[00:00:05]
>> You know it's my pleasure to be editors Professor Andrew Conway for our talk this morning is a professor of cognitive psychology at Claremont Graduate University and his research is aimed at understanding individual differences in cognitive ability is research is concerned with how cognitive abilities are defined and measured and what role they play in real world code mission received his bachelor science and computer science and psychology from Union College in Schenectady New York then received both masters and Ph d. in experimental psychology from the University of South Carolina on receiving his doctorate he began teaching at the University of Illinois Chicago where he served as both assistant and associate professor from there he worked as a lecturer and later as a senior lecturer at Princeton University before moving to where my Graduate University in 2015 he has won numerous teaching awards throughout his career and 2001 he was the u.i.c. nominee for us a professor of the year.

[00:01:15]
Record also awarded the excellence of teaching award in 2000 and in 2013 over 150000 students and rolled in his online course the to 61 he has published extensively on the topics of intelligence working memory and cognitive control his research integrates cognitive psychology 2nd commit tricks and neuroscience He's a former associate editor of The Journal of Experimental Psychology general and he's currently an associate editor at The Journal of intelligence though it's my great pleasure to introduce Andrew and please take it away.

[00:01:58]
Thank you very much for that kind introduction if this makes me feel old. Talking to this or that for protect Ok so. In this talk I'm going. I can for myself I guess in this talk. I'd like to tell you about some more recent work I've been. Particularly this new year I go framework my collaborator and I Christoph coach proposed called process overlap theory which is designed to account for the general factor of salt so to tell you about a process overlaps a real start by talking about some psychometric models of intelligence as well as cognitive models of cognitive ability to launch into an overview of the possible and then I'll show you how pop explains away the general intelligence on the other hand if there's time I packed in a lot near this and I overly ambitious top of my.

[00:03:10]
Pop access to specific prediction that other theories of intelligence most serious of a cause do not make and that's about something called ability differentiation I'll talk about program. Terry how do we want to handle questions. There really see very well. Well I'll tell you what I'll keep track of that shot Andrew and.

[00:03:38]
And if you do would be you happy with the questions during the talk. Yeah I think if we could limit the answer to questions of clarification and not deep seated questions there's a number of deep issues that I'm going to get into that I think would be better for the Cure I mean yes but question of a clear vision Absolutely yeah please Ok well so I'll keep track of the chat and I can feed you some questions when we get to that point but and the general quick q. and a everybody please feel free to on mute and go ahead and speak up.

[00:04:14]
Ok so. Just to preview you know my my approach to my research is really lies at the intersection of cycle Earth's cognitive psychology and part of neuroscience and it's one of the reasons why I'm so attracted to this department at time I think because this is department is one of the few that does a really nice job of integrating these 3 levels of analysis and I hope study that that's what my program of research jobs.

[00:04:44]
So. I'm going to start a cycle matters because process overlap theory was designed to explain the general factor intelligence of course General factor of intelligence starts in psychometrics So I'd like to begin a story with the positive mental So the positive manifold just refers to the pattern of all positive correlations that are server typically observed when you master of diverse fabric toss to a large group of heterogeneous healthy young adults or healthy children.

[00:05:23]
So you need a large very toss large group of people I always you know especially when I talk to more popular audiences I don't really need to say this here to the school psychologist tact but just a calculus of course these correlations are not are not what I'm right so I'm not saying that if you do well on 100 times you're absolutely going to do well on another and then also I mean there's this phenomenon of ability differentiation so those patterns of all public relations it tends to be with her higher levels of ability something I'll get through trivia So just to demonstrate this is this is probably not necessary for this crowd but just in case there are some people who are not familiar with psycho matters I just want to quickly walk you through this because this is what the theory is designed to explain so let's just do a simple simulation where we assumed that there are 9 manifest variables these are of service.

[00:06:25]
Areas if you think of the Minister for kinds of cognitive tests designed to measure 3 probabilities So let's say crystallize intelligence fluid intelligence and visual spatial abilities or. Let's say we have about 250 people. So if I. Go over all that data of pull of a dart and run out scatter plot make sure.

[00:06:50]
This is typically what you're seeing now these are simulated data that's your polling perfect rights are or are actual data don't look like this claim but wait and see is where the 1st 3 surface loads of say these are all tests of crystallised intelligence you see very strong positive correlations among those Here's 3 houses say fluid intelligence you see very the same power and down here for his Will special you see the same patter but overall is the positive manifold rights of the slope of all of these Red Guards going up and so all of this house or possibly poorly it that's what Spearman 1st discovered in 100004 I so this is you know tomorrow I'm going to talk about the replication crisis and this is probably the most widely rotated finding we have a positive mental so I can I can look at that correlation matrix right and so while our this is nice pattern of conversion by person political rights I see this as nice convergence of hard core Alicia's here for the free pass to crystallize intelligence Likewise for fluid and likewise for his Will spatial I see this pattern of all possible solutions proffered and again this is a this is a simulation so correlations are pretty high because we don't see that in.

[00:08:14]
Real and real studies so that what a lot of people and myself included psychometrics do is we represent this through the traditional approaches we represent us in what's called the late variable model right so I can say that there's 3 basically wrong abilities that we're measuring here. And again crystallized fluid and visual spatial and the idea here is that there's there's this ability this late ability that we haven't observed directly but we can infer from the covariance of the performance of these out to her kinds of chast Now this model doesn't doesn't fit the data a perfectly because these probabilities themselves are correlated right so we need this higher order general factor that explains the covariance among the broad abilities and so this is sort of the Us conventional wisdom in the field of psychiatric.

[00:09:18]
And when it comes to intelligence now I know there's a couple of people in this department who just agree with that. And you'll find that I disagree about too so. But this is I would say the Patel I think it's safe to say that it's health care a lot of oil is one of the is probably the dominant model and the fuel the psychometrics intelligence since since the ninety's so again here we are crystallized fluid visual spatial you know working memory retrieval or long term memory there's some debate about the number of type of these rather buildings but there are several probabilities and there's this higher order factor the 4th or $1000000000.00 question that's been around for over a 100 years down is what is it so I split.

[00:10:09]
Theories into 2 broadcasts here on this slide I think quite a simple as this but just for the purposes of this talk. Though there's just broadly there's general ability theories and what I refer to as sampling models or sent me theories. Though by far the general ability here is a dominate the 1st century of research on intelligence it is my hope that we're changing mat and this new century hopefully there's been a ship o.c..

[00:10:40]
So this started as fireman says of the general ability it reflects a general mental ability so a psychological after the 1st. Positive has a causal effect on performance on a range of different kinds of tasks a very different approach which I'll get into the pop model is essentially all.

[00:11:01]
It argues that instead there's no one general mental ability there are lots and lots of cognitive processes that are just sampled in an overlapping matter of cross about our past which all explain more detail in a few minutes I said Gee as general mental ability has really been the dominant Turkish and for over a 100 years and so tomorrow in my vision talking I'm going to take a look back at the last 25 years since you know since I was a visitor person at Georgia Tech in 1960.

[00:11:37]
Sadly not a whole lot is changed in the field of intelligence so this is from 9094. Op ed by me by a bunch of my 52 intelligence researchers I just want to point out the who starts right away and say it's a very elemental capability. And that really hasn't changed very much for the years.

[00:12:02]
So. Now I'm going to shift gears and talk about kind of psychology and cognitive approaches to this question What is so as many of you know I'm trained as a cognitive psychologist I'm not trying to psychiatry I've picked up that skill along the way not quite for this stuff some of you have.

[00:12:24]
But also my training as it is in straight talking to experiments psychology microbrews and experimental psychology. And so from a cognitive perspective of course this question has been addressed for years you know probably most seriously starting in the seventy's with the work of people like Jim Pellegrino and boss her murder and bus hunts and others so there's really 3 top candidates I would argue if there's prophecies were to marry past the Congress her troll or what Randy would call intentional control.

[00:13:05]
I'm going to focus in on would be remembered so I started graduate school and I think anyone and this paper was a poser 9090 and I went to work in a lab that started working memory. So I'm like working memory be a candidate for Gee why would we think that well.

[00:13:24]
This is from a recent paper by Larry Baron Rosenblum and I really like this paper because they make the case for the need for a standard model sort of like using physics as an analogy and then we know this isn't perfect I remember this is really simplistic this model but like if we want to hand off our science to say like education we're not doing very good job of I think we need a standard model for people to understand what we're talking about right so if you just look at the standard model you know working memories of that all right so it's it seems fair to think that's a candidate for this general ability right and as I said I started growth of school in a working memory lab in 99 why this paper was published in 1009.

[00:14:17]
12500 citations so I kill them in Crystal Palace. Correlation between measures were every facet of and measures of lose all of. Now to understand this argument and to understand where people are coming from with respect to like why they think we're being very passive He is such a strong predictor of intelligence you need to understand.

[00:14:48]
The is these memories and past and I guess I don't have to go through a lot of these Hal with this group what memories that house are but just in case you know in my Especially early in my career I would be a big distinction between what we call simple fantastic this is just a media serial recall tasks and.

[00:15:17]
Strictly because I'm here and complex spent pasts so I think most people in this room know how this work. But here's just an example of a complex bandpass so he operations fantastic what happens is people are presented with an operation on their head and given Mansour they have to say for all whether that answer is correct and then they get a letter they have to remember for later recall and then they go through a series of these and after a series they were told others were the symmetry spanners is a spatial version of this.

[00:15:58]
Screen like this they have to make a decision as to whether that symmetrical about the vertical axis has to be so. That on sabbatical and they remember a location or better poll that's in contrast the simple fan or immediate 0 call where you just get a list of digits and revolt them back or just get a series of locations every ball back you know I think this crowd is familiar because half.

[00:16:25]
So I guess I answered graduate school in my city while. The senior person and live at the time was a woman and you can't or. I feel very fortunate to be underway so she was she was sort of the leader of allowed at that time and this was her dissertation.

[00:16:46]
And so I just sort of followed your how often I was 21 years old and has served. Try try to annul everything into it he was doing she seemed just for a very successful student to laugh and this was her dissertation. And so. She sort of promoted the idea that working memory capacity was equivalent to sort of the amount of activation that someone who brings her very much half of this was motivated by the act are models so right around the time John Anderson published the rules of his rules of mind book he was a year before that or 2 years before that.

[00:17:26]
And so a lot of that were the Jonathan was doing with the arts are always based on the fan effect right so the older people are in the crowd will remember all the sort of the facts the what is the fact for those of you were younger. We would just studies like this where we would how people memorize a series of sentences like this they would have to memorize the lawyers and the books and the plumbers and the park and the plumbers and the church thereafter memorize all these different sciences to get to the point where if I just gave them a sentence and said the lawyers in the back there's like you know a lawyer is not there if I hear mortars and both they say yes.

[00:18:10]
And 5 minutes you're late and these benefactors studies is the number of places a person is associated with so you can see that teacher has a fan of for a plumber has a fan of 3 and so all. They want to be sure of her dissertation was something really really compelling.

[00:18:32]
So which I'm sure was of course as fan size increases the time it takes to verify a sentence so these are just target sentences like the lawyers on the boat. Yeah Boyer's of the most very fast you're faster to verify that science. Teachers are the road to teachers on the road over here now with what Judy did this year measured people's working memory capacity before they did this and the fact experiment and she compared people who were high span people who are in the upper court file working on the stand fast as complex fantastic and she compared them to people in the lower court trial I did some work you see as you get a really steep found for the lower working memory people who are Colo spam and you goes more shallow than effect for high fat now it's what's important is that you still get better rates so the slope this is still significant it's still it's still going up for hasn't become automatic for these high spam participants does that happen if you can see if you could create a situation model for Ga rubber bands these work righteousness then you don't get a fanfic at all but that's not happening in this experiment and so what was even more compelling about you is just situation is that you could then calculate the slope for each individual soldier and when you partial that slope out of the relationship between working memory and safety the relationship went away so the the family fact completely explained the relationship between working memory capacity and scores on the s.a.t. Ok so now you can start to understand why people are looking at working memory capacity as a candidate Bridgie So this was my 1st project in the lab.

[00:20:31]
It turned out to be my master's thesis What we noticed was that. There's a compound in these materials right so the wires in the boat so is the teacher right and the plumbers in the market so it's the teacher so there's conflict there's interference in these materials what would happen if we take that conflict out right so if the lawyers in the boat only lawyers in the boat so that's what I did in my master's thesis and what I argue that is that working memory capacity is not the total amount of activation it's the ability to resolve that interference you know why did I argue that out there here's the here's the abstract.

[00:21:21]
We presented a view that individual differences and attentional resources lead to this is the oil age for hitter's process of intervention so there's 2 conditions here that those 0 2nd edition The once again one second condition ignore those what's important is that you see this really steep slope this is the lowest fans and then the shallower slope this is the high school so what I did here is I mean I replicated if you use dissertation but if you take that interference if you're if you're account for that confound in the materials then move that interaction goes away right so so high spans are still found for the most mass but now the slopes are across groups so that's what led us to this idea that working memory capacity isn't really about passive it's about the ability to resolve interfere and that once you as everybody in this department knows the executive attention theory after very pointed this is belabor this point is that that early work really provided there would be to really support for these I think times are theory which you all know through transfer.

[00:22:34]
And so just remind you you know why do we think it was attention because the previous work was learning a bunch of sentences and retrieval from on from every right well we then pushed it down to a low level task so here's a task Here's an experiment that I did with Nelson County where we had subjects do or die product listing tasks were they had to attend to the left year and they're just they just had a shadow whatever they heard and then what we did is we tracked it is somewhere along the way I think really the task we presented their own.

[00:23:07]
The writer which they're supposed to be blocking are getting right and what we found which was really I think counterintuitive to a lot of people predicted by this theory is that lower the memory capacity of people are more likely to hear their name then high working memory people right so the people with higher working memory process this is better able to block out or hit that information and then there's another paper that we did with Michael Chaney where we showed a very similar fact with the entries because House contact is real simple it just fixate in the center of the screen and a flash pops up in the for it and this other job is to make it I'm a bit in the opposite direction it's actually really really hard.

[00:23:55]
And so we found it. Memory subjects are slower to detect a target when the after making that. So called The Opposite Direction no difference and of in the process of God There we have sort of a visual attention and an auditory attention task and we're seeing the same sorts of differences that we saw in resolving interference from retrieval from arm from Eric again this is why we think it's a really general ability and so that we do these very volatile studies again I know you guys are probably all familiar with this all over this quickly but you know we see that performance on complex found is highly related fluid intelligence where performance on a simple span is not thank you Alex for this.

[00:24:45]
And then this so this is a paper I did when I was an assistant professor at u.i.c. we see the same sort of thing. And then this is still after his favorite side I'm sure Summerville the sockets better from your perspective. So we didn't notice and 2005. Correlation between these 2 concerts about what's out there.

[00:25:13]
So I'm guessing that most people in this room are very familiar with Oxford where they're probably not familiar with it may be surprised to hear me say that is the folly. Is there a lot of problems about it so what are the problems with this idea working memory capacity is g. or even executive actions he Well 1st of all we're talking about coral but earlier it was with correlations we were remembering about the in fluid how it's not g. So there's a there's a 6 of our.

[00:25:45]
It and further words only point 7 Sure it's not well. The correlation between cognitive control influence elegance all that stuff difficult very can ask some of the grad students you're got off. I have point 6 here that's a really rough estimate it's not a formal not analysis as we know it's extremely rare.

[00:26:07]
But it goes deeper than that so we know that working memory is not a unitary system I saw it consists of verbal stories across the spatial story process the central executive and storage is it you know Terry so that we can when you look at working memory there's mechanisms for active meet ends as well as rapid retrieval what I would call robbery free will from outside memory or the pocket of control itself isn't unitary So we go from the office model that there's a new edition shifting up to the top rate was more recent we're coming through my business control argues that there's proactive and reactive most of the troll I did so.

[00:26:46]
This system that I was talking about obviously is to say this is too simplistic approach here 1st sample is the after our recent incarnation of the x. our model and what you see is this fractionation of working memory so the so the ground is working memory and what you're seeing here is years ago doors a lot of prefrontal cortex which is maintaining whole there's there's there's retrieval there's sort of a sort of visual attention and then there's motor and action right so we there's we're fractionating this system and so can we fractured it into different kinds of abilities I think.

[00:27:30]
Now this one I saw a note here because I do want to stress the fact that my work really does it's a great so I could measure its partners like causing harm or science there are a. Great deal support for this idea a fraction of the working memory stroll comes from Congress science so I don't think it's an accident that you know I think the changed on this when I went to Princeton because I was surrounded by.

[00:27:55]
Scientists and sort of hard quarter of scientists. And again I'll mention the work of me off in Friedman and talk river I just want to show you one study that I did during that time that led to sort of push I think you know a different direction and I was up and i Phone or i studied it with Jason China my private student at the time.

[00:28:17]
Where we looked at apps. Sort of neural activity as subjects perform these simple and complex than tasks. So this is this study it's sort of complex I just want to show you want you result. So in this study we had people do these sort of stories and process that ATSIC stands house but it's on the scanner right so we have to do we have to control the timing and in a way that's different from art for my you regularly administer these tasks so we just had suddenly it's do lexical decisions have.

[00:28:57]
Of what's the lexical position items are 4 seconds and then they got a letter to be remembered for later recall so this is basically the equivalent of something like operations batteries that it's a verbal complex that task or there's a verbal processing task and over to store it and then over here is the signature wristband house that I'm sure her lawyers who are making in 4 seconds how many of these judgments and they may and then they get a location a half remember earlier.

[00:29:26]
What I want to point out is we looked at neural activity comparing say top class fantastic to this condition you know it was just storage only so there was no processing component they just did the storage. So this is basically an immediate 0 per call faster simple fantastic.

[00:29:48]
So what we looked at we can we do the contrast between basically complex and simple. If you look down here during 2 phases of the tasks so that you see refers to coding make sense and coordination so that's during this whole process of doing the decisions of maintaining the items but then we also looked at the recall stage of the tasks and just when they got to the point where they had to recall the letters back because that's where we are you that we are is that it is complex and cast stuff that's are not holding items in active maintenance that those items are just a painting away and that retrieval then subjects are going to have to recall them back and so we argue that that is more like an episodic memory process and so in working memory task we should see that the Cadillac tippity which is a which is a pretty novel prediction of the time though here the behavioral data so this is the simple story as condition only.

[00:30:55]
And what you see is you know people do worse when the when the domains back when it's a verbal oral pass there's more interference but they do worse than when there's crossed I mean and you see that same effect in the. Specialty. Areas just the contrast between complex span and supposed back in the coding meetings and coordination stage and this is what you see all the time working memory and Hassen f.m.r.i. studies you see dorsal lateral prefrontal cortex you see.

[00:31:27]
I relly some accuracy we saw that in her contrasting verbal complex suburban example and spatial complex special simple that's been done a 1000000000 times what was novel as a society because when you look at the recall fish right so whether people have to recall those letters back and this is me go back you know so here's the recall stage it's about It starts at 28 seconds it piece at about 32 seconds you look at this peak at 30 seconds or you're seeing is this is medial temporal lobe typical activity so this is our argument for this fractionation of working memory there's something yes active maintenance story but there's also retrieval from long term memory.

[00:32:20]
And of course there's Congress control which we don't think of as unitary either. So. Finally there's all this you know Congress science evidence suggests that notion fractionate the system and there's probably multiple abilities but now it's back out into philosophical right and so psychology remembers biology and philosophy right so let's remember it's like Richardson's in the room it's like Richardson's or fell asleep during image during the brain's lives you can wake up again soon to.

[00:32:54]
Remember in the factor analysis or at least variable model of actor is not a cognitive process right so we can say if John performs better vocabulary house than most people it's likely that he'll perform better on the mental rotation test as well. That's not controversial that's not shopping that's just off the medical Now this is more controversial to say that Johnny used his general intelligence to correctly answer items on both the vocabulary house and majority should pass this this I'm not arguing against this I am going there.

[00:33:38]
Do my times are. Just a few minutes behind where I want to be. So this introduces the possible I struggle with probably you know my early years of Princeton. Was Why does the they were gay should people in test performance are also people who test for a cure so massively don't need each other why do we get all these positive correlation is where and if you look at the mind in the brain everybody seems so thorough job a specific right so this is the tension I would argue between psychometrics where you keep seeing this general ability and cognitive psychology part and there are signs where you see all these arguments from popularity so how do we how do we deal with that tension how do we resolve how do we solve this puzzle that's process of appeared so process over life theory attempts to explain and enter into visual differences and intelligence in terms of intra individual psychological processes and again it's largely motivated by my early research on working memory hard to control classic experimental psych work individual differences were an f. m.r.i. So starting point for for profit is over life theory and my graduate students and everyone at her office probably sick of hearing me say this is talking to cost or never process cure is even something as simple as the n.t. cost is not process if you're.

[00:35:10]
So. In any task there's lots of doing specific process if there's lots of them in general processes now what we argue is that there sample the overlapping manner across about every of tests. Furthermore we assume that exactly the attention process is our sample more often so the probability of sample you make that a potential process is higher than the sampling say some domains the subjects they burrow process and executive attention processes are sample more often tests of fluid intelligence and you get that seems reasonable because they're not a whole problem solving task.

[00:35:56]
Again what process or a lot of theory argues is a shield is not real and I'm not the 1st argue this and they've got it Bill in this department has been making this argument for years so what i'm should be clear what I'm saying is like I'm better she obviously is real that's unfair fastbreak me that's it I mean it's an unfair final.

[00:36:17]
So sorry Mr g. obsolete is real and it's important to predict lots of possible life outcomes no denying that. But psychological Jeev radials so it does not reflect in general mental ability it is not a cycle psychological attribute. What that means is a composite cognitive abilities Knox's. Again back to the question What is she there are these 2 broad camps process overlap there is a modern version of Thompson's original sampling model and again let me for those who may not be familiar with it unfortunately I can't see the crowd here there are no.

[00:37:00]
But I usually do at this point as I ask people you know who in the room has heard of spearmint. And then I ask who is who has heard of Godfrey Thompson and usually not a whole lot of people in this room maybe people know who got for tossing but most places like.

[00:37:17]
So not every Tom said introduce the idea of sports simply models of this is Thompson. Former boss introduce the Mutual is a model of 2006 which is also sort of a sampling model and this paper by our fall here period was really influential and I think there is are 2 Cypress papers that sort of revive the sampling wall and so I think the best way to explain a sampling model is just to show you this illustration so sooner a bunch of tests.

[00:37:48]
And we have lots of different talking to prophecy. And so these 2 tasks are going to reveal a high correlation because they share several profits the correlation between just one in Test 2 is going to be higher than the correlation between tests one test 6 because test one in Test 6 only share one process what's key to point out here is that there is no one process that is just a sample by all the tests I so there's no one ability the sample by all the tests is dead the positive manifold emerge is due to the overlap of prophecies across hats and process overlap very.

[00:38:36]
Domain specific prophecies. Both domain specific and open in general our sample so in verbals house you get verbal processes but also go in general executive prophecies and so on. Process over level theory of our $26.00 the paper it's expressed as both the structural model a very reliable and as an idea responsible.

[00:39:04]
If you want to think of the cognitive model it stresses both domain and domain specific prophecies it has a central role or memory I mean decide that a central role of talking to control and work or executive function they're here for the structure a lot of the sort of the groovy images I think it's forgive me.

[00:39:28]
The things I know is this is not how we normally reside of labor. This is for teaching purposes but the 1st thing to know is that the arrows going from these broad ability go maids are going up towards g.-d. not the other way around so we think of g.-d. as a formative factor that is it's just that in that sense of one's overall cognitive ability and so this again is exists it's real but it should be modeled as an index of one's overall ability not one's general mental ability not down here whether these are reflective or formative is still sort of it open question in this in this illustration there are reflector then expresses a response all I don't really have the time to go into all the details of this I know there are some people in the crowd here who are experts on hierarchy so so those people later probably read through this quickly but I just want to point out 2 things about this this is a fire to model it says that there's the probability of a person getting an individual Imus has arrived is a function of their ability and true parameters the discrimination parameter for that item and the difficulty parameter for it I don't what I want to point out is that we assume that there are multiple domains though like Per Wolf facial.

[00:41:00]
And we assume that there are multiple talking to processes within I don't mean like within verbal You could do verbal rehearsal or you could you Chunky. Or lots of other things the other thing to notice about this model is that process ease and abilities are compensatory within the domain so if one verbal process fails I can substitute in another girl process but they are not.

[00:41:28]
Cross dummies So if a verbal process fails I can not compensate with the spatial process that's where most of the really important with respect to executive attention is it means that executive attention serves as a bottleneck because executive attention is is is it's own during. The slightest tell you are prepared to 7.

[00:41:52]
Hours of think about. The know how it is in general it's intelligence still explained away we've done a series of simulation studies to show this we've also goes read it with a bunch of network models and then again there are several predictions the pot makes that have been supported.

[00:42:11]
His word I'm going to tell you about ability differentiation. Going a little longer so let me I'll just walk you through the simulation study quickly so this is like how I started the talk at the beginning. Assume we have now let's put up we have a 1000 subjects but let's again assume the same thing I said at the beginning there are 3 verbal counts 3 spatial tasks briefly are exhausted and here we're going to assume one lots of prophecies per Jony we in our simulation studies we didn't do it this so we looked at how many processes during the I'm just showing you one stimulation here for the sake of time.

[00:42:54]
So in this one that I'm sure we had 50 verbal 50 spatial 50 fluid 50 executive or kinds of processes these are Ptomaine general. Now what's really key is we give each subject a score on each one of those 200 processes so for those of you are like really use of the data analysis side of this you can think of a giant data measured sort of 1000 rows for all the subjects 200 columns are all these abilities and that's a very normal distribution and we're talking so all these abilities are our that's key because we want to derive a g. out of this and sure there is no correlation to start from the general model so to contrast pots it's a serve to serve as a baseline is the straw man we have this what we call a general sampling model this is actually the original Thompson model of the Euro Zone Thompson model didn't make a distinction between like executor prophecies and spatial prophecies and didn't vary the probability of them being South pull this was just a very simple model back in the day you actually get by rolling dice and so all we do is say the probability of any one process example is point one and.

[00:44:11]
We have then 20 processes sampled by the pot algorithm what we did is for fluid intelligence to house we said the domain general processes are sampled with greater probability of a domain specific they get that seems reasonable be reasonable because because food Intelligence tests are not novel problem solving tasks we don't have we can't bring our previous nots to bear to solve those sets in contrast to more domain specific tasks like Perl and spatial pass so here we flip the probabilities and now the probability of a specific domain specific process being sampled is higher and probably a bit of a general.

[00:45:00]
That sampling algorithm is what generates. The type of late very long haul so I'm just going again for the sake of times going to jump to their results here. So here's the general sampling. Algorithm the 1st it's just going to give you one general factor because all prophecies are sample with the probability but here's the result you get when you use the sampling procedure according to the pa algorithm so again this is completely based on Poc are out of I don't respond to all this completely based on sure 100 orthogonal abilities and what happens you get a higher order if you model where she is most strongly related to fluid intelligence remembered there is no general ability underlying this generated model it just looks like there's a general ability and I'd argue that's what's going on in our in our studies so just to summarize that simulation.

[00:46:07]
So. The test scores generated by the pile group of they fit a model that peers if there's a general intelligence a way. But there is right and so that higher order of the model that I just serve you that's the kind of hollow most that only intelligence researcher but as you guys know and or are right so in organizational psychology and education across the board and psychological science this is the sort of the beast and the conventional wisdom and I would argue that it's a verb.

[00:46:46]
Ok just a just to give you a glimpse I'm not going to talk about the network model but I just want to show you that. This this model we could fit it right but this is actually incompatible with our theory right so this says there's a general ability that causes his performance and verbal pasts and so on right so we don't assume that there's an overarching general ability so we now don't use these kinds of models like Steve to use a network model and so this is just showing you a network model of psychometric that were possible which just fits the data very well and it shows you that fluid intelligence is more central to this network than the others you see there's more connections between fluid and spatial and fluid and verbal than there are cross domain and actions we've done lots of these are here is a network model of the way where you see working memory and fluid reasoning so we're remembering yellow blue blue those are tend to be in the center and that we're we've done this with the woodcock Johnson we've done this with tons and tons of old late very Wattles before you run them as number bottles and they look like this.

[00:47:58]
So I see that I don't have that much time or you want to open up her for a few may. So I just want to the points of this network model the more consistent with our theory and I I'm sort of a strong advocate of trying to make trying to achieve compatibility between a theory and as a test of all.

[00:48:23]
You know I'm just going to zip over this because of this because of the time but I did want to say that the hot model makes a very specific specific prediction and all ability differentiation for those of you who are the sort of the feel the psychometrics you know this as Spearman's law of diminishing returns so what Spearman discovered in 132 is that the correlations become smaller.

[00:48:50]
As you go higher in the general ability. Distribution boat Derren diagonal just show this very clearly in 1809 with respect for intelligence. So what you see is these are that average correlations across different kinds of tests and they just go down as i.q. goes up against ability differentiation there as people's i.q. or general ability goes up the crossover correlations go down and so the impact or goes down and so I just want to point out that we make that British we should observe that not only in House of intelligence but also in tests of working memory capacity and so recently Christophe and selling all of our I was of the University of Amsterdam the sort of the wizard on moderate factor analysis which showed that results are just for this I want to get to the human race I'm just going to show.

[00:49:53]
These are the factor loadings and what you see is that if you look at the model where you use complex fan. The load is due to the go down as you go up on fluid intelligence and you Joe don't see that when you model just simple sound tasks and again we were predict that because we don't mix well you don't think there's as much of an executive attention influence on these kinds of tasks as there are over here and so this is a very specific prediction that only the hot model were big hits and it's been supported.

[00:50:30]
By this is just very quickly I'll say there's a lot of other evidence in support of pot Here's just a sampling and if you're not convinced already going on this is pretty good support I would say we're thrilled with this but the Journal of intelligence is putting on an entire special issue together and she in its underlying executive prophecy has said it's according to coaching Conway's hot model but now there's an entire special issue not only of the Journal of intelligence but also the Journal of intelligence is now going to Specialist 3 this year the current process is ongoing but I think it's fair to say that we've made impact which I'm proud of so again your take home message is positive the Urkel framework.

[00:51:17]
That accounts for the general backer intelligence iterates in for ideas and evidence from psychometrics partners like ology and resigns as big of Haitians I think for a lot of things what I'm just what I like here is talking to testing and I want to give a hat tip to Phil who has been arguing this for decades thanks are enjoying my time here or really interesting talk.

[00:51:44]
Why don't we open the the floor to questions. Yes you might know I do study cognition psychometrics and especially using on our team models but I'm wondering how you come to the conclusion that you have a particular process in a test. Have you considered doing and just I don't like.

[00:52:14]
It so just just recently I started getting entirety and sorry my my my professor my crusty face so I wasn't trying to that all of us are oblivious for a long time unfortunately we have visited a person here if he tired a couple years ago and all my questions are here and so we have a paper that's about to be submitted we're looking at its iron speed of these different kinds of memory standouts.

[00:52:45]
And so but but that said I think you know we're never going to be able to infer process. You know from a psychiatric model directly so we need to rely on commercial operations converging evidence and that's why I try to integrate these different levels of analysis in my work.

[00:53:06]
So I think I think those are those are 2 answers to your question. But I do think this field in general is that the the I or c. And I think you know the kind of work that you do. I think it sorely lacking from this promise feel like I'm surprise you know my grad student was not by snow never done this before.

[00:53:31]
I was like well none of us were trained it as her words are say I'd like to see more work I'd like to see a lot more iron to work and I think our model of and to help you know God. That's our Feehery. Ok Well just consider I'm difficulty modeling I have difficulty model.

[00:53:57]
And so long as you're considering different i.r.t. models another one you you might want to consider is an unfolding model some of the. Things you've presented are very consistent with what psychometricians has known about proximity data so when you showed that sir complex pattern of correlations between different domains.

[00:54:24]
There is work out there that would show if you try to model that way s.. Some type of linear s.c.m. or factor analysis model what you're going to produce is a artificial 2nd factor which to me seems like that's what general the general I Q factor is it's just applying a linear model to I'm not nonlinear data and one way to get around that might be through an unfolding mechanism and there IOW t.m. folding models and then and.

[00:54:59]
Not in the eye or t. base or more classical based unfolding models that you might want to look at because you mention convergence and I think that would give you some convergence just from another body of literature Yeah absolutely and. You know one of the reasons I'm attracted to the Department of there are people like you I can go to this.

[00:55:22]
For those of you don't know if you're going to grab. The stage or. Any other questions from the audience it want to keep this on the mythological front but you know another methodology has questions. A factory model a typical take nor term for the square of the factory loadings of a factory model is the commonality and then their term for very unique variances is very unique this is so like a one to one factor model supposes that there are commonalities between the general factor and all of the tests and it seems like there's a.

[00:56:23]
High degree of overlap between this process overlap theory and the one factor model. Could you elaborate a little bit on where you see unique quantitative predictive differences between the 2 models Yeah so I think I mean if I may boil over this too quickly I'll start I did so a really important part of the process all out there is the idea of.

[00:56:57]
Executive attention or of what I would call it talking to the troll talking control serves as a bottle that performance and so it's not simply additive where so what you're describing is like you can only get simulations if we put it in that interaction component into the Iraqi model that it doesn't matter like how many prophecies you have like you could I want got a 1000 because the 1000 is just matter and it's so I think that's what you're saying and so what's important is how did that come how did that component in there the fact that you need to have the cognitive controller executive attention process if the dummy general winds.

[00:57:46]
How to serve as a constraint. The rest of the rest of the mile and so that is that's the part that leads to the prediction of ability differentiation and so that's what distinguishes it from I think I'm not sure I answered your question but I think that's what distinguishes it are there data that would speak to the falsifiability of that supposition Yes I think that's that's why I don't want to present your ability differentiation results up yeah right so if you did observe ability differentiation right then that would falsified we do observe ability to Parisian things yeah but if it's hard it's hard to come up with like strong tests I'll admit there's there's something to steer it right you know we're not specifying what the current prophecies are.

[00:58:45]
So like I was the 1st time I ever presented this. Now it's in pounds that to me civil law or the cognitive process that I was like well that's for you to figure out and you'd like that after the. Other person's anybody else that would call it from me.

[00:59:10]
So what what is the best counter-argument is that you've heard when you've done this presentation going to this point you know. Well they don't have this eerily silent right now so. Maybe they could argue. You know they're I think they're sort of the champions of the idea that you know cognitive control or what they call intentional control is right.

[00:59:40]
So. But I'm. Funny What's funny is you know whenever I present this I'm typically presented to talk to the psychologists here comes. Actually presenting it to cognitive psychologists or Congress scientist who really really like this right because they say Now it's with me in the sense I think this is I believe the users of my celerity and dummies the specificity and so I pretty much ignored standard answer it's because it doesn't make any sense and this is I mean this is this is what my friends to colleagues told me there is like you know she doesn't make the sense there's this g. in the brain and so I think it's we've got I know it's a warm reception I think this crowd is a little harder because there's a hardcore psych MacPherson's here.

[01:00:35]
And people who are I think I'm bored with the idea that talking to a troll this is. Sort of the dominant factor. If you will. And one from an angle. Well angle angle will chime in How's that Ok I'm sitting at a call in a car in Atlanta you're listening to your talk on my i Phone just leaving it on most the most appointment.

[01:01:17]
You know 1st of all all my colleagues I love having students who disagree with me. And who can can can weigh in and argue about that and. Even when I think they're wrong. And I think you're thinking of g. more broadly than I mean I don't I think a fluid intelligence as you know as a component that you you know you think of g. as fluid and crystallized basically combine and I think given that that approach to the g. the model says a great deal that I agree with I think it doesn't speak to you know the fluid intelligence or to attention control and I've argued that those are distinguishable I think there is a good neuro science evidence for the ability to control attention and that that is really key to an awful lot of things does it account for everything that is important and human cognition of course not.

[01:02:24]
But you know so I you know I've sort of made a career out of getting. More and more precise in the way I've talked about this stuff and you know I think it really is one of the real benefits of your Christophe's work is focusing at a bigger level and you know but I do think attention control and you know you know me I am perfectly willing to admit when I'm proven wrong but I have a high bar for evidence that does that.

[01:02:58]
But I think attention control is really important and many many many tasks. That said that when you look at the relationships that it's virtually impossible for me to see how there would be specific. What you're calling processes I would call for skilled information processing different skills at processing different kinds been from Asian you know so why would a.

[01:03:31]
Low attention control person be better at detecting their name. And in the diet dichotic listening task it's hard for me to see how you know how that's process specific and I think that's where Ted general tension control does come in you know it and you can look at the huge array of.

[01:03:54]
Jobs and paths and in the you know from motor control to to discover general discrimination I don't see how. The relationship between measures of attention control and sensory discrimination ability would emerge from a pot theory without without thinking about some specific mechanism for attention control so I'll leave it at that right good talk.

[01:04:24]
But though again I mean here in the simulation these are attempts to control properties and are distinct from what really broke Yes and there are some when you go through a fluid reasoning past it's a combination of attention control problem and other properties that are referred to as fluid and this is what is true with you this alliteration of the right because we agree that there is a distinction between attention for a little respect it's just that a test and then again we agree that attempts to control is important because the tissue for it was being sample everywhere across the boundary.

[01:05:05]
And no other process is like that so it seems you're prolife best so it's just not. I agree. Ok well but slickness had a great discussion here is really a terrific talking Andrew and so a little bit out of my realm but I got a lot of new information from it so I really appreciate it and.

[01:05:34]
That we all give him a virtual applause for great. Great talk there on.