Cookie Time »« Garter Porn

Genes for IQ Found?

Well, no.

The authors tried to replicate published associations between particular genetic variants (SNPs) and IQ (specifically the g factor). They looked at three datasets, a total of about 10,000 people, and didn’t confirm any of the 12 associations.

As Razib Khan says in his post on this, “My hunch is that these results will be unsatisfying to many people.” I’d go further and say that no-one will be happy with these.

For those who believe that IQ is purely environmental and not genetic, any satisfaction they might feel will be short lived because these authors did replicate the recent finding that genetic variants explain about 50% of the variance in IQ. Looking at all SNPs together, there was a strong correlation between “genetic similarity” and similarity in IQ. That independently confirms what the much-criticized twin studies of IQ said – IQ is about 50% heritable.

But for people who do believe in the genetics of intelligence, this shows us that we have no idea what the genes are, and that everything published so far has been pretty much for naught.

This is consistent with prior studies that have tried to replicate gene-IQ associations. It just doesn’t happen. They turn out to be statistical artifacts, a common problem encountered when doing the number of tests involved in genome association studies.

As for the strong correlation between genetic similarity and similarity in IQ? I’m not remotely disappointed in the finding. It’s exactly what we would expect if those who were genetically similar shared a large portion of their environment as well. They do, in fact, both because families and immigrant groups tend to cluster in similar environments and because a number of things more directly encoded in genes (illness, skin color, etc.) have an effect on the environment in which intelligence is developed.

I’m only disappointed in those who think that this is somehow conclusive evidence of subtle genetic effects on IQ.

Comments

  1. jamessweet says

    So Greg Laden was touting some much lower statistic for the percentage of IQ that was heritable just the other day — IIRC it was in the single digits! I was skeptical, because it seems implausible to me that intelligence would be so uniquely environmentally-driven.

    50% heritable is higher than I would have guessed, but in the ballpark. I frankly have difficulty understanding people who want to make mental characteristics out to be virtually all nature or virtually all nurture*. I mean, every single one of our physical characteristics are clearly affected by a blend of genetics and environment, and since the brain is a physical organ… well, it just seems like an odd thing to argue that it would be uniquely one or the other.

    *To be clear, I am using “nurture” as a shorthand for environment, which includes things like biochemistry in the womb and early life, unpredictable contingent events, etc., in addition to what many people normally think of as childhood environment.

  2. says

    James, you understand that if the uncorrected variability associated with shared genes is 50%, that is the upper bound of how much could be genetic, not how much actually is, yes? You also understand that the brain actually is a uniquely plastic organ, and that humanity’s astoundingly extended childhood appears to exist to maximize that plasticity? Given that, why wouldn’t intelligence (and all the other factors included in IQ testing) be incredibly responsive to the environment?

    And your estimate of how much intelligence should be heritable is based on what?

  3. says

    Also, heritability is a measure of how much of the variation in a trait is due to genetics. If the brain was 100% determined by genes, and everyone had exactly the same level of intelligence, the heritability would be 0.

    It’s just like how the heritability of having two arms in humans is almost zero.

  4. jamessweet says

    First things first:

    And your estimate of how much intelligence should be heritable is based on what?

    Oh, only gut intuition, and obviously any sound data trumps it in a heartbeat, obviously. But there are only so many hours in a day; if a normally trusted source (such as an FTB blogger) makes a claim that “sounds right”, I’m less likely to dig into it than a claim that “sounds wrong” to me.

    James, you understand that if the uncorrected variability associated with shared genes is 50%, that is the upper bound of how much could be genetic, not how much actually is, yes?

    No, I had missed that’s what you are saying. Thanks for the clarification.

    It’s just like how the heritability of having two arms in humans is almost zero.

    Yes, but heritability of arm length and circumference and whatnot is not almost zero. Well, I guess I don’t actually know that to be true, but I would be shocked if it were not the case. To be sure, environment plays a tremendous role in that as well, but you were to tell me there was data showing virtually no genetic role in arm dimensions, that would rather surprise me.

    As to the link, and the stuff that links to, unfortunately I have to go to a meeting in five minutes (that damnable “so many hours in a day” thing again!) so an in-depth reading will have to wait. I’m open to changing my mind, but realize one reason I was so unguarded in the original comment is that I thought I was agreeing with you. I guess I misunderstood!

  5. jamessweet says

    Shit, I’m going to be late. I just want to say that when I said “I thought I was agreeing with you”, what I meant by that is I thought I was agreeing in saying that the picture regarding heritability of intelligence is not nearly so simple as “It’s mostly genetic, and here’s some of the genes involved” nor as simple as “It has virtually no significant genetic component at all.” That’s what I thought I was agreeing with. If you are asserting that it is actually the latter, I am open to finding out more!

  6. Bryan says

    PZ, H squared would not be zero in your example, but undefined…as like there’s no variation to explain.

  7. unbound says

    Another statistical meta-analysis tells us that the correlation is either indeterminate or has a low correlation coefficient. Not really surprised (not really sure why poorly accomplished statistics are so heavily relied upon with some of the sciences).

    Even if there was a decent correlation coefficient associated with the meta-analysis, it would only mean that the theory being examined has some degree of merit…not confirmation of the theory itself. There would still be a ton of work required to make the theory meaningful beyond applicability to large groups of people (so good for marketing something to them perhaps).

    In any case, I do remember my college psychology professor stating something along the lines that IQ was very good at what it measured, but they are not sure exactly what it is measuring. Is that no longer the case?

  8. says

    Jamessweet, I think it is extremely likely that intelligence is mostly environmentally driven. If you consider how humans evolved, it pretty much has to be that way.

    Humans mostly evolved in small tribes, essentially extended families where everyone was related to everyone else. There were a number of bottlenecks where relatively small populations founded larger populations.

    The two archetypal and even defining characteristics of humans are language where complex ideas can be expressed using syntax and grammar, and tool making and tool using. These defining characteristics of humans require the very large brains that humans have. Large brains with substantial plasticity which allow learning across the lifespan. Humans have just about the largest brains per body mass of any mammal. Humans are also the only mammal that experiences high maternal mortality during birth. In the absence of medical c-section a percent or so of women die in childbirth each pregnancy. The very large brain that human infants have at birth often kills their mother through cephalopelvic disproportion.

    Because humans can communicate, a tribe can share cognitive resources. Each member of the tribe doesn’t need to be able to do every cognitive task that the tribe needs, there can be specialization and division of labor. Division of cognitive labor requires that tribe members have different neural substrates to do those different cognitive tasks.

    There is no doubt that the more and greater “intelligences” that a tribe had at its disposal, the better that tribe would do compared to other tribes with fewer or lesser “intelligences”. Once a tribe can share tools, the tribe only needs one tool-maker for each type of tool. The tribe that had 10 different tool-makers, each making different types of tools that the whole tribe could share (stone tools, wooden tools, cloth tools, food tools, hunting tools, woven tools, fishing tools, agricultural tools, fire tools, communication tools) would do better than a tribe where there was no diversity of expertise. In other words, a tribe with diversity of cognitive abilities would out compete a tribe where everyone had the same cognitive abilities.

    So how does a tribe get diversity in cognitive abilities? If the tribe is all related to each other, it is hard to get diversity in genetics. A gene needs to mutate, have a positive effect, be selected for and become fixed. But in a small group, a gene that codes for a specific type of intelligence doesn’t produce diversity, it only produces individuals with that specific trait. That is not what a tribe needs to be successful.

    What could code for diversity would be making neurodevelopment in utero exquisitely sensitive to environmental effects. Development comprises myriad coupled-non-linear pathways. Development is inherently chaotic and inherently exhibits the butterfly effect (or in this case the flagella effect) where the whoosh of a flagella or an intracellular movement at a certain instant affects the binding of a transcription factor or the retrotransposition site or something else. Since there do seem to be a large number of retrotransposition events (see below), the idea that diversity in cognitive ability is related to SNPs would seem to be rather unlikely, perhaps even quaint.

    Where neurodiversity likely mostly occurs is differentiation and neurodevelopment in the first trimester, when the number of minicolums is determined (about 8 weeks gestation). That is where the relative sizes of the different brain compartments are formed, the basic wiring together, and the ultimate total size. It may be a crap-shoot as to which part of the brain is biggest and best connected, but that doesn’t matter. The tribes needs individuals with all types of cognitive expertise, good vision, good sense of smell, good hearing, skill at throwing rocks, spears, what ever.

    Everything that the brain does is necessarily (to some extent) limited by the quantity of neural substrate devoted to that task. The quantities devoted to individual abilities cannot be arbitrarily increased because all that does is produce a baby’s brain that is too big to be born and the mother dies. Evolution fail.

    Then there is this paper which shows that much of the brain is mosaic, with retrotransposition causing differences in protein expression throughout the brain.

    http://www.ncbi.nlm.nih.gov/pubmed/22037309

    This type of retrotransposition greatly increases the number and behavior of different nerve cells by tens of thousands? Hundreds of thousands? Millions? What relevance does the single parent genome have to brain function if the brain has 100,000 different daughter genomes produced by retrotransposition?

    And then there is epigenetic programming laid on top of all of this.

  9. JL says

    It’s exactly what we would expect if those who were genetically similar shared a large portion of their environment as well. They do, in fact, both because families and immigrant groups tend to cluster in similar environments and because a number of things more directly encoded in genes (illness, skin color, etc.) have an effect on the environment in which intelligence is developed.

    You seem to have misunderstood the study. All the subjects were ethnic Swedes, and none of them were closely related to each other (those who were found to be related in a genetic test were excluded from the study). So, what the study shows (like the previous Davies et al. study whose results it replicated) is that the more genetically similar overall two unrelated persons from the same population are, the more similar their IQs are. This strongly suggests that IQ is influenced by at least hundreds of different common genetic variants. They estimated that the heritability of IQ in the sample was 47%. When the classical twin method was used, the larger sample from which the study’s sample was drawn yielded a heritability estimate of 57%. The 10 percentage point difference suggests that some of the heritability is non-additive and/or that the new method would require a more extensive sampling of markers for more accurate results.

    James, you understand that if the uncorrected variability associated with shared genes is 50%, that is the upper bound of how much could be genetic, not how much actually is, yes?

    The method in question actually yields the lower bound of additive heritability, not the upper bound of broad heritability.

    is a uniquely plastic organ, and that humanity’s astoundingly extended childhood appears to exist to maximize that plasticity? Given that, why wouldn’t intelligence (and all the other factors included in IQ testing) be incredibly responsive to the environment?

    While IQ is highly malleable in principle, in practise it is one of the more stable human characteristics across the lifetime.

    This study makes the mistake of assuming IQ tests actually measure anything useful…

    Actually, you’re making the mistake of assuming that they don’t.

  10. says

    Actually, JL, I don’t think you fully understand the general research behind this. You’ve seen those big, pretty charts where you’re shown how Germans cluster over here and sub-Saharan Africans cluster over there and Japanese cluster somewhere else when you look at SMPs? If you look at more fully sampled versions of these, you’ll see that groups that live closer to each other blend into each other. Particularly in places without much migration, people who share more of their genomes live more closely together–even if they aren’t immediate family.

    People who live close together share a large number of things unrelated to their genes. Genes don’t select for the quality of local schools or health care or nutrition. They don’t select for industries or environmental conditions and contaminants. They don’t select for cultural practices and values. Where you live, however, does affect all that–and it correlates with genetic relatedness.

    That’s where you get into heritability estimates. Heritability estimates are only as good as what you’ve subtracted from them. You can say that something covaries with relatedness, but until you control for the other factors that also covary with relatedness for other reasons, the most you can do is say, “Here’s the covariance our study hasn’t accounted for.” That is heritability, but it’s only an upper bound for genetic contribution.

    As for stability of IQ scores, you’re confusing that with correlation. Individual test scores can change quite a bit and still produce a correlation between earlier and later test scores of, say, .7. Beyond that, there’s actually very little data on the stability of IQ scores for any substantial length of time and none I know of that covers people from very early childhood through a career. Also, any sort of longitudinal test selects for stability of environment, cutting out a potential source of variability, which again makes results upper bounds rather than the most reasonable estimate.

  11. Bryan says

    Stephanie, do you just make stuff up when in over your head?

    You said:

    “As for stability of IQ scores, you’re confusing that with correlation.”

    How else do you calculate reliability?

    “Individual test scores can change quite a bit and still produce a correlation between earlier and later test scores of, say, .7.”

    You realize that’s variance explained / this is social science / this translates to a simple correlation of .84?

    “Beyond that, there’s actually very little data on the stability of IQ scores for any substantial length of time and none I know of that covers people from very early childhood through a career.”

    How about this study? It’s only been cited 200+ times and covers 66 years longitudinally.

    http://www.sciencedirect.com/science/article/pii/S0160289699000318

    (and see table 1 where they cite a dozen other such studies).

  12. says

    It’s only been cited 200+ times and covers 66 years longitudinally.

    Bingo. Not saying anything specific about that study, but this is a great, frame-able example of the institutional dishonesty that Brian Pesta and his Muddled Mafia of psychometric researchers rely on to keep their jobs. Take a handful of studies that say what they need said, and no matter how many times they are shown to be inadequate (and it only takes once or twice) site it as often as possible in papers uncritically published in the journal “Intelligence” so that later you can say how many times it is cited.

    Bryan, citing yet another paper from “Intelligence” in an argument for why you are right and everybody else is wrong is the exact behavior that Huxley was doing just now with his lower lip stuck out and yelling unintelligibly that made me decide to check my email and stuff instead of having my first cup of AM coffee. So, the best analogy I can find to your strident comment is a toddler that is demanding something he wants but can not articulate.

    Is it possible for you to take a breath without citing that rag?

  13. JL says

    If you look at more fully sampled versions of these, you’ll see that groups that live closer to each other blend into each other. Particularly in places without much migration, people who share more of their genomes live more closely together–even if they aren’t immediate family.

    There’s so much wrong with that argument that I don’t know where to start.

    For one thing, the fact that people from the same region bunch up together in a PC map does not mean that each person in the same region is more similar to other people in the same region than to people elsewhere. For example, a Swede will often be more genetically similar to a Norwegian than to another Swede who lives next door.

    This paper demonstrates the above. If that’s too heavy-going, I recommend this blog posting which empirically shows that, for example, Frenchmen are frequently genetically more similar to Italians and Greeks than to other Frenchmen.

    Moreover, even if local subpopulations were highly uniform genetically, your argument would not make sense, because it would follow from it that people in the same locality would have to have very similar IQs. In fact, of course, the full range of IQ scores can be found in any local population. Hypothetically, it’s possible that there are highly inbred communities in Sweden where the IQ range is very restricted, but they would have been excluded because individuals who were found to be related were dropped from the study.

    Furthermore, behavior genetic studies of IQ heritability, some of the best of which have been conducted in Sweden, have utterly failed to discover the sort of enormously potent environmental influences you suggest to exist.

    I could go on, but in fact the above points are not even relevant here. The sample in the study consisted of one half of all male twins (i.e. one twin from each pair) born in Sweden between 1936 and 1950. They all took the Swedish armed forces IQ test upon conscription at about age 18. Thus the sample is representative of all regions of Sweden, and there is no local clustering of the sort you suggest.

    As for stability of IQ scores, you’re confusing that with correlation. Individual test scores can change quite a bit and still produce a correlation between earlier and later test scores of, say, .7.

    Nope. The short-term test-retest reliability for the same IQ test is >.9 and >.8 for different tests. If an individual’s test score changes substantially (because of, say, a neurodegenerative disease) the correlation will probably be much lower if not negative, but that has nothing to do with test-retest reliability.

    Beyond that, there’s actually very little data on the stability of IQ scores for any substantial length of time and none I know of that covers people from very early childhood through a career.

    Wrong again. There are several long-term studies of IQ stability. The longest are those by Ian Deary and colleagues in Scotland. They showed that the correlation between IQ scores measured at ages 11 and 80 was .66, or, after correcting for ability range restriction in the follow-up sample, .73. In fact, even that figure is an understatement, because it assumes that the test is perfectly reliable–the true stability coefficient may thus be as high as .80 over 69 years. Needless to say, this level of stability is mind-boggling, rivaled by few if any complex human traits.

    Also, any sort of longitudinal test selects for stability of environment, cutting out a potential source of variability, which again makes results upper bounds rather than the most reasonable estimate.

    Could you elaborate on that? How does a longitudinal study select for stability of environment?

    Heritability estimates from studies using molecular genetic methods as described above are indeed lower bound estimates, because they do not test for non-additive effects nor for all genetic markers.

  14. says

    What did I make up, Bryan? Are you saying that individual test scores can’t vary widely over time in such a data set, when people who tested at an IQ of 100 at 11 scored anywhere from 40 (or 75 discarding the outlier) to 120 at age 80? Are you saying that study was unlikely to select for people with stable environments? Are you suggesting that 11 years old is very early childhood?

    Or are you just making vague sneers about my statements because you need a new hobby?

  15. says

    JL, that’s impressively obtuse. First off, the first paper you pointed me to specifically says:

    In a similar vein, Romualdi et al. (2002) and Serre and Pääbo (2004) have suggested that highly accurate classification of individuals from continuously sampled (and therefore closely related) populations may be impossible. However, those studies lacked the statistical power required to answer that question (see Rosenberg et al. 2005).

    That means it doesn’t support what you’re claiming. But again, a study like that in the original post doesn’t require highly accurate classification to produce a correlation that looks like heritability but isn’t. It only requires a trend toward more closely related people living in similar environments.

    Nor is “In fact, of course, the full range of IQ scores can be found in any local population.” any kind of argument about environmental versus genetic factors, at least for anyone who understands that “environment” is more than just one thing.

    Blah, blah, blah test-retest reliability. You understand that this is a different thing than long-term stability, right? If so, try to have something to say about long-term stability instead, please.

    I love that both you and Bryan try to point me at Deary’s studies. What do you think I was looking at when I was talking about large changes in individual test scores in a group with a relatively high correlation of (later) childhood and late adulthood scores? Have you seen a scatterplot of that data? Do you know what that kind of childhood-adult score changes can be seen in that data? How many individuals–the level at which IQ test scores make a difference in treatment–can change their score by well over a standard deviation of total scores and still fit very comfortably within that data set?

    And do you really not understand how a test that requires finding your subjects decades later and testing them selects for people who are in stable environments; i.e., easy to find?

  16. Bryan says

    Stephanie: Google “stability reliability”.

    btw, stability and heredity are two separate questions. You seem to be confused / blending them.

    Greg. You’d be hard pressed to find a more impressive vita (in social science) than Ian Deary’s, even if you discount his papers in Intelligence.

    Happy holidays!

    B

  17. Bryan says

    Here’s Deary’s vita. Just look at the IN PRESS articles (42 of them, some in Nature).

    I suspect his in press articles *alone would get him full professorship anywhere (my quick search shows 8 articles in Nature t/o his career).

  18. says

    Bryan, do you really not understand that the relatively short-term test-retest reliability for a test of a trait is separate from the stability of the trait itself over a long period of time? Really?

    But do feel free to explain how I’m mixing up stability and heredity. That should be entertaining. You might want to look at JL’s claims that I’m responding to before you try, though. JL is the one claiming that the long-term stability of a trait (not the test-retest reliability) is an indicator that the trait is controlled genetically. I’m not the one making that claim.

  19. Bryan says

    Ok, Stephanie, here’s the litmus test to prove how clueless you are. Email Razib (you quote him above)and ask him to characterize how well you represented his position and thinking in that quote.

    Ask him why he said people might be dissatisfied by the study in question (i.e., is it because the study supports blank slate, and so incompetent naturists will ignorantly dismiss the study, as it contradicts their world view? This seems to be your take…).

    Post reply here!

  20. says

    Bryan, if you can’t tell the difference between me characterizing Razib’s position and me disagreeing with it, as I’ve done here, you’ve got no call to be calling me clueless. You can knock off the name-calling and get to substantial disagreement anytime now–if you’re capable.

  21. JL says

    JL, that’s impressively obtuse.

    Well, it seems that it’s too difficult for you.

    First off, the first paper you pointed me to specifically says:

    “In a similar vein, Romualdi et al. (2002) and Serre and Pääbo (2004) have suggested that highly accurate classification of individuals from continuously sampled (and therefore closely related) populations may be impossible. However, those studies lacked the statistical power required to answer that question (see Rosenberg et al. 2005).”

    That means it doesn’t support what you’re claiming.

    It neither supports nor refutes my argument, because it has nothing to do with it. Why the hell did you quote such an irrelevant passage? The whole paper is about classification algorithms that rely on aggregate properties of populations (e.g. those PCA results you referred to earlier) versus actual genetic similarity between individuals. The new molecular genetic methods of heritability estimation rely on the latter, not on any classification algorithms. Based on classification results, you erred into believing that people who get classified similarly are also genetically highly similar.

    But again, a study like that in the original post doesn’t require highly accurate classification to produce a correlation that looks like heritability but isn’t. It only requires a trend toward more closely related people living in similar environments.

    For your argument to make even a little sense there would have be an extreme degree of genetic similarity between unrelated people living in the same locality, and they would have to have very similar IQs. Genetic similarity explained 47% of IQ variation in the Swedish sample, not some trifling amount that could be accounted for by any weak trends towards genetic and phenotypic similarity. Your argument presupposes that the Swedish population consists of highly genetically distinct local subpopulations each of which has a different and very restricted range of cognitive ability (e.g. everybody who lives in village A has an IQ in the range 80-85, while in village B the range is 125-130, etc.). That’s absurd.

    Moreover, you conveniently ignored (again) the fact that the sampling method of the Swedish study precluded the inclusion of many people from the same locality. The subjects were one half of all male twins born in Sweden between 1936 and 1950. As Sweden is a large country whose population in those days was scattered between about 2,300 municipalities, it’s unlikely that very many people in the sample had even met each other, let alone grown up in the same locality.

    Nor is “In fact, of course, the full range of IQ scores can be found in any local population.” any kind of argument about environmental versus genetic factors, at least for anyone who understands that “environment” is more than just one thing.

    The fact that the full range of mental ability is present in any local school is a deathblow to your argument that genetically more similar Swedes have similar IQs because they come from the same place.

    Blah, blah, blah test-retest reliability. You understand that this is a different thing than long-term stability, right? If so, try to have something to say about long-term stability instead, please.

    What a strange comment! In the very next paragraph, I discussed long-term stability at length, yet you accuse me of not having anything to say about it!

    I love that both you and Bryan try to point me at Deary’s studies. What do you think I was looking at when I was talking about large changes in individual test scores in a group with a relatively high correlation of (later) childhood and late adulthood scores? Have you seen a scatterplot of that data? Do you know what that kind of childhood-adult score changes can be seen in that data? How many individuals–the level at which IQ test scores make a difference in treatment–can change their score by well over a standard deviation of total scores and still fit very comfortably within that data set?

    Of course there are changes in individual scores. The correlation is not unity. The outliers are interesting, and sometimes it’s pretty clear what caused the large score changes. For example, the scatterplot you mention is printed in Deary’s book “Intelligence: A Very Short Introduction”. Deary notes that the individual in the plot with the largest downward deviation from his or her childhood score was shortly after the follow-up study diagnosed with Alzheimer’s. That is the likely reason for that outlier.

    And do you really not understand how a test that requires finding your subjects decades later and testing them selects for people who are in stable environments; i.e., easy to find?

    Oh, that’s what you meant. I don’t disagree with that. The sampling process in the Scottish follow-up studies principally selects for individuals who are, firstly, not dead yet, and, secondly, who are not too infirm to participate (incidentally, both of these outcomes correlate with IQ at age 11 independently of e.g. social class). Yet when you consider that at age 11 intelligence is not yet fully developed, and that by age 80 all people must have experienced some intellectual deterioration, the stability coefficients are truly astounding.

    JL is the one claiming that the long-term stability of a trait (not the test-retest reliability) is an indicator that the trait is controlled genetically. I’m not the one making that claim.

    You are putting words into my mouth. Stability and heritability are of course separate questions, with no necessary connection with each other, and I have never claimed otherwise. (Although, funnily enough, Deary and colleagues have a paper coming out soon in Nature called “Genetic contributions to stability and change in intelligence from childhood to old age”.)

  22. says

    Ender –

    I would just like to note that IQ testing isn’t actually useless, it’s just useless in terms of what a lot of people consider it useful for. The tests used in IQ testing are integrated into occupational psych assessments that help determine how best to help kids who are underperforming in school or who have specific disabilities that interfere with their ability to function normally. I would go as far as to say they are critical elements of such assessments and that such assessments are extremely useful in determining the best course of action in terms of treating mental illness, disability and general academic dysfunction. It isn’t going to make everything perfect, but it does go a long ways towards helping kids reach their best potential.

    That said, IQ tests as they stand alone aren’t used in occupational assessments. They take the various elements of IQ tests and mix them into several parts of those assessments. They aren’t the whole of occupational assessments either, just important parts. This only applies to comprehensive assessments. Typically these run between 6-12 hours, over several sessions.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>