so
44:13we'll never explain it all and a
44:15real-world significance even now is 25
44:17percent of the kids in the Loess group
44:18go to university and 75 percent of those
44:20in the top group go to university
44:22so this is a real difference and we need
44:25to think about how we're going to deal
44:27with this in education and in parenting
44:29so right now all we've got is this
44:30course variable of educational
44:33attainment but there's a lot of work
44:35being done on more specific polygenic
44:37scores for reading for STEM subjects and
44:39for ADHD and I think it might help with
44:42personalized education where we don't
44:45just have a one-size-fits-all
44:47educational system we recognize that
44:49kids are different and we try to go with
44:51that
44:53and if you can predict problems as in
44:57medicine I think that will lead towards
44:59more preventative work rather than
45:01waiting until problems occur and here's
45:03something people don't recognize is
45:04siblings in a family are 50% similar
45:07genetically that means they're 50%
45:09different and people don't really
45:10realize how different that is so you can
45:14have kids in a family where one kid did
45:16very well at school and the other one is
45:18not doing well but it could be they have
45:20a very different polygenic score they
45:23could you know it could be all over the
45:25place you could have one that's two
45:26standard deviations above the mean and
45:27one that's two standard deviations below
45:28and this is the thing that always comes
45:31up in education is selection and my view
45:33just based on values is we shouldn't
45:34have selection but if you do I don't see
45:37any logical reason why you could argue
45:39against using polygenic scores to sup
45:42and test because at least they're not
45:43biased and you can't get a better score
45:46if you buy an expensive tutor for
45:47example so they add something to the
45:50prediction and I'm particularly
45:51interested in the possibility that they
45:53could be useful in the most socially
45:55disadvantaged families where the kids
45:57don't have the demons felt to deal with
46:01academic you know training so what I've
46:06tried to say is that inherited DNA
46:07differences are the major systemic force
46:09making us who we are making our children
46:13being responsible for how our children
46:15do at school the environment is
46:17important but it's not systemic it's
46:19these random chance variables not
46:21systemic effects of families there's
46:23schools and that polygenic scores will
46:26transform size science it's already
46:28happening all big studies now are
46:30including DNA to add a genetic component
46:33to what they're doing society parenting
46:36schools and also how we understand
46:40ourselves thank you
46:42[Music]
46:43[Applause]
46:57hey thanks for the talk
47:00um so on the genome-wide Association
47:02studies you mentioned that these are
47:04correlational but that generally implies
47:06causation and it's true there's no
47:08reverse causation here but don't things
47:10like a sort of that mating increase
47:12those bias those estimates upward and I
47:14guess the extreme case would be you
47:16could imagine June genome-wide
47:17Association study that would test for
47:19country of birth right like it is
47:21impossible the genetics would have yeah
47:24yeah well this is you know like
47:26condensed from three hours of talk and
47:28it's discussed in the book that this
47:30very good point that as I said at the
47:32beginning we only describe particular
47:35populations at particular times and the
47:37populations that are in these big
47:39genome-wide Association studies are
47:40prude almost entirely northern European
47:44American Australian their Caucasian
47:46populations so these scores describe
47:49that population pretty well so it
47:50describes people in England but it
47:53doesn't describe other ancestry groups
47:55for example and the way you deal with
47:58this is you take out principal
48:01components from these analyses so these
48:02these are the you know this is all the
48:04life sciences doing the stuff and some
48:05of the brightest people around so that's
48:07well covered the issue of a sort of
48:09mating is a tricky one because we're if
48:13you're interested in predicting how well
48:14kids do at school part of their genetics
48:17is whether or not there's a sort of
48:19mating you know like begets like
48:21opposites attract it's only like begets
48:23like it's a sorted at positive assorted
48:25mating so you know if you don't believe
48:28in intelligence be single go to a
48:31singles bar and in a few minutes the
48:34first thing you're picking up is how
48:35bright someone is especially verbal
48:37intelligence so for personality the
48:39correlation is point one between spouses
48:41for nonverbal ability its point four and
48:46for verbal ability is 0.6 I mean you
48:49don't really know somebody's spatial
48:51ability very well when you talk to them
48:52but you very quickly pick up on whether
48:54they're going to be worth talking to in
48:55the morning so that's a sort of de
48:57mating and it's part of the genetics
49:00it's getting at mechanism if you want to
49:03understand the mechanisms by which this
49:04occurs great but if you're interested in
49:07prediction
49:08it's okay assorted of mating and and it
49:11really is only for cognitive abilities
49:13that you get this assorted mating it's a
49:15great question and just you know tons of
49:19stuff about this and so it'd be nice to
49:20be able to go on about it some more but
49:22I think we'll have to leave it there for
49:24this thank you
49:28so in the study about the Bolsheviks
49:31cause and essentially the effect sizes
49:33if you have a million people and you
49:35have a lot of snakes how can you like
49:37detect the effects and like interactions
49:40between different sized yeah great
49:44question I'll repeat it because that mic
49:46is only for the recording there's a
49:48question about if you have these small
49:49effects how do you correct for looking
49:51at millions of snips and the second part
49:54was interactive effects so again a lot
50:04to say about that that's a very
50:05pertinent question but as I showed you
50:09we're correcting for a million tests so
50:11that line of significance corrects for a
50:13million tests even though you have more
50:15than a million snips
50:16those are linkage groups so it's thought
50:19that a million correcting for a million
50:20tests work and it does because what you
50:22do is you create these polygenic scores
50:24from the genome-wide Association data in
50:26one study but then you apply them in
50:29independent samples and what people
50:31realize right away is if you take a
50:32hundred snips doesn't predict nearly as
50:34well as a thousand doesn't predict
50:36nearly as well in independent samples as
50:3810,000 but as you rightly are intuiting
50:41these are additive effects of each snip
50:44and most people think genetic effects
50:47must interact well they do at a
50:49mechanistic level but fortunately for us
50:51they don't when it comes to adding up
50:54the effects of genes you get most of the
50:55effect of heritability from adding up
50:58these genes and in a way that's the way
51:00selection works in evolution it works on
51:02additive genetic variants but you know
51:04it's hard to believe but quantitative
51:06genetic studies twin and adoption
51:08studies and animal selection studies are
51:10all consistent with the notion that
51:12added that genetic effects are largely
51:14additive
51:17because the problems of power would be
51:18so much greater if you have to deal with
51:20not just interactions between two genes
51:22but what about ten genes or a hundred
51:24genes
51:29[Music]
51:31thank you thank you for the thought I'm
51:35a physicist by training science it's
51:37somewhat denser I try to always to look
51:39somewhat more fundamental perspective
51:41say if I understand correctly DNA
51:44determines what proteins are produced by
51:47the body which means that if we inject
51:50those proteins which made people include
51:53smarter then that that study potentially
51:57you know this study could be potentially
51:59interpreted like you know today you can
52:02do a test even in the UK form before the
52:06childís bird is born for the severe
52:09genetic disorders and potentially it
52:11terminate the pregnancy early and one
52:15can imagine that some people want one
52:17sort of kids which have pretty supposed
52:20to be less perform at the school but at
52:23the same time the studies could be
52:25interpreted as you know a way to develop
52:28medicines which
52:32make people's cognitive abilities higher
52:36make sense yeah so you started off by
52:40saying you know about you're talking
52:41about single gene simple gene protein
52:44sort of relationships and it's all a lot
52:47more complicated all these genes do many
52:50many different things and each thing
52:53even in the brain is influenced by many
52:55many genes so it's going to be hard to
52:58take like a gene editing approach for
53:00these complex disorders and traits
53:01influenced by thousands of DNA
53:03differences where people are really
53:05thinking about it those in terms of
53:07single gene disorders that's where gene
53:08editing comes in but you know you have
53:11there was this crazy BBC documentary on
53:13beauty and how we could do gene editing
53:14for beauty but you see you've got
53:17trillions of cells how are you gonna
53:18change all of your cells as an adult all
53:21you can do with gene editing is get in
53:22there at the first stages but you'd have
53:24to change every cell but people are
53:27trying to do that you know for single
53:29gene disorders with gene editing this
53:31CRISPR technology which is amazing but I
53:33don't think it's going to work when
53:35you're dealing with these thousands of
53:37tiny DNA differences but another good
53:41question
53:52last one we can finish here so thanks a
53:57lot again my pleasure thank you all
54:00[Applause]
54:12you

About this transcript

This page contains the full transcript of How DNA Makes Us Who We Are | Robert Plomin | Talks at Google by Talks at Google, generated from the public captions YouTube serves with the video. The transcript has 9,453 words across 1,399 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.