difficult questions about g
TRANSCRIPT
1. The notion of g has been established from apsychometrician’s point of view for the better partof a century now - but it has almost been totallyignored within cognitive psychology – why?
Here are some suggestions:a. The desire to produce a universal architecture
had primacy. The logic was that characterisingindividual differences would involve variations inthis universal architecture which had to bespecified first.
b. A majority of cognitive/experimentalpsychologists are very narrow in their scope. Mosttheories of short-term memory phenomena, forexample, while focussing on the phonological basisof such phenomena, have historically failed to takeserious account of the cognitive structures requiredfor speech production. Cognitive consideration ofintelligence requires a broad picture since it ispervasive.
c. The IQ battles of the 60’s were traumatisingfor the first generation of cognitivists who couldnot bear having sociologists yelling at them. Sothey stayed away.
d. People working in problem solving, probablythe closest cognitive area, got tied up in thetechnology of implementation or in charting thepeculiarities of human logic.
e. Developmentalists were confused about thedistinction between behaviour and cognition.
2. Why should cognitive psychologists pay anymore attention when John Duncan claims thatfrontal functions and g are synonymous?
The claims are in need of careful analysis. Forexample, the design of the Duncan experimentsmakes it clear that there is no simple task analysisthat could point to cognitive factors that wouldmap onto g. What we do not know, however, iswhether the selective recruitment of lateral frontalcortex that Duncan has found when subjects areperforming high g tasks corresponds directly to ahigh g function. To quote Duncan et al. (2000) thiswould be “a specific frontal system important inthe control of diverse forms of behavior.” If itdoes, then the notion of synonymity is plausible.However, we are still no closer to a cognitivecharacterisation. On the other hand, this selectiverecruitment might actually correspond to some sideeffect of high g tasks, such as the need for specialattentional control, special workspace manipulationor similar computational requirement.
3. Under what conditions shouldneuropsychologists take g seriously – when it
has been localized or when it has a cognitivetheory?
This seems to me to be a category error. “g” isa psychometric derivative. It should not occur as alabel on an information processing model or on amap of the cortex. We can ask what function(s) itcorresponds to in cognitive theory and then drawthe conclusion that this function is implemented inthe frontal lobes (treating the Duncan claims asdata). This might seem pedantic but we have toavoid the urge to think of g in the singular outsidethe confines of psychometrics. To start with, thereare likely to be developmental factors to be takeninto account – which, inevitably, will have broadramifications. Such factors would in effect specifyparameters that would affect a range of processes.Speed of processing is an example of this. Inaddition, there will be specific factors that willrespond to what Duncan et al. call “a broad rangeof different cognitive demands.” Such constructscould have the same psychometric status.
4. What might a cognitive theory look like?One framework that might be useful is Causal
Modelling (Morton and Frith, 1995; Morton, 2004).This makes an explicit distinction betweenbehaviour and cognition and between cognitive andbiological representations and constructs. A samplemodel is given in Figure 1. The arrows in thisdiagram are to be interpreted in terms of “X isessential for the development/operation of Y”. Theframework is a device for representing specifictheories, though in the figure I have imagined atheory with elements from a number of placesrather than taking any particular one.
We start with genetic influences which are shownas affecting the development of four brain regions aswell as something I have called “factor G”. This, inturn, is shown as affecting the four brain regions. Iffactor G is a variable, then all the brain areas will becorrelated with respect to that parameter.
Of the four brain areas specified, three of themmap onto relevant cognitive functions. Thus br1 isspecified (in this hypothetical theory) as thelocality where the cognitive processes responsiblefor knowledge acquisition can be found. The moreefficient the neural functioning of br1, the moreefficient will knowledge acquisition be. Theacquisition of knowledge will also require an inputfrom the environment, of course. The area br2 hasan equivalent relationship to the acquisition ofsymbolic skills. A frontal area of the brain ismapped onto executive functions. Of course, in any
Cortex, (2005) 41, 232-233
FORUM ON INTELLIGENCEDIFFICULT QUESTIONS ABOUT G
John Morton
(Institute of Cognitive Neuroscience, University College, London)
Cortex Forum 233
proper theory the term “executive function” wouldbe seen as far too vague, and we expect the precisenature of the function to be specified in thecontext, since there is ample evidence from theliterature on executive dysfunction to showdissociations between functions.
The final stage of the causal model is theinclusion of behaviour. In this case we include fourdifferent IQ tests. Two of them have been specifiedas knowledge rich, two of them as involvingsymbolic skills and three of them as involvingexecutive functions. The fourth test is envisaged(idealistically) as not being dependent on anycognitive function.
Within this particular causal model, we can seethat performance on all four tests would becorrelated by virtue of the common influence offactor G. Performance on the first three of the testswould be more highly correlated because of thecommon influence of executive functions/br3 – if,that is, that we assumed that it was the same
executive function involved in all three cases. Ifdifferent executive skills were involved in the tasksthen the theory, and so the model, would have tochange.
Of course, the model put forward is far toosimple minded, but gives an indication of thedegree of complexity we have to be able torepresent as cognitive psychologists if we are tobecome interested in “g”.
REFERENCES
DUNCAN J, SEITZ RJ, KOLODNY J, BOR D, HERZOG H, AHMED A,NEWELL FN and EMSLIE H. A neural basis for GeneralIntelligence. Science, 289: 457-460, 2000.
MORTON J. Understanding Developmental Disorders: a CasualModelling Approach. Blackwell: Oxford, 2004.
MORTON J and FRITH U. Causal modelling: A structural approachto developmental psychopathology. In D Cicchetti and DJCohen (Eds), Manual of Developmental Psychopathology.New York: Wiley, 1995, pp. 357-390.
John Morton, Institute of Cognitive Neuroscience, University College, London.e-mail: [email protected]
Fig. 1