26 February 2013

Personality Isn't Random

All parents are highly attuned to the ways that their children are distinctive and different from each other.  We notice that one child is a bit better at learning a foreign language, while the other is a bit better at sports.  We notice that one child is a bit more of a morning person, while the other is a bit more prone to stay up late at night.  We notice that one child is more aware of what adults in a room are talking about, while the other tends to be relatively oblivious.

But, particularly if a parent doesn't regularly interact with large and diverse groups of children the same age the way that a school teacher or youth group leader or sports coach does, that it is far harder to notice the extent to which your own children are similar compared to children in general. 

But, the psychological make up of children is not random.  A large component of a child's personality in inherited from that child's parents, and siblings tend to have far more in common with each other than two people their age in the general population.

This week, for fun, I had both of my middle school aged children take a Myers-Briggs personality test, which strikingly proved that point, as I note below the break, after explaining the what the test is and what it purports to measure.  After that I discuss a bit about what we do know about the role of inheritance in personality, and a few other bits of anecdotal evidence that I've acquired as a parent.

Background: Myers-Briggs Personality Types

The Myers-Briggs type indicator is a paper and pencil self-questionnaire originally designed in 1962 by Myers and Briggs based on concepts articulated by Carl Gustav Jung in a book published in 1921 is designed to measure four dimensions of personality.  For two of these dimensions, one variant that is about three times as common as the other variant.  For the other two of these dimensions, both variants are equally common.  The four personality dimensions in the Myers-Briggs personality typing system  are: 

* extraversion (75% of the population) v introversion (25% of the population)
* sensing (75% of the population) v. intuition (25% of the population)
* thinking (50% of the population) v. feeling (50% of the population)
* judging (50% of the population) v. perceiving (50% of the population).

Crudely speaking:
* extraverts tend to be social, while introverts tend to be loners;
* sensing people are practical and concrete, while intuition guided people are theoretical and abstract;
* thinking people make decisions rationally, while feeling people let their emotions guide their decisions;
* judging people clearly segregate work and play and focus on meeting goals, while perceiving people mix work and play and focus on the process of reaching goals.

These four dimensions, in combination, permit sixteen different personality types, some more common than others, and the combinations play out in distinctive ways.

The current standard in personality psychometric testing is now the "Big Five" personality dimension model, rather than Myers-Briggs, although there is some overlap (both have an extraversion-intraversion dimension, and some of the other four of the Big Five are correlated strongly with Myers-Briggs dimensions), but Myers-Briggs is still widely used and understood, and it is an easy test to conduct with easily available resources.

While the Big Five system is the most widely accepted by personality psychologists, a lack of a consensus set of definitions of personality traits is one of the central issues in the field that leaves it struggling to establish itself as more than pop psychology and pseudoscience.  There are many competing systems to define personality traits.

One of the more sophisticated critiques of personality testing, for example, argues that personality tests are often designed so that personality traits appear to be equally common in both me and women, even though the particular questions that men with a particular personality trait answer that cause them to be associated with that trait often differ greatly from those that women answer that cause them to receive the same composite score for a particularly personality trait. Contrary to the usual conclusion that personality types are only slightly gendered, this critique found that there was only a 10% overlap between male and female personality test responses that was masked by the aggregation of subscores and blending of question types on a single personality test (i.e. you could determine someone's gender with 90% accuracy from their answers to a personality test).

The Results

The test results showed that both of my children had precisely the same one of the sixteen Myers-Briggs personality types.  I know the Myers-Briggs personality types of myself and their mother as well.  She and I are different on three of the four dimensions and identical one of the four dimensions.  The trait that she and I share is found in only 25% of the general population.

The trait that their mother and I share is also shared by both of our children. With regard to the other trait that has a 75%-25% ratio in the general population, for which their mother and I differ, both children have the trait that is more common in the general population.  With regard to the other two traits (thinking v. feeling and judging v. perceiving), both children match my type with regard to one of the traits, and both children match their mother's type with regard to the other.

It is also worth noting that the Denver Public Schools conducts IQ tests for all of its elementary school students (Raven's Progressive Matrices) in order to identify if they would benefit from gifted and talented programs and informed us of the results.  My two children are less than two percentiles different from each other in tested IQ.  Both of them are within the IQ range of myself and their mother as estimated based on upon our SAT and GRE standardized college entrance exam scores.

What Are The Odds?

These results would be stunningly rare if Myers-Briggs personality types were due to random chance without any hereditary component.  The odds that two people drawn at random from the general population would have the particular personality type that my two children share is one in 455.  The odds that all four of us would also be identical with regard to one of the less common version of one of the personality dimensions, would also be very low.

But, they would be quite typical results, if Myers-Briggs personality types arose from a simple Mendelian inheritance model.

The overall percentages in the general population, and the mix of these traits among the four of us, is precisely what one would expect if extraversion v. intraversion and sensing v. intuition were governed by two equally frequent alleles in the general population with alleles for extraversion and sensing being dominant genes, while intraversion and intuition were recessive genes.

There are four possible combinations of these traits (TJ, TP, FJ and FP).  Half of those combinations involve one trait from each parent.  Half of those combinations are a match for one parent or the other.  In a simple genetic probability model where there was an equal chance of taking after either parents in regard to one of these traits, the odds of the both children matching each other is one in sixteen, and the odds of a shared combination having one trait matching one parent and the other trait matching another parent is one in two, for a combined probability of one in thirty-two (about 3%).

This is uncommon, but not particularly uncommon.  The most frequent possible outcome (that the children do not match each other on these two Myers-Briggs personality dimensions and that one is an exact match to one parent, while the other is not), would have a probability of 15/64 (about 24%).  So, relative to the most likely outcome, the odds of their particular combination are 15-2.

If intelligence were distributed randomly, the odds that my children would be so similar in IQ would be about one in 2500.  In contrast, their similarity in IQ is perfectly consistent with a model in which IQ is highly hereditary (as it is for children who are not in poverty) and based upon many genes (so that the hereditary components varies gradually and is subject strongly to the law of averages, rather than by a few genes that produce big jumps in IQ between people who have it and those who don't).

Personality traits and IQ are significantly hereditary.

In reality, a significant amount of research has been done to determine the extent to which personality traits like those measured by the Myers-Briggs test are inherited, by methods such as conduct personality testing people who are related to each other by blood, people who are unrelated to each other, and identical v. fraternal twins. 

We know from psychological research such as the Terman study that followed a cohort of high IQ school children in California throughout their lives, that personality types are largely independent of intelligence (at least among people of high intelligence).  In the case of the Big Five personality dimensions, four recent twin studies found that about 47%-57% of the variation in that personality dimension is inherited, while the remainder is not.  But, some personality traits do correlate strongly with a number of DSM-IV psychiatric disorders.

Adult IQ is about 85% heritable according to a 2004 meta-analysis, although factors such as differences in the pace at which children develop their IQ potential makes child IQ somewhat less heritable (about 45%).  The heritability rate is lower in poor families.  There is strong evidence that the deprivations experienced by children in poverty suppress adult IQ.  Maternal folic acid deficiencies and neonatal or early childhood exposures to lead and to air pollution, for example, have negative impacts on IQ. 

The heritability estimates for the Big Five personality dimensions almost certainly understate the extent to which some personality subtraits are inherited.  Almost all comprehensive personality and IQ tests that profile a person with scores on a small number of dimensions aggregate scores from questions on a number of subtraits.  In reality, some subtraits may have greater heritability than others, or for example, may test vaguely similar but actually gender specific traits that are independent of each other genetically.  As long as some of the subtraits are not very heritable, the heritability of the overall personality trait that is aggregated from the subscores will be lower than a personality trait more carefully defined to match phenotypes empirically known to be heritable.

The empirically determined heritability of Big Five personality dimensions and many other personality traits is also probably reduced by testing errors in what are usually self-reported paper and pencil testing instruments asking people to describe themselves (something that is often inherently subjective),rather than third party evaluations or tests that are designed to measure behavior in a controlled experimental setting (like the famous marshmallow test) rather than measuring how someone perceives himself or herself.

Improvements in testing instruments to avoid the limits of self-reporting of subjective self-perception,

As researchers continue to identify genes associated with particular behavioral phenotypes, it may be possible to more precisely define the behavioral phenotype associated with particular genes and those precisely defined phenotypes may prove to be far more heritable than the personality traits 

In some cases, such as anti-social behavior in young boys the percentage of the variation that is hereditary is even greater (a total of 72%, of which 61% of the total variation is specific to anti-social behavior and the balance of which is attributable statistically to the hereditary components of IQ and ADHD).  So there is a strong genetic component to personality.  Shared environment plays a minimal role in determining personality types.  Random chance and non-shared environmental factors account for the remainder of the variation. 

The Congenital But Not Genetic Factor Associated With Digit Ratio Distinguished

Even when personality and IQ don't have heritary sources, they frequently have specific, biological, congenital causes.  The cognitive and personality impacts of testosterone exposure in utero plays a role in a many congenital behavioral traits that do not have a genetic basis.  This neonatal exposure is particular well studied because an easy to measure trait, index to ring finger length ratio, can reveal what a person experienced in the womb.

ADHD and the Big Five conscientious personality dimension are also strongly associated with finger digit length ratio   So is high IQ.  Moderate sized studies have also linked sexual orientation and gender identification in both men and women, including subidentifications within a sexual orientatin like the self-identification of lesbian women as "butch" or "femme" to digit ratio.  Digit ratio is also associated with assertiveness, male aggression, masculine appearance, right handedness, autism, ADHD in males, male depression and anxiety, schizophrenia, eating disorder prevalance, female psychopathy, aptitude for sports in women, alcoholism, leadership in management settings, and innovation by managers.

Personality and IQ Are Not Mostly Caused By Mendelian Inheritance Of Common Genes

A handful of genes have been identified that have been associated with character traits that could be described as personality traits or psychiatric conditions, such as resiliency in the face of bullying or abuse, and "novelty seeking."  For example, the CHRNA4 gene has been associated with addictive behavior such as smoking and Internet addictions, AVPR1A (the "ruthlessness gene") and MAOA (the "Warrior gene").  Other studies have looked at 5-HTTLPR (see also here), a serotonin transporter gene, as well as DRD2 and DRD4 which relate to the dopamine system, and SERT to examine their involvement in anti-social behavior and resilience  There are at least 542 rare genes that have been associated with schizophrenia risk, which has a strong hereditary component, although a family history of psychosis is still a better predictor of schizophrenia risk than any genetic test.

But, a significant but not exhaustive effort to link well demonstrated hereditary elements of personality that are known to exist to genes has mostly come up empty handed.  The effort so far has been sufficiently probing, however, that we can experimentally rule out any minimal Mendelian inheritance model for inherited personality traits of the kind that I have suggested above.

This issue isn't particular to personality studies.

 For example, there is a huge gap between the exhaustively studied extent to which high IQ is inheritable and has a genetic component, and the number of genes that have been linked to high IQ, although we do know that the genetic component of IQ involves many genes of small effect rather than a a few common variants that each have a great effect. 

There are, however, a number of well understood, simple, single genetic variants with large effects known to cause mental retardation. Advanced genetic tests can identify genetic causes for mental retardation in about 54% of mental retardation cases where it is conducted and a large percentage of the remaining mental retardation cases can be attributed to known environmental factors such as folic acid deficiencies and lead poisoning.

The so called "missing heritability problem" in which we have established a far larger genetic component to particular kinds of observable forms of gene expression called phenotypes than we can identify particular genes to explain, is pervasive in the study of human genetics. 

1 comment:

andrew said...

Neuroskeptic has a decent post on the problems with relying upon self-reporting in psychiatry, which is a pervasive practice despite being obviously problematic lots of the time.