Merenje inteligencije

[betty]

evo esejcine ako nekog zanima. u principu se sastoji iz dve celine. jedna je verovatno zanimljiva samo uskom polju strucnjaka i tice se distribucije inteligencije a druga, dotaknuta na ovoj temi vise puta, se tice odnosa udela nasledja i sredine u razvoju inteligencije. tu je predstavljen model koji je po mom misljenju jedini iole plauzibilan, but only time will tell.

THE g FACTOR AS AN INVERSE MEASURE OF DEPRIVATION

It is a common research assumption that intelligence is normally distributed in the population. Methods used to measure intelligence assume a normal distribution and with this assumption they manage to yield reasonable patterns of results and predictions. The distribution of intelligence is rarely elaborated from the point of view of a specific theory about its origins or development. The aim of this essay is to relate intelligence to effects of both genes and environment within a theoretical framework that reconciles several problems present in the nature-nurture debate, and to elaborate on why these effects would lead to a negatively asymmetric distribution.

 

The facts and the fictionMan is an intelligent beast. Through many successive generations of interacting with the environment, he has come to bend it to his will. It is a common folk notion that man is 'more evolved' than the rest of the animal kingdom, largely due to the fact that his problem solving skills are more pronounced. After all, it is clear that it was not our agility, sharp claws or muscle power that protected us from other predators. Evolutionary pressures have honed us towards highly conspecific-interactive behavior, developing over time a refined symbolic function that further boosted our ability to generalize, memorize, imitate, self-reflect, communicate ideas, teach and be taught. We use our brains much more deftly than any other species. Zooming further into our own kind, it becomes evident that humans differ among themselves in cognitive skills. Some children may quickly pick up knowledge in all their classes while others may struggle with many subjects. Some can't wait to leave school while others spend a lifetime of reading and learning, always enjoying new mental challenges where detecting patterns or discovering novel strategies is required. Very often this difference between people in handling complex information holds regardless of the content of that information: those who have a larger vocabulary are also likely to be better at memorizing general knowledge facts, at calculating, assembling pictures out of blocks, imagining what an object would look like from a different angle and at a host of other mental feats. Intelligence is the term that we use for encompassing these different behavioral measures into a single construct. The idea behind the construct is that there is an ability behind the behavior: the more of this ability people have, the better they are at all sorts of cognitive tasks. In order to better study it, intelligence researchers quantify it by expressing it in terms of variance that different cognitive tests share. They create a fictitious variable the variation of which follows the common variance in the assorted tests, and apply an arbitrary scale to it while basing its units on proportions of this common variance. This variable is known as the g factor, or general intelligence factor. G factor scores, as well as their strong indicator, IQ, are the best across-the-board predictors of performance that differential psychology has ever yielded. If we had to pick a hundred people out of a thousand to do basically any imaginable task, we would do better to pick the hundred most intelligent than to make a random sampling. Intelligence garners respect, but its latent and multifaceted nature and the fact that it spills over into countless areas of behavior introduces difficulty in studying it. One of the many unknowns is what its distribution in the population really looks like.

 

What is normal about intelligence?When running a factor analysis (as we do when we extract the g factor) one of the assumptions is a multinormal distribution of the manifest and latent variables. Parametric statistics need an assumption about the distributions in order to compare real and expected outcomes in terms of probability. It is often safe to assume a normal distribution because many psychological variables follow it: in most domains, average turns out to be most common, and being above or below that is less and less probable as we move away from the mean. Also, as we increase the number of variables under consideration in a factor analysis where we use linear methods of extraction, the probability of getting a (multi)normal distribution increases as a consequence of the central limit theorem. Another reason that it is safe to assume a normal distribution is that the worst that can come from being wrong is that our model will explain less variance. Therefore, assuming a normal distribution of the g factor or IQ scores will give us useful measures with which we can successfully predict intelligent behavior. But none of this means that the distribution of intelligence really is normal.

 

Where am I intelligent?In order to come to an idea about how intelligence is spread across the population we need to have a theory about where it comes from and where to look for it. G is covert, it is a mathematical artifact that we utilize as a decent single-value approximation for a useful construct, and nowhere within the numbers it is represented by will we find the answer to what it really is and how to look at its real distribution. The situation with IQ is no better: we purposefully add and remove items until a normal distribution is reached ? but this is then just the distribution of test scores and not of intelligence per se. Unfortunately, pinpointing the nature and origins of intelligence is the moment where theories begin to diverge. Individual differences are reflected in behavior, but are they reflected somewhere in the makings of our brains and bodies? And even if they are, is it because of our inborn talents, our caring parents, our varied diet, the school we went to, the kids we played with, the air we breathed? There is no consensus on this issue. Many theories about the origins of intelligence center on the nature-nurture debate. Within it is often implicitly embedded a determinist/non-determinist position. Everybody agrees that the extent to which people in our surroundings see us as intelligent is reflected in an idea of our personal worth. The issue dovetails with political values people hold dear regardless of whether or not they are intelligence researchers. If individual differences in intelligence can mostly be linked to our genes then it makes sense to assign different set roles in society fairly early in development. If differences can mostly be linked to environmental causes, then efforts should be turned towards finding high-risk groups and adjusting their environments in such a way that their mental capacities can develop further. But it is not a simple matter to determine the input of genes vs. environment. A way of approaching the issue is to collect data on cognitive tests performed by twins (monozygotic and dizygotic) and fraternal siblings reared together (in the same environment) or apart (in different environments) as well as biologically unrelated children adopted into the same family or reared in different families. The elegance of this idea lies in the fact that we know how much genes each of these pairs share, we know how much variance in their g scores explains their performance, and we can assume that any variance unrelated to genetic commonality should be attributed to the environment, be it shared or unshared among the pair. The proportion of phenotypic variance thus ascribed to genetic variance is termed heritability. The higher the heritability of a trait, the more of the variation in its expression across individuals can be linked to genetic causes. And not only is intelligence a highly heritable trait but its heritability rises as people get older, reflecting a gradual diminishing of the power of environmental effects to explain individual differences.

 

Sifting through intelligent genesUnfortunately, there is something very wrong with the line of reasoning presented above. Genes come into being through an interaction with the environment, as an adaptation to the environment, as components that influence behavior which can again be wiped out by environmental pressures. Genetic influences on behavior are triggered, stifled or nuanced by environmental cues, or they have some bearing on the individual changing the environment to better suit it to his needs. To paraphrase Hess, intelligence is 100% genetic and 100% environmental. There is no aspect of the phenotype that is not the joint outcome of genes and the environment (Cosmides & Tooby, 1997). To link part of the population variation to genetic influences merely means that within this part of the variance we are able to glimpse an existing gene-environment interaction. It tells us nothing about how the interaction plays itself out and it certainly cannot miraculously split the effects of genes and environment into independent components. True, if we make a comparison across traits, higher heritability would mean the gene-environment interaction is more visible and it could make sense to roughly say this corresponds to the variance of one trait being more genetically determined than another. This still does not mean it makes sense to ignore the interaction within any given trait. To muddy the waters even more, size of heritability has nothing to do with how likely we are to find genetic markers of a trait. Matt Ridley (2003) uses the example of variation in number of fingers to illustrate this. Most people have five fingers, so the variance is the extent to which there are people who have more or less than five. A small fraction of the population is born with polydactilia and has six. A much bigger fraction of the population has had accidents in which fingers were severed. Therefore, most of the variance in finger numbers is environmentally determined; therefore its heritability is very low. Obviously it would be wrong to conclude from this that the number of fingers we have has nothing to do with genes. It would be convenient to flip the logic and assume the inverse is true: that highly heritable traits are more environmental in their makeup while less heritable traits are more genetic. But this is an easy way out because it again ignores the gene-environment interaction concealed within the heritability estimate. And this is not the end of the story! If the environment happened to change in such a way that accidents where fingers get crushed or cut off began to decrease, heritability would rise. This should not be taken to mean that heritability tells us nothing. The higher it is, the less we know about how the interaction plays itself out but the lower it is the less of this problem we have. Tooby and Cosmides (1990) have argued that it is safe to assume that low heritability traits are evolutionary adaptations: if a trait is important for survival then environmental pressures will constrain its variance over generations, ultimately leading to a universal expression within a given species. For this reason they maintain that individual differences cannot be considered evolutionary adaptations. Unsurprisingly, billions of dollars invested in searching for genes that make us more or less intelligent have led, to put it bluntly, to nothing. But this presents us with a problem. The more related two people are, the more closely their IQ scores correspond to each other ? and this has been the outcome of thousands of studies all through the past century. For randomly sampled completely unrelated pairs of people there is no correlation. It grows as the degree of shared genes grows going all the way up to 0.8-0.85 for monozygotic twins reared together. This is roughly the size of test-retest reliability: the scores of the same person doing an IQ test at two moments in time will be that close. There must be something genetic about intelligence but we know this because individual differences get carried over from generation to generation. How can differences between people be genetic?

 

Bridging the gapTo summarize, individual differences in intelligence are closely linked to genetic influences yet intelligence doesn't behave like an evolutionary adaptation. Evolutionary pressures lead to genes coding for traits through selecting behaviors important for survival. The more important a trait is for survival, the more clearly it will be linked to a genetic basis. The more clearly we can link genes to it, the less variation between people we should expect. Traits which vary across individuals should not be important for survival and either their expression has little to do with genetic influences or this variation is an epiphenomenon of some other bodily process and the trait itself is not directly fitness-relevant. Yet intelligence is hardly fitness-irrelevant. It evidently varies across species and determines success in many behaviors related to survival: building shelter, hunting, growing plants, communicating needs, etc. Marital partners are often of similar IQ level, implying that it is a trait the expression of which is favorable in one direction. But where are these 'intelligent genes'? Why can't we find them and how could it be that they defy evolution? Is it possible that intelligence is after all a purely social construct?Penke, Dennisen and Miller (2007) have come up with a solution that reconciles all of these issues. It stays within an evolutionary framework, allows intelligence and differences in intelligence to be coded by genes, assumes selection pressures to work towards constraining variance by eliminating genetic combinations that lead to the lower end of the intelligence distribution, and explains why we should not expect to find "intelligence genes" in spite of all this. Their solution assumes that what is reflected in performance on IQ tests is overall functioning of the brain. G is nowhere in particular, there is no "intelligence center", and the way to deduce its value would be to look at phenotypic behavior. As soon as some aspect of the brain is not functioning perfectly performance level goes down. Genes of course code for all the intricate aspects of brain functioning. There are about 30.000 genes in the human genome, coding for approximately 100.000 proteins. By some estimates up to 50% of these genes may be connected to brain functioning. Intelligence, on the genetic level, is dispersed among these 15.000 genes. The mechanism that differences in intelligence relate to is not different functional variants of the same genes ? brain functioning is extremely important for fitness and on each of these loci only one set of nucleotide pairs will be selected for by evolutionary pressures. What can alter the functioning of any of these genes are mutations. In general, if a trait is coded by a small number of genes, one mutation will make a big difference. Since evolutionary pressures have shaped us for so long, mutations will most likely have an adverse effect because the optimal combinations of base pairs within each nucleotide have probably been found already. But if the behavior under consideration is polygenic to such a great extent, each individual mutation will only be mildly harmful. These harmful effects would accumulate because they are carried over from one generation to the next, but selection pressures keep wiping them out at a rate proportional to the harm they cause. The reason they don't cease to influence intelligence is that new mutations keep springing up. It is estimated that in our species 1.67 new mutations arise per individual. A large proportion of these will be in the part of the genome coding for brain functioning. The rate of newly surfaced mutations on the one side will be balanced by selection pressures on the other side, resulting in a state of equilibrium; hence the name mutation-selection balance. Since the aggregate functioning of the brain depends on so many genes, these mutations will not stop it from carrying out its various functions but they will have deleterious effects on overall information processing capacity. The extent to which mutations exert their influence will be visible in all sorts of cognitive tasks independent of the type of information processed. The exact genes in question will not matter much; it is the overall mutation load that will lead to a diminished information processing capacity. Implicit in this is the assumption that not only are the mutations independent of each other, but that their effect on the phenotypic level is also largely additive. This is exactly what Penke et al. find ? that the genetic variance of intelligence is predominantly additive. The effects of this proportion of genes contributing to additive variance in intelligence are nicely illustrated by the Watershed model forwarded by Cannon and Keller (2005; figure 1).Figure 1The Watershed model

The model depicts effects of mutations on traits. In the case of intelligence, we can assume there are many "upstream" genes coding for numerous brain functioning features. In the model they are illustrated by the narrowest streams. The effect of each mutation is small and can be followed upstream to the particular gene, but as soon as any aspect of the mutated trait is needed for some more complex mental feature, i.e. as soon as it is combined with the functioning of some other genes, it becomes less easy to attribute specific effects to concrete genes. If another aspect of the same trait is reliant on some other mutated gene, then the effects trickle downwards together, producing a confluence which is a pure sum of the two effects, but within this sum it is impossible to distinguish the part of the effect that comes from one gene from the part of the effect that comes from the other. The more "downstream" traits are the ones more relevant for survival, and the larger the number of genes they will be coded by. Intelligence, if regarded as overall information processing capacity of the brain, should be very much downstream, and the effects of mutation load on intelligent behavior should be inextricably intertwined with each other. One prediction stemming from this model is that variation in intelligent behavior among humans should be more g-related than variation among members of other species, i.e. that common variance diminishes as we go towards species with less developed brains.

 

Back to normalityAn important implication of regarding mutation-selection balance as the source of genetic variance in phenotypic intelligence is that it allows us to take a closer look at the gene-environment interaction. Admittedly, gene-gene interactions are not taken into account in this model and are a complete unknown even though it is reasonable to assume they exist. It is the additive genes that produce an effect we can predict, and we can further predict how the environment comes into play.If we had only (additive) genetic influences to account for, a normal distribution of intelligence would be the expected outcome because there is no reason to assume anything other than a normal distribution of mutations. Very few people carry few mutations, very few carry a large mutation load, and most of us are somewhere in between. It is important to note here that variance in intelligence is introduced only through deprivation in the sense that a person is deprived of a functional gene coding for brain functioning. The proper gene is either there or it's not, the expected condition is that it's there ? if it's not, we observe an effect on intelligence level. If no mutations existed, no individual differences in intelligence would be found. But there is the remaining 100% of intelligence that is environmental that we still need to account for. Intelligence, being a fitness-relevant trait, endures unidirectional selection pressures. Also, being a fitness-relevant trait, it has a requirement of sufficient environmental conditions to fully develop. Consider what it would be like if highly nuanced environmental conditions produced highly nuanced information processing capacities ? we could expect that variance in intelligence between groups living in highly dissimilar geographical locations would be higher than variance within these groups or that the correspondence of sibling IQs would be less dependent on their shared genes and more on the place they were born and raised ? and all of this would be a result of brains being less capable of handling information if they are not stimulated enough. It should not be overlooked that brains process an abundance of information that is not a part of the construct of intelligence: the main bulk of processing concerns functioning of inner organs and analyzing environmental inputs on sensory organs. It is only a small part of this overall processing capacity that we tap into when measuring IQ and most of it is so clearly related to survival that to have it widely fluctuate depending on environment would make as much sense as to allow kidney functioning to fluctuate for the same reasons. This could be the main reason that IQ is strongly connected to overall health. Sufficient conditions, on the other hand, are a common way that fitness-relevant traits and environment interact. When Hubel and Wiesel covered the eyes of newborn kittens with a patch the result was that capacity for vision deteriorated and hadn't returned even after the patch was removed. But paying special attention to visually stimulating a normal kitten will not endow it with superior vision. If the mutation-selection balance model of intelligence holds, we also shouldn't expect environmental inputs to influence the distribution on both ends. Environment should not give us any special brain capacity which is over and above that already written in the genes, and sufficiently met environmental conditions should allow brain processing capacity to fully blossom. If no mutations and no environmental deprivation existed, again no individual differences in intelligence would be found. This leads to viewing the g factor in a new light: as an inverse measure of total deprivation. Basically its value indicates how lucky we have been in terms of escaping mutation load and having been presented with sufficiently met environmental conditions for growing up in. Coming back to the distribution of intelligence in the population, we have a starting point in the normal distribution of mutations coding for brain functioning and an overlying effect of environmental conditions which can be detrimental to processing capacity but can't especially boost it. Going from the mean given by mutation load only, it is easier to become less intelligent than more intelligent - leading to an asymmetric distribution with a thicker tail on the low end (figure 2). Figure 2Distribution of intelligence if g is an inverse measure of total deprivationJust like the mutations, the necessary environmental conditions can be of a wide variety. Not having a varied enough diet or ample food intake is detrimental for the development of the whole body, brain included. Not stimulating the brain enough with problem-solving activity in an early period, not communicating enough so that the symbolic function develops late, and even not engaging in motor activity during critical periods could all alter brain functioning for the worse.

 

PredictionsIt is impossible to directly check the shape of the distribution of intelligence not only because we know it as a latent variable but also because it likely is not located in some tangible (thus quantifiable) neurological substrate. It is a functional product and we will probably never be able to directly measure its quantity and completely bypass having the person perform cognitive tasks. Perhaps someday it will be possible to assess information processing capacity without actually processing information but that time is not in the foreseeable future. Nevertheless, predictions can be made based on the assumption of a negatively asymmetric intelligence distribution. This essay will address two such assumptions, one related to ability differentiation and the other to the Flynn effect.

Ability differentiation

Ability differentiation, formerly known as Spearman's law of diminishing returns, concerns the fact that the g factor extracted from lower-scoring participants explains more of their common variance than a g factor extracted from higher-scoring participants. Existing explanations rest on the idea that g is something one can have more or less of, and when there is more, it can be allocated to different specific abilities in a greater number of ways. It is possible to be intelligent in a variety of manifestations but being dull is unexcitingly predictable. Mutation-balancing selection of intelligence would predict quite the opposite, yet ability differentiation could be a straightforward consequence of the proposed distribution of intelligence. On the lower end of the scale, an increment of intelligence comes with a more uniform spread than a corresponding increment on the higher end (figure 3).Figure 3Ability differentiation in intelligenceOn the higher end, mutation-selection balance theory predicts our brains are more uniform in information processing capacity because we are closer to the evolutionary limit. Variance is introduced from the lower end, and more of it can be assigned to g there because g is a measure of common variance. To make a specific prediction, the IQ correlations of monozygotic twin pairs should be higher for pairs below the mode than for pairs above the mode.

The Flynn effect

The Flynn effect concerns the fact that mean levels of IQ have been rising in the last fifty years at a rate of about 10 points per generation. As this clearly can't be an effect of genes changing, and Dickens and Flynn (2001) believe it is not a pure effect of environment either, they introduce the idea that environmental and genetic effects interact in a multiplicative way so that small inborn differences get greatly pronounced with just a small change in the environment over the years. The mutation-selection balance account of intelligence would offer a somewhat different explanation. Although such gene-environment interactions are conceivable they probably don't account for the best part of intelligence variation. Instead, the Flynn effect springs from larger numbers of people having sufficient environmental conditions to grow up in. This releases the deprivation due to environmental circumstances. A small change in the environment leads to a large effect in the phenotype because as soon as sufficient conditions are met, ability fully develops (constrained only by mutation load). This scenario comes with three specific predictions: a) The Flynn effect is mainly due to changes in the lower end of the intelligence distribution because that is where environmental influence was introduced in the first place. If a shift does exist along the whole distribution, it should get progressively smaller for higher intelligence levels. b) The Flynn effect has a clear end when sufficient environmental conditions are met. Continuing to introduce change in the environment cannot make us superintelligent and we should not expect the most intelligent people of today ? or of the near future ? to be especially cognitively skilled compared to the most intelligent people from a century ago. Sunder, Barlaug and Torjussen (2004) believe there is evidence that meets both of these predictions, at least in present day Norway. Also, the connection between GDP and national IQ points in this direction (figure 4).Figure 4National IQ by Gross Domestic Product

It should be noted that it is unwise to put too much belief in national IQ data as the low end is underestimated and the high end overestimated due to different sources of bias. Even so, the pattern is clearly exponential and if we look at average IQ level of a hundred we see that it is not at all informative about GDP value ? it ranges from nearly minimum GDP to maximum GDP in the sample! If mutation-selection balance rests on solid assumptions then this same pattern should be observed within any one culture when relating IQ to wealth. c) Lowering the amount of adverse environmental effects leads to normalizing of the intelligence distribution. This can't be directly measured but since factor analysis relies on a normal distribution of the latent variables then the closer the real distribution is to normal, the more discriminative the g factor will be. It will explain more variance in societies where there is less poverty and hunger, more schooling, etc. It is indeed true that the g factor explains more variance in intelligence tests in Westerns societies than in African countries; it would be interesting to see whether it explains more variance in the Netherlands today than it did fifty years ago.

 

ConclusionTheories that attempt to relate proportions of variance in intelligence to genes or the environment suffer from the problem of relying on a fuzzy heritability estimate which always encompasses gene-environment interactions. These problems are circumvented by assuming that variance in intelligence depends on a balance of selection pressures and mutation load within genes coding for overall brain functioning. G, in this framework, is an inverse measure of total genetic and environmental deprivation on brain processing capacity. Mutation-balancing selection theory comes with no prediction about the distribution of intelligence across the population, but a negatively asymmetric one can be inferred. This distribution leads to specific consequences: that the g factor will explain more variance in the lower than in the higher end of the IQ scale, that the Flynn effect will be temporary and restricted to the lower end of the distribution and that the g factor will explain more variance in intelligence in those cultures where the effect has ended.

[BraveMargot]

betty, najzad sam stigla da procitam tvoj esej o inteligenciji.poprilicno mi se svideo, dosta je pregledan i jasan i ne zahteva neko podrobno predznanje, niti preterano citiras radove, tako da bi mogao uz male izmene pravo u neki naucno-popularni casopis. imam samo jednu primedbu: iznosis tezu da se inteligencija zasniva na nekom uopstenom neuroloskom supstratu, tj razvijenosti mozga kao takvog koja se kodira u skupu od 15000 gena, ako sam dobro shvatila. Medjutim, postoji veliki broj radova koji povezuje g faktor (posebno njegov fluidan deo) s testovima prefrontalne funkcije, posto su oni najvise zasiceni tim faktorom. Sta mislis o tome, vidim nisi se u radu osvrtala na te nalaze? isto, deluje mi malo kontradiktorno sto prvo navodis da u donjem delu distribucije postoji manje varijabilnosti, i da g faktor bolje objasnjava razlike u tom delu distribucije. S obzirom da se iz grafikona vidi da je gornji deo strm i da u njemu vlada veca varijabilnost, onda nije ni cudo da se donji deo moze uspesnije modelovati nekim takvim testoloskim konstruktom koji je verovatno samim izvodjenjem bio pod vecim uticajem tog sporo uspinjuceg 'glupljeg' dela distribucije. dakle, to mi pre lici na artifakt merenja - testovi ustanove varijacije, koje onda objasnjavamo faktorom izvedenim iz istih tih testova. Takodje, ljudi u holandiji imaju veci IQ od ljudi u Africi, zasto onda g faktor objasnjava bolje inteligenciju ovih prvih? Valjda bi, ako su principi oko distribucije bazirani na nekoj 'pravoj' tj organskoj a ne matematickoj podlozi, trebalo da bude obrnuto, tj da se bolje objasnjava distribucija ciji donji deo je dominantan. da zakljucim, mislim da je taj g faktor zaista mnogo zloupotrebljavan u raznim interpretativnim zahvatima koji bi trebalo da objasne sta je inteligencija i na cemu se zasniva. Faktorska analiza sama po sebi je tek jedan malo slozeniji korelacioni pristup kome se neosnovano pripisuju nekakve eksplikativne moci. volela bih da vidim grafikon koji mi objasnjava kako se g faktor distribuira u populaciji - da on stvarno postoji, to bi se jos moglo i uciniti jer bi se mogao napraviti test koji bi ga bio u stanju adekvatno meriti. Svidja mi se jako sto se iz eseja vidi da su testovi skrojeni da daju odredjenu distribuciju i odredjene rezultate, ali mi se ne svidja sto i sam g faktor onda nije vise kritikovan na isti nacin.

[betty]

margot! :)hvala na osvrtu, odgovoricu kroz koji dan. na neke stvari imam odgovor ali o nekima moram da razmislim.nazalost, nije tacno da rad nema puno referenci, zapravo ih ima gomila ali ih nisam navela jer nisam imala vremena da ih trazim. i nije to toliko problem za popularni tekst, problem je sto se radi o stvarima koje nisam licno procitala a nemam poverenja u bas sve podatke. npr. onaj da 50% gena kodira nesto sto ima veze sa radom mozga, to je veoma neprecizno i ni ja ne znam tacno sta znaci. verujem profi da se dobro secao brojke, ali pitanje je npr. da li pricamo o onom delu genoma koji varira izmedju ljudi ili o citavom genomu ili o citavom minus junk dna itd. no, moguce je da cu jednom nesto uraditi sa tekstom. vise razmisljam u pravcu modelovanja sa distribucijom koja nije normalna, ali ne znam kako se to radi i da li je trenutno uopste moguce.

[betty]

imam samo jednu primedbu: iznosis tezu da se inteligencija zasniva na nekom uopstenom neuroloskom supstratu, tj razvijenosti mozga kao takvog koja se kodira u skupu od 15000 gena, ako sam dobro shvatila. Medjutim, postoji veliki broj radova koji povezuje g faktor (posebno njegov fluidan deo) s testovima prefrontalne funkcije, posto su oni najvise zasiceni tim faktorom. Sta mislis o tome, vidim nisi se u radu osvrtala na te nalaze?

uvek mi je zanimljivo kad dobijem ovakav fidbek, koji upucuje na ideje i pitanja koja se nekim logicnim sledom javljaju kod citaoca a meni ni ne padnu na pamet. i sad kad vidim primedbu mogu jedino da se slozim da taj deo ocigledno nedostaje.elem belem, prvo i najbitnije, u ideji Penke-a ne postoji neki jedan mozdani supstrat (~g(?)), nego se bas radi o funkcionalnom svojstvu koje je emergentna (recimo) funkcija hiljada i hiljada sitnih strukturnih elemenata ciji se efekti sabiraju na nivou fenotipa. povezivanje g s jednim centrom je, iz perspektive 'prirode g' (sto je zapravo bila okvirna tema eseja) sasvim drugacija perspektiva jer se odmah dovodi u vezu sa idejom da je g neko tamo jedno svojstvo kojeg u mozgu mozemo imati vise ili manje. iz te perspektive je i ocekivano da se moze videti u mozgu. iz penkeovog ugla ne bi trebalo da je vidljivo u samom mozgu, ili, vidljivo je na milion razicitih nacina (dakle prakticno nevidljivo). ovo sto ti navodis je delimicno tacno. naime, jeste tacno da postoje radovi koji povezuju prefrontalnu funkciju sa intelektualnim zadacima u smislu da je to deo mozga koji je aktivan kada angazujemo vise kognitivne funkcije. medjutim nije tacno da je to povezano s g. da bi bilo povezano sa g, potrebno je da to bude neka mera individualnih razlika: da inteligentniji ljudi pokazuju vise (ili manje) prefrontalne aktivacije od manje inteligentnih onda kada se intelektualno naprezu. takvih radova ima veoma malo, posto se mozdana aktivnost retko posmatra kao funkcija individualnih razlika i obicno osnovna jedinica analize nije ispitanik nego mozdani centar. neki radovi ipak postoje, i citala sam pregledni clanak i meta-analizu takvih nalaza i u najmanju ruku je neuverljiva teza da postoji centar u kom se odrazavaju individualne razlike u inteligenciji. prvo, gomila radova nema testove inteligencije nego neke zadatke koji jesu intelektualni ali nikad ne bi usli u iq test, i drugo, u najboljem slucaju se do 30% nalaza slaze da ima negde nesto (valjda) parijetalno. odnosno 70% se ne slaze. otprilike, u toj sumi podataka mozes da vidis nesto samo ako pokusavas da nadjes nesto odredjeno i gledas samo one nalaze koji potvrdjuju tvoju ideju, inace nema sanse da ga vidis. mene svakako nisu ubedili.

isto, deluje mi malo kontradiktorno sto prvo navodis da u donjem delu distribucije postoji manje varijabilnosti, i da g faktor bolje objasnjava razlike u tom delu distribucije. S obzirom da se iz grafikona vidi da je gornji deo strm i da u njemu vlada veca varijabilnost, onda nije ni cudo da se donji deo moze uspesnije modelovati nekim takvim testoloskim konstruktom koji je verovatno samim izvodjenjem bio pod vecim uticajem tog sporo uspinjuceg 'glupljeg' dela distribucije. dakle, to mi pre lici na artifakt merenja - testovi ustanove varijacije, koje onda objasnjavamo faktorom izvedenim iz istih tih testova.

nisam sigurna da razumem prvi deo. mozes li da me citiras :) gde kazem da u donjem delu ima manje varijabilnosti? [ne mislim da fenotipski ima manje varijabilnosti u donjem kraju distribucije, dapace! mislim, da se izrazim nekorektno, da u populaciji ima vise glupih nego pametnih - ali da unutar donjeg kraja distribucije postoji vise nacina na koje covek moze biti kognitivno usporen nego sto na gornjem kraju ima nacina na koje moze biti bistar] pazi, sama negativno asimetricna distribucija ne postoji kao nalaz - to je moja ideja, da bi distribucija mogla biti takva, uzevsi u obzir penke-ovu ideju o balansu novih mutacija i prirodne selekcije i jos nekoliko osnovnih postavki teorije evolucije. distribucija faktora se ne moze meriti, moze se samo pretpostaviti i onda gledati koliko model sa takvom distribucijom objasnjava postojece varijanse na nivou grupe za koju postoje podaci. s tim sto se uvek, koliko ja znam, pretpostavlja normalnost latentnih varijabli.

Takodje, ljudi u holandiji imaju veci IQ od ljudi u Africi, zasto onda g faktor objasnjava bolje inteligenciju ovih prvih? Valjda bi, ako su principi oko distribucije bazirani na nekoj 'pravoj' tj organskoj a ne matematickoj podlozi, trebalo da bude obrnuto, tj da se bolje objasnjava distribucija ciji donji deo je dominantan.

vidi gornji pasus. u modelovanju faktora (a faktorska analiza je uvek modelovanje), pretpostavlja se da je g normalno distribuirano i onda se gleda koliko takav model odgovara podacima. ako 'realno' g odstupa od normalnosti, model ce objasniti manje varijanse. dakle, ako imas dve grupe podataka sa dve distribucije gde je jedna bliza normalnoj, isti faktorski model ce u njoj objasniti vise varijanse. u tom smislu, u siromasnim zemljama ocekujem (ako je moja ideja o distribuciji ok) da je 'realna' distribucija vise asimetricna nego u bogatim zemljama gde nema toliko sredinske deprivacije, i razlika u objasnjenoj varijansi bi trebalo da sledi samo iz toga.postoji i alternativno objasnjenje pretpostavljene negativno asimetricne distribucije koje nisam uzela u razmatranje ovde, a to je da reziduali nisu nezavisno i normalno distribuirani duz cele g skale. to moze da iskrivi distribuciju g na isti nacin. ukoliko bih dobila neku numericku potvrdu ovako iskrivljene distribucije morala bih da nadjem nacina da odbacim ovo objasnjenje.

da zakljucim, mislim da je taj g faktor zaista mnogo zloupotrebljavan u raznim interpretativnim zahvatima koji bi trebalo da objasne sta je inteligencija i na cemu se zasniva. Faktorska analiza sama po sebi je tek jedan malo slozeniji korelacioni pristup kome se neosnovano pripisuju nekakve eksplikativne moci.

slazem se, kauzalnost se krije u nasim idejama o inteligenciji a ne u numerickoj analizi koja izbaci model s nekim tamo faktorom. opet, zato mi se svidja penkeova ideja, jer g kod njega ima jasnu vezu sa mozgom i sa genima (a inteligencija je sigurno povezana sa obe stvari) a inteligencija je opet funkcionalno svojstvo - fenotipska uspesnost. g samo po sebi ne bi trebalo da bude vidljivo i direktno merljivo. ali iz tvog osvrta mi deluje da to nisam dovoljno naglasila u tekstu. hmmm.

[BraveMargot]

betty, kad budem imala vise vremena odogovricu na tvoj prethodni post, hocu da nacrtam i uskeniram primer distribucije na koju sam mislila, posto mi se cini da se ne razumemo u potpunosti. na zalost, ne verujem da ce to biti pre aprila, maja.Splitovala sam ovaj topik kako bi diskusija o inteligenciji postala preglednija, u medjuvrmenu naisla sam na jedno zanimljivo poglavlje koje govori o promasenim naucnim istrazivanjima veze izmedju velicine mozga i pameti (link), vodjena vrlo slicnim zabludama kao i prethodno diskutovana ispitavanja o inteligenciji i rasi. Naravno, posto se radi o velicini mozga koja sama po sebi nije PC-relevantna pa se zato sme povezivati s podrijetlom vrsta, ovde itekakvog udela u pogresno iskonstruisanim zakljuccima ima i jedna od najpausalnijih naucnih oblasti ikada - bihejvioralna genetika.

[betty]

nema frke, a ja mislim da cu u novembru imati jedan slobodan vikend kad cu moci dalje da elaboriram :)

[Indy]

...ne verujem da ce to biti pre aprila, maja...

... mislim da cu u novembru imati jedan slobodan vikend ...

:DDok nam se ekspertice ne vrate, evo nesto lakse za nas sa jeftinijim ulaznicama:Istraživanje: Najmanje inteligentni su oni koji slušaju hip-hop i R?n?B

[Pikoleta]

svako ima neko svoje misljenje, neko svoje vidjenje glede tog ikua. jednima je svejedno i nemaju neki stav o tome, II bi ga stavili u sve i svasta, u bunar znanja, u jamu hiphopa, u cisternu r'n'b...u dno gdje nema kraja i ostali bi suvi, III, oni s najvecim, dali bi, skupo platili za komad dobrog ikua, za komad svjeze expertize pa i za staru koku dobrog starog r'n'r, IV bi dali pare i poslije testiranja trazili nazad, kao reklamaciju, kupili, platili, dobili rezultate i poslije dobrog skontavanja kad ih uhvati 'mala snaga', pokusali izgledati nezadovoljno ne bi li dobili pare nazad. V su uvijek u nekom elementu i sad mogla bih nabrajati kakvih sve ima, ne bih znala i ne bih nikad stigla do kraja, VD su recimo experti, kad ih vec spomenuh, VD slusaju punk i na to gledaju cisto kao nekakav posao, posao u zadovoljstvu tj. zadovoljstvo u poslu iliti posao i zadovoljstvo idu skupa. oni plate i poslije odradjenog cina gibaju dalje, nista do oktobra, eventualno razglednica sa zelenortskih ostrva, jerbo tamo gdje su jednom bili vise se ne vracaju. samo dalje placaju, novom sadrzaju, novoj analizi, novom modelu, uvijek istom svrsetku. mozda svi ovi opisani izgledaju isto ili barem slicno, mozda u svakom od njih lezi, zivi onaj II, ili III, no jedno je sigurno, jedno ih veze, svaki od njih dao bi koji dolar za jedan smotuljak ikua oko njegovog hipotalamusa.

[Indy]

Mladom hakeru ostaje jos da izvede korelaciju izmedju SAT testova i inteligencije. Sto se mene tice, inteligentniji sam za pet standardnih devijacija od slusalaca Betovena, iako vise cenim kvalitete Bijonsice. Odatle bi se dalo zakljuciti da su slusaoci hip-hopa daleko devijantniji od ljubitelja klasike.

Не знам за курвејшс Бијaнсeицу, али ја бих увек типовао на Снуп Догидога и Грендмастер Флеша када је у питању тест из шизл ма низл интелигенције на улицама Бронкса, радије него на било ког Бетовен-слушајућег ајвилиг дипломца. С друге стране, и Бетовен је испунио своју квоту девијантности кроз литерарно дело Ентони Бурџиса.Науплејинг - Де Ла Сол - Ринг ринг ринг. Уопште не звучи глупо, осим телефонског броја 2222222. Edited March 18, 2009 by Indy

[billadni]

Nego, da se vratimo na temu.Kakva je korist od statistike, kada ja uvek srećem pojedinačne slučajeve? Za izolovan slučaj, statistika je bezvredna, to svi znamo. Mislim, iako znam koliko ću šestica da dobijem u 1000 bacanja kockice, na koji broj da uložim sve pare ako se s nekim opkladim u jedno bacanje?

nema slučaja. pravi slučaj bi bio uzročno nekorelisana pojava. staistika pomaže da odredimo verovatni ishod složenih dešavanja. naprimer, kada je verovatnoća veća od 50% da dupliranjem dobiješ. ti staviš 1000 puta po 1EUR. dobijaš. imaš u džepu. daklem, edukativna je samo za statističare. Edited March 18, 2009 by billadni

[Aineko]

imam jedno pitanje (mozda glupo, ali sto bi rekli u South Parku - nema glupih pitanja, samo glupih ljudi ) :kakv je stav zvanicne phisohologije prema Mensinim IQ testovima (ako uopste postoji nesto kao 'stav zvanicne psihologije prema Mensinom testu'...) i prema njihovoj tvrdnji da je to jedan od najpouzdanijih* testova jer je i jedna od najkoriscenijih (ili vec tako nesto)?*sta god se podrazumevalo pod pouzdanoscu IQ testa, ovde je vec puno o tome receno.po svemu ovde recenom, meni se kao laiku za psihologiju namece zakljucak d ataj test ne meri nesto mnogo vise od sposobnosti da resis tih 36 zadataka u datom vremenu .

[betty]

slusala sam nekoliko kurseva o inteligenciji u dve razlicite zemlje i cini mi se da mensa nikad nije pomenuta. a ni ja ne znam skoro nista o njoj. ali: da li si sigurna da mensa koristi posebne testove? ja ne znam, ali sam pretpostavljala, da koriste vec postojece testove. tipa katelov ili ravenov.

[Д]

Управо тако, Менса (и друга слична удружења која су иоле озбиљна) немају никакав "свој" тест. Раније је већина националних Менси употребљавала Равенове матрице RAPM-II, а сада Данијелсов Figure Reasoning Test. Поред тога, у Енглеској и САД су у оптицају и Кателов вербални тест и Кателов culture fair, трећа скала. У осталим озбиљним удружењима, поред наведених, у оптицају су још и различите Векслерове скале, Станфорд-Бине (новији, дакле, скале 4 и 5, итд. и сл.). За разлику од озбиљних удружења, у неозбиљним удружењима се користе некакви тзв. тестови који нису прошли никакву стандардизацију, већ само посредно нормирање на нерепрезентативним узорцима од по пар десетина људи.

[boshoku]

Управо тако, Менса (и друга слична удружења која су иоле озбиљна) немају никакав "свој" тест. Раније је већина националних Менси употребљавала Равенове матрице RAPM-II, а сада Данијелсов Figure Reasoning Test. Поред тога, у Енглеској и САД су у оптицају и Кателов вербални тест и Кателов culture fair, трећа скала. У осталим озбиљним удружењима, поред наведених, у оптицају су још и различите Векслерове скале, Станфорд-Бине (новији, дакле, скале 4 и 5, итд. и сл.). За разлику од озбиљних удружења, у неозбиљним удружењима се користе некакви тзв. тестови који нису прошли никакву стандардизацију, већ само посредно нормирање на нерепрезентативним узорцима од по пар десетина људи.

jadno, ne ti vec testovi.pokusaj rasijalizma da se vaspostavi, kao u-nauci-ukotvljenji-metod, uspostavljanja, na bazi prirodjenog potencijala, legitimne taksonomije drustvenog potencijala, pojedinaca koji ga cine. otuzna glupost.

[Aineko]

Управо тако, Менса (и друга слична удружења која су иоле озбиљна) немају никакав "свој" тест.

aha, OK. tnx za pojasnjenje,, betty takodje.(a da li je mensa 'ozbiljno' udruzenje je sto se mene tice - diskutabilno, ali nije tema :) )