164

Review

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

Human intelligence differences: towards a combined experimental–differential approach Ian J. Deary Despite the fact that much is known about the taxonomy and predictive validity of human intelligence differences, there has been relatively little progress in understanding their cognitive bases. However, some recent firm findings mark the beginnings of a cognitive reductionism in human intelligence. Progress towards discovering ‘cognitive components’ that, firstly, show individual differences and, secondly, relate to psychometric intelligence differences is described here at different nominal levels of analysis: ‘psychometric’, ‘cognitive-experimental’ and ‘psychophysical’. The field of intelligence differences remains a fertile yet seriously under-developed demesne in which cognitive scientists should collaborate with differential psychologists.

A companion article to this one (in last month’s issue) examined the structure and predictive validity of human intelligence differences1. An earlier article examined progress towards discovering biological bases of these differences2. The present article addresses the cognitive bases of psychometric intelligence differences. All of these topics are dealt with at greater length in Ref. 3. Cognitive components or elements that might relate to intelligence differences are derived from at least three sources. First, mental ability test items themselves have been fractionated, putatively to reveal mental components: here called the psychometric approach. Second, procedures from experimental cognitive psychology have been used. These are typically reaction-time procedures in which some manipulation of the stimulus–response relationship is said to include (the pure insertion method) or exclude (the subtraction method) one component or element of mental processing: here called the cognitive approach. Third, procedures at an arguably lower level are used in which some aspect of the person’s processing of sensory information is examined to reveal processing elements: here called the psychophysical approach. Psychometric approaches

Ian J. Deary Dept of Psychology, University of Edinburgh, 7 George Square, Edinburgh, UK EH8 9JZ. e-mail: [email protected]

In the early years after psychology’s cognitive revolution the alliance between the individual differences approach to human intelligence and a cognitive approach was well exemplified by R.J. Sternberg’s componential studies of human reasoning4. Starting with analogical reasoning he used a partial cueing technique and regression models to isolate components from items long-used in mental tests (such as ‘sugar is to sweet as lemon is to ?’).

Components such as ‘encoding’, ‘inferring’, ‘mapping’ and ‘application’ were said to be the mental elements of the reasoning process (see Box 1). Linear combinations of the components’ timings predicted performance on paper-and-pencil (i.e. standard psychometric) tests of reasoning. R.J. Sternberg’s reasoning components bore resemblances to Spearman’s ‘eductions of relations’ and ‘eduction of correlates’, which were published in the 1920s (Ref. 5). (Spearman’s work on the ‘principles of cognition’ was dubbed the first cognitive psychology text6.) The components were a lively research topic for a decade or so7, but seemed to run out of steam. Their validity was never demonstrated; it was never clear whether they were or were not arbitrary slicings-up of mental test items3. The original components did not always reliably appear in studies of different reasoning procedures; sometimes these had to be bundled as a ‘reasoning’ (!) component7. For a series of supposedly elementary aspects of mental functioning the components attract little current attention in research on intelligence, or in cognitive psychology. Recent reviews of this and related efforts to split psychometric test items into their components appears unclear about the achievements made using this approach8. More recent research on intelligence differences favoured two more general processes as possible sources of key individual differences, some would say in the general factor. These processes are working memory and general control processes. The former is better characterized and more studied than the latter. Readers might wonder why these are deemed to be ‘psychometric’ rather than ‘cognitive’ in origin. The reason is that within intelligence research these processes have been extracted from, or implemented in, investigations using essentially psychometric types of tests. The rationales of these investigations vary. Some involve careful manipulation of the working memory loads of psychometric test items with the extraction of latent traits that can account for psychometric test battery variance9,10, some involve applying a simple theory of information processing to the assembling of psychometric test batteries11,12, some involve protocol analyses of people’s responses to standard IQ-type test items and their implementation in computer programs13,

http://tics.trends.com 1364-6613/01/$ – see front matter © 2001 Elsevier Science Ltd. All rights reserved. PII: S1364-6613(00)01623-5

Review

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

165

Box 1. Components of analogical reasoning? R.J. Sternberg developed a method for testing models of the cognitive components that are involved in reasoning by analogy, which is common in psychometric intelligence tests and is generally considered to be a central aspect of intelligencea. His components of analogical reasoning include the following: (1) Encode (2) Infer (3) Map (4) Apply Thus, to answer the question ‘A is to B as C is to ?’ with reference to Fig. I, we ‘encode’ the stimuli. We may ‘infer’ that there are two A to B transformations on the left-hand side of the double colon: large circle goes to small circle and two crosses

++

+

A

B

**

*

C

D

TRENDS in Cognitive Sciences

Fig. I. An analogy item like that used by R.J. Sternberg in his componential analyses. The format is ‘A is to B as C is to?’ (D, the correct answer, is shown here.)

go to one. If we ‘map’ the large circle in A to the large square in C and the two crosses in A to the two asterisks in C, and then ‘apply’ the two A to B transformations from C to D, then we should be looking for a small square with one asterisk (as shown) among any answer options. Another example can be found in Sternberg’s later volumeb. The key theoretical claim was to have found these elementary ‘components’ of reasoning. Some criticized these components for being true a priori, rather than an empirical discoveryc; others suggested that they need but lack empirical evidence (see Ref. d, Chapter 5). These types of reasoning items are sometimes said to be inappropriate for such mechanistic analyses (Ref. e, Chapter 9). The issues of people adopting different strategies or shifting strategies demand flexibility in what can otherwise seem to be deterministic modelsf. References a Sternberg, R.J. (1977) Component processing in analogical reasoning. Psychol. Rev. 84, 353–378 b Sternberg, R.J. (1985) Beyond IQ: A Triarchic Theory of Human Intelligence, Cambridge University Press c Kline, P. (1991) Sternberg’s components: non-contingent concepts. Pers. Individ. Diff. 12, 873–876 d Deary, I.J. (2000) Looking Down on Human Intelligence: From Psychometrics to the Brain, Oxford University Press e Mackintosh, N J. (1998) IQ and Human Intelligence, Oxford University Press f Lohman, D.F. (2000) Complex information processing and intelligence. In Handbook of Intelligence (Sternberg, R.J., ed.), pp. 285–340, Cambridge University Press

and some involve tests in which people with certain types of brain damage perform poorly14. Whatever the origins of these measures, there emerge high correlations between working memory and psychometric intelligence15. Some take this to mean that working memory is basic to intelligence differences, others the reverse; and yet others take it to mean that they are two names for the same concept (see Box 2). With regard to the field of intelligence research it is true that some tests of ‘working memory’ look suspiciously like common-orgarden IQ-type psychometric tests. On a more positive note, the realization that general intelligence and working memory might be closely linked brings together two concepts with massive psychometric evidence on the one hand and massive cognitive and neuroscience evidence on the other. A combined research programme on these twin topics must be a priority for research. Cognitive experimental approaches

People with higher psychometric intelligence tend to have faster and less variable reaction times3,16,17. This well-replicated finding tells us about a fairly small part of the variance in intelligence and tells us only as much as we understand about (1) the origins of reaction-time differences, and (2) the cause(s) of the correlation between reaction times and psychometric intelligence. A small but significant association between reaction time and psychometric http://tics.trends.com

intelligence was found in the first three decades of the 20th century18, but the finding was properly established in the 1980s16, by way of a diversion. What attracted the interest of intelligence researchers in the 1970s was not so much the overall reaction times or their variabilities, it was the idea that manipulations of the reaction-time method might afford the extraction of mental elements. Three such procedures were adopted quite widely by psychometric intelligence researchers: the Hick19, S. Sternberg20 and Posner21 procedures. The study of these reaction-time procedures in the context of psychometric intelligence differences is discussed in detail elsewhere3,22. The Hick reaction-time procedure steadily increases the number of stimulus choices in a choice reaction-time setting. If the reaction-time duration is plotted against the base-2 logarithm of the number of stimulus alternatives a straight line function emerges (Fig. 1). The slope of this line was supposed by Hick (it was previously discovered by Merkel23 in the 19th century and also reported by Hyman24 nearly contemporaneously with Hick) to assess a person’s ‘rate of gain of information’. This parameter correlated significantly with intelligence differences in a German study25 from 1964 brought to the English-speaking psychological world by Eysenck26. Replications were attempted by many researchers, especially Jensen16,27. Jensen used a reaction-time device intended to separate the overall response time into decision and movement times. Decision times,

166

Review

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

Box 2. How closely related are working memory and psychometric intelligence differences? Kyllonen and Christal found that general reasoning and working memory differences were very highly correlateda. They assembled a group of processing tests based upon a cognitive architecture, the ‘four sources’ model (Fig. I). This test battery (‘cognitive abilities measurement’ or ‘CAM’) has a general factor that is almost perfectly correlated with the general factor (g) from a battery of psychometric testsb (Fig. II). Kyllonen later discussed the overlap between g and working memoryc. It was argued that general fluid intelligence and working memory both reflect ‘the ability to keep a representation active, particularly in the face of interference and distraction’ (Ref. d, p. 309).

References a Kyllonen, P.C. and Christal, R.E. (1990) Reasoning ability is (little more than) working memory capacity?! Intelligence 14, 389–433 b Stauffer, J.M. et al. (1996) Cognitive components tests are not much more than g: an extension of Kyllonen’s analyses. J. Gen. Psychol. 123, 193–205 c Kyllonen, P.C. (1996) Is working memory capacity Spearman’s g? In Human Abilities: Their Nature and Measurement (Dennis, I. and Tapsfield, P., eds), pp. 77–96, Erlbaum d Engle, R.W. et al. (1999) Working memory, short-term memory, and general fluid intelligence: a latent variable approach. J. Exp. Psychol. (Gen.) 128, 309–331 Fig. I. The cognitive architecture used by Kyllonen and Christal to examine associations between working memory and psychometric intelligence.

Environment (input)

Perceptual processing cycle

Cognitive processing cycle Working memory

Procedural memory

Declarative memory

Cognitive processing cycle

Motor processing cycle

Response (output) TRENDS in Cognitive Sciences

0.994

g

g

1.00

0.89 0.81 0.92

0.85

V/M

0.95

PK

0.96 DK

SPD

TK

PS

WM

TRENDS in Cognitive Sciences

movement times, and the variability of decision times show significant correlations with mental ability test scores around or above 0.2 (Refs 16,22). The parameter that evoked interest in the Hick procedure – the slope function – does no better and often rather more poorly in correlating with intelligence. The Hick slope promised an apparently tractable element of human information processing and was thought to http://tics.trends.com

Fig. II. Structural model showing the association between general factors (g) obtained from a set of ‘paper-and-pencil’-style psychometric mental tests, the ASVAB. The factors of the ASVAB (shown in blue) are verbal/mathematical (V/M), clerical speed (SPD), and technical knowledge (TK). The factors for the CAM (green) are processing speed (PS), working memory (WM), declarative knowledge (DK), and procedural knowledge (PK). Note the almost perfect correlation between the two general factors.

afford a small window on what it means to be of high ability. Once it is realised that all manner of reaction-time parameters correlate significantly with mental test scores it has to be asked how much we have really understood about intelligence. Opinions differ. Alhough they are not exclusive, researchers tend to go for one of two explanations3,22. Some suggest that the correlations between reaction

Review

Lower ability

450

167

are themselves theoretically tractable. Otherwise one has merely linked an unknown to another unknown. The statistical problems with ‘slope’ measures as sources of components that might then be the sources of individual differences are discussed by Lohman30,31.

500

Reaction time (ms)

Fig. 1. The Hick reactiontime procedure and intelligence differences. The middle line (red) shows that response time increases as a linear function of the base-2 logarithm of the number of stimulus alternatives. This is the so-called ‘Hick law’ and the slope was said to reflect an individual’s ‘rate of gain of information’19. The lines on either side demonstrate the hypothesis that people with different mean levels of psychometric intelligence had different slopes. This was first tested and verified by Roth25. Subsequent research found that there were more consistent correlations between intelligence and mean response time and response-time variability than with the slope function.

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

400

Psychophysical approaches 350 Higher ability 300 0 1

1 2

2 4

3 ‘Bits’ 8 Stimuli

TRENDS in Cognitive Sciences

times and intelligence test scores occur because brighter people have faster ‘speed of information processing’. The problem with such a construct is that it is variously operationalized. On the other hand, some suggest that reaction times and intelligence test scores share variance because they both involve some similar higher-level ability such as motivation, attention, test sophistication, and so forth. Studies of this set of ideas offer them little support3,22,28. In support of more reductionistic ideas, the correlation between reaction times and psychometric intelligence is substantially mediated by genetic effects29. Similar stories may be told about the research on intelligence and the S. Sternberg memory scanning task and the Posner letter-matching task3,22. The former task claimed to isolate the time it took to scan a single item in short-term memory; the latter assessed speed of access to long-term memory stores. These ‘mental elements’, it was thought, might show individual differences and these might in turn be correlated significantly with mental test score differences. Again, the overall reaction times in these procedures correlate significantly with mental test scores, but the key theoretical parameters have no special correlation with intelligence test scores, and their effect sizes are often a bit lower than other less theoretically interesting parameters. Individual differences psychologists who have adopted reaction-time procedures from cognitive psychology are in something of a fix. They do find replicable, interesting, surprising and significant associations between mental test scores and apparently lower-level task parameters. The parameters that produce the strongest correlations are invariably not the interesting ‘slope’ or ‘difference’ variables; rather, they are the less explicable intercepts and overall reaction times and their variabilities. This leaves the question of what these variables mean in terms of brain processing. Linking mental test scores to cognitive variables is only really productive when the cognitive variables http://tics.trends.com

The visual information processing task most used in intelligence research is inspection time32 (see Box 3). Individual differences in inspection time are significantly correlated with mental test scores at about 0.4 or above33–37. Correlations are stronger with ‘performance’ as opposed to ‘verbal’ tasks38. There is current debate about the group factor of intelligence with which inspection time correlates most strongly39; it relates to the general factor also. The correlation between inspection time and intelligence test scores does not appear to be caused by higher level factors such as motivation, test anxiety, personality, or cognitive strategies3,22,40,41. There is debate as to whether practice on the inspection-time task enhances or decreases the association with intelligence42,43. A study using a latent-trait approach suggested that the correlation between intelligence and inspection time is an association with the efficiency of visual processing in general rather than a specific ability to perform well on inspection time (see Fig. 2). Performance on the inspection-time task is related to the gradient of the N140–P200 excursion of the brain electrical potential evoked by inspection time and other simpler stimuli, such as the auditory oddball technique3,44. Psychopharmacological agents that decrement or enhance cognition tend to have parallel effects on inspection time45–47. Auditory tasks, intended to assess processing limitations analogous to those tapped by visual inspection time, also show modest correlations with mental test performance3,48–50. Individual differences in something as simple as inspection time are able to mediate much of the variance in ageing and psychometric intelligence51. In summary, the study of inspection time has proved fruitful in linking psychometric intelligence to simpler aspects of information processing, but the nature of inspection time and the mechanisms of the association between inspection time and intelligence are far from fully understood. Issues in applying cognitive reductionism to intelligence differences Causal direction

When psychometric intelligence test scores correlate significantly with apparently lower-level mental tasks the conclusion, easily but incorrectly arrived at unfortunately, is that there has been a successful reduction from the level of psychometrics to components (elements/aspects/parameters) of a cognitive architecture. The lessons of the last 30 years have produced an appropriate scepticism about such

Review

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

Box 3. The inspection-time procedure Typical cue, stimulus and mask from the inspectiontime procedurea are shown in Fig. I. Figure II shows the resulting psychometric function. A cue warns the subject about an impending stimulus. The stimulus seen either as here or as its mirror image are thereafter presented for a duration that can vary from a few milliseconds to a few hundred milliseconds. A mask replaces the stimulus immediately after stimulus offset. The subject indicates (without being required to make a rapid response) which of the two stimuli was presented, typically by stating the side on which the longer line appeared. The mask shown here is only one of various configurations used in different

experiments. Stimuli, too, have varied across experiments, although this ‘pi’ figure is the most common. Tachistoscopes, computer monitors and light emitting diodes have variously been used to present the stimuli. (For a review of inspection-time research see Ref. b, Chapter 7.) References a Vickers, D. et al. (1972) Perceptual indices of performance: the measurement of ‘inspection time’ and ‘noise’ in the visual system. Perception 1, 263–295 b Deary, I.J. (2000) Looking Down on Human Intelligence: From Psychometrics to the Brain, Oxford University Press c Deary, I.J. et al. (1993) Nonstationarity and the measurement of psychophysical response in a visual inspection-time task. Perception 22, 1245–1256

100 90

*

Percent correct

168

80 70 60 50 40 0

Cue

Stimulus

10

20

30

40

50

60

Duration (ms)

Mask

TRENDS in Cognitive Sciences

Time

TRENDS in Cognitive Sciences

Fig. I. Cue, stimulus and mask used in the inspection-time task.

claims3. It is necessary to prove rather than assume that a cognitive variable is causal to intelligence differences, and it is necessary to test the reverse hypothesis and the possibility that some other variable or variables is causal to both. Mental speed

Maturely, differential psychologists have eschewed the attitude which says ‘declare victory and withdraw’. At one time it was common to demonstrate a correlation between intelligence and, say, inspection time, reaction time, nerve conduction velocity, P300, or whatever, and to state that the reason was that these latter variables assessed ‘mental speed’ or ‘speed of information processing’ and that this construct was a part cause of differences in intelligence52. More recently, the proper approach has been to question the constructs being assessed by the supposedly more basic task, frequently with the discovery that they can be almost as mysterious and intractable as psychometric tests. http://tics.trends.com

Fig. II. The relationship between stimulus duration and the probability of a correct response in an inspection-time experiment, using the ‘pi’ stimulus shown in Fig. I. The stimulus duration is varied between zero and 64 ms. Data are shown for two observers, Subject A (blue symbols; 600 trials per data point) and Subject B (red symbols; 800 trials per data point). The data are extensions of those described in Ref. c.

What is elementary?

In most of the cases of attempted cognitive reductionism of intelligence differences that were adumbrated above the research began by a differential psychologist ‘buying’ the idea that some ‘element’ of cognition had been discovered. R.J. Sternberg’s components, the slope of the Hick reaction-time procedure, and the inspection-time construct were originally promoted and accepted by some as benchmarks of the brain’s information processing performance. The broadly accepted position nowadays is that they are probably more basic than intelligence tests, but that they are still poorly understood themselves3. Valid parameters of validated cognitive architectures

Perhaps differential psychologists have been naive. They were eager to accept that cognitive psychologists discover valid elements of cognition and that these might be related to higher-level abilities. Lately, though, intelligence researchers

Review

AliceHeim Non-verbal

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

Visual change detection

0.40 0.56

Alice Heim Verbal NART (verbal)

0.78 F2

0.90

Visual movement detection

F1 0.46

0.79

0.81 0.57

Inspection time

Contrast sensitivity

TRENDS in Cognitive Sciences

Fig. 2. Visual information processing and intelligence. Well-fitting structural model of the associations between psychometric ability tests and visual information-processing tasks. The model tests the hypothesis that a latent trait from three different visual processing tasks relates to intelligence differences, but that variance attributable solely to the individual tasks does not63. The three tasks share the characteristic that the stimulus information is time-limited. Contrast sensitivity lacks the element of time-limited stimulus availability, but is in other respects a demanding visual task. Psychometric tests (blue): Alice Heim 4 Part I (verbal and numerical), Alice Heim 4 Part II (diagrammatic) and the National Adult Reading Test. Visual processing tasks (red): visual change detection, visual movement detection and inspection time. Contrast sensitivity had no significant correlation with other tasks. F1 and F2 are latent traits. (See Ref. 63 for description of the fit statistics.)

have found themselves busily attempting to validate the reaction-time and inspection-time procedures, to see what they are made from53. Had these procedures been based on a proven cognitive model in the first place this would not be necessary. Reducing intelligence differences relies upon there being a valid cognitive architecture with valid constructs and processes. To the extent that cognitive psychologists deliver these, and that differential psychologists become aware of them, the cognitive reductionism of intelligence differences has a chance of success. Intelligence: a multi-level understanding

There is now a tendency for researchers in intelligence to employ cognitive and biological approaches in the same studies. That is, in addition to relating psychometric intelligence to a cognitive or psychophysical variable, they will, in addition, attempt to discover whether these two also relate to some biological aspect of information processing. For example, psychometric intelligence and inspection time have been examined in relation to evoked potentials and psychopharmacology3. An end to ‘greedy reductionism’

Dennett’s concept of greedy reductionism refers to attempting to understand a biological process in terms of unvalidated constructs and mechanisms (‘skyhooks’)54. It contrasts with good reductionism, which is harder and rarer, and means understanding phenomena in terms of validated constructs and mechanisms (‘cranes’). Greedy reductionism occurred when researchers, using reaction times and inspection times, and so http://tics.trends.com

169

forth, stated glibly that intelligence was caused by ‘mental speed’ or ’speed of information processing’. These attempts at reductionism are ‘greedy’ because they falsely suggest that we know what reaction times and inspection times measure. And the lesson is clear. It is not difficult to obtain significant correlations between psychometric intelligence tests and simpler cognitive and biological parameters – it happens with brain size, evoked potential indices, reaction times and inspection times3. The difficult questions follow: (1) what valid construct or constructs is the ‘simpler’ index measuring?; and (2) what are the mechanisms of association between those constructs and intelligence differences? Conclusion

The coming together of working memory and intelligence differences exemplifies the possibilities of cognitive and differential psychologists’ combining forces, something that has been urged many times yet failed fully to materialize: in 1957 by Cronbach55, in 1964 by McNemar56, in 1967 by Eysenck26, in 1978 by Carroll57, Glaser and Pellegrino58, and R.J. Sternberg59, in 1995 by Eysenck60, and in 1996 by Kyllonen12. In 1998, Carroll was still urging the union: ‘Obviously, what we need at the present time is a full-scale investigation of possible microlevel cognitive processes and the ways in which they affect performance on more traditional types of tests in various cognitive domains.’ (Ref. 61, p. 20.)

In 1999, Gustafsson wrote, ‘During the past couple of decades, research conducted along experimental lines has contributed greatly to our understanding of the nature of G [general intelligence], both in terms of low-level reductionistic models and in terms of higher level cognitive models. There is, however, still a long way to go until the major individual-differences constructs have been adequately accounted for in theoretical terms. It would seem, however, that an approach combining experimental methodology with multivariate modeling techniques may prove useful in researchers’ future attempts to achieve such an understanding.’ (Ref. 62, p. 287.)

Cognitive scientists with a credible cognitive architecture that contains isolable, testable parameters can expect a warm welcome from differential psychologists if they wish to study individual differences in intelligence.

170

Review

References 1 Deary, I.J. (2001) Human intelligence differences: a recent history. Trends Cognit. Sci. 4, 127–130 2 Deary, I.J. and Caryl, P.G. (1997) Neuroscience and human intelligence differences. Trends Neurosci. 20, 365–371 3 Deary, I.J. (2000) Looking Down on Human Intelligence: From Psychometrics to the Brain, Oxford University Press 4 Sternberg, R.J. (1977) Component processing in analogical reasoning. Psychol. Rev. 84, 353–378 5 Spearman, C. (1923) The Nature of Intelligence and the Principles of Cognition, Macmillan 6 Gustafsson, J-E. (1992) The relevance of factor analysis for the study of group differences. Multivariate Behav. Res. 27, 239–247 7 Sternberg, R.J. and Gardner, M.K. (1983) Journal of Experimental Psychology: General, 112, 80. 8 Lohman, D.F., Complex information processing and intelligence. In Handbook of Intelligence (Sternberg, R.J., ed.), pp. 285–340, Cambridge University Press 9 Embretson, S.E. (1995) The role of working memory capacity and general control processes in intelligence. Intelligence 20, 169–189 10 Embretson, S.E. and Schmidt-McCollam, K.M. (2000) Psychometric approaches to understanding and measuring intelligence. In Handbook of Intelligence (Sternberg, R.J., ed.), pp. 423–444, Cambridge University Press 11 Kyllonen, P.C. (1996). Is working memory capacity Spearman’s g? In Human Abilities: Their Nature and Measurement (Dennis, I. and Tapsfield, P., eds), pp. 77–96, Erlbaum 12 Kyllonen, P.C. (1996) Aptitude testing inspired by information processing: a test of the four sources model. J. Gen. Psychol 120, 375–405 13 Carpenter, P.A. et al. (1990) What one intelligence test measures: a theoretical account of processing in the Raven’s Progressive Matrices Test. Psychol. Rev. 97, 404–431 14 Duncan, J. et al. (1996) Intelligence and the frontal lobe: the organization of goal-directed behavior. Cognit. Psychol. 30, 257–303 15 Engle, R.W. et al. (1999) Working memory, short-term memory, and general fluid intelligence: a latent variable approach. J. Exp. Psychol. Gen. 128, 309–331 16 Jensen, A.R. (1987). Individual differences in the Hick paradigm. In Speed of Information Processing and Intelligence (Vernon, P.A., ed.), pp. 101–175, Ablex 17 Deary, I.J. et al. Reaction times and intelligence differences: a population-based cohort study. Intelligence (in press) 18 Beck, L.F. (1933) The role of speed in intelligence. Psychol. Bull. 30, 169–178 19 Hick, W.E. (1952) On the rate of gain of information. Q. J. Exp. Psychol. 4, 11–26 20 Sternberg, S. (1966) High speed scanning in human memory. Science 153, 652–654 21 Posner, M.I. and Mitchell, R.F. (1967) Chronometric analysis of classification. Psychol. Rev. 74, 392–409 22 Neubauer, A.C. (1997) The mental speed approach to the assessment of intelligence. In Advances in Cognition and Education: Reflections on the Concept of Intelligence (Kingma, J. and Tomic, W., eds), pp. 149–173, JAI Press http://tics.trends.com

TRENDS in Cognitive Sciences Vol.5 No.4 April 2001

23 Merkel, J. Die zeitlichen Verhältnisse der Willenstätigkeit. Philos. Stud. 10, 499–506 24 Hyman, R. (1953) Stimulus information as a determinant of reaction time. J. Exp. Psychol. 45, 188–196 25 Roth, E. (1964) Die Geschwindigkeit der Verabeitung von Information und ihr Zusammenhang mit Intelligenz. Z. Exp. Angew. Psychol. 11, 616–622 26 Eysenck, H.J. (1967). Intelligence assessment: a theoretical and experimental approach. Br. J. Educ. Psychol. 37, 81–97 27 Jensen, A.R. (1998) The suppressed relationship between IQ and the reaction time slope parameter of the Hick function. Intelligence 26, 43–52 28 Jensen, A.R. (1998) The g Factor: The Science of Mental Ability, Praeger 29 Neubauer, A.C. et al. (2000) Genetic and environmental influences on two measures of speed of information processing and their relation to psychometric intelligence: evidence from the German Observational Study of Adult Twins. Intelligence 28, 267–289 30 Lohman, D.F. (1994) Component scores as residual variation (or why the intercept correlates best). Intelligence 19, 1–11 31 Lohman, D.F. (1999). Minding our p’s and q’s: on finding relationships between learning and intelligence. In Learning and Individual Differences: Process, Trait, and Content Determinants (Ackerman, P.L. et al., eds), pp. 55–76, American Psychological Association 32 Vickers, D. et al. (1972) Perceptual indices of performance: the measurement of ‘inspection time’ and ‘noise’ in the visual system. Perception 1, 263–295 33 Nettelbeck, T. and Lally, M. (1976) Inspection time and measured intelligence. Br. J. Psychol. 67, 17–22 34 Nettelbeck, T. (1987) Inspection time and intelligence. In Speed of Information Processing and Intelligence (Vernon, P.A., ed.), pp. 295–346, Ablex 35 Kranzler, J.H. and Jensen, A.R. (1989) Inspection time and intelligence: a meta-analysis. Intelligence 13, 329–347 36 Deary, I.J. and Stough, C. (1996) Intelligence and inspection time: achievements, prospects and problems. Am. Psychol. 51, 599–608 37 Grudnik, J.L. and Kranzler, J.H. Meta-analysis of the relationship between intelligence and inspection time. Intelligence (in press) 38 Crawford, J.R. et al. (1998) Evaluating competing models of the relationship between inspection time and psychometric intelligence. Intelligence 26, 27–42 39 Burns, N.R. et al. (1999) Inspection time correlates with general speed of processing but not with fluid ability. Intelligence 27, 37–44 40 Egan, V. (1994) Intelligence, inspection time and cognitive strategies. Br. J. Psychol. 85, 305–316 41 Stough, C. et al. (1996) The relationship between intelligence, personality and inspection time. Br. J. Psychol. 81 255–268 42 Chaiken, S.R. (1993) Two models for an inspection time paradigm: processing distraction and processing speed versus processing speed and asymptotic strength. Intelligence 17, 257–283

43 Bors, D.A. et al. (1999). Inspection time and intelligence: practice strategies and attention. Intelligence 27, 111–129 44 Caryl, P.G. (1994) Event-related potentials correlate with inspection time and intelligence. Intelligence 18, 15–46 45 Thompson, J.C. et al. (2000). Effects of the nicotinic antagonist mecamylamine on inspection time. Psychopharmacology 150, 117–119 46 McCrimmon, R.J. et al. (1996). Visual information processing during controlled hypoglycaemia in humans. Brain 119, 1277–1287 47 Stough, C. et al. Examining neurochemical determinants of inspection time: development of a biological model. Intelligence (in press) 48 Deary, I.J. (1995) Auditory inspection time and intelligence: what is the direction of causation? Dev. Psychol. 31, 237–250 49 Parker, D.M. et al. (1999) Auditory inspection time and intelligence: a new spatial localization task. Intelligence 27, 131–139 50 Olsson, H. et al. (1998) Auditory inspection time: on the importance of selecting the appropriate sensory continuum. Pers. Indiv. Diff. 25, 627–634 51 Nettelbeck, T. and Rabbitt, P.M.A. (1992) Age, intelligence, and speed. Intelligence 16, 189–205 52 Stankov, L. and Roberts, R.D. (1997) Mental speed is not the ‘basic’ process of intelligence. Pers. Indiv. Diff. 22, 69–84 53 White, M. (1996). Interpreting inspection time as a measure of the speed of sensory processing. Pers. Indiv. Diff. 20, 351–363 54 Dennett, D.C. (1995). Darwin’s Dangerous Idea: Evolution and the Meanings of Life, Allen Lane 55 Cronbach, L.J. (1957) The two disciplines of scientific psychology. Am. Psychol. 12, 671–684 56 McNemar, Q. (1964) Lost: our intelligence? Why? Am. Psychol. 19, 871–882 57 Carroll, J.B. (1978) How shall we study individual differences in cognitive abilities? Methodological and theoretical problems. Intelligence 2, 87–115 58 Glaser, R. and Pellegrino, J.W. (1978) Uniting process theory and differential psychology: back home from the wars. Intelligence 2, 305–319 59 Sternberg, R.J. (1978) Intelligence research at the interface between differential and cognitive psychology: prospects and proposals. Intelligence 2, 195–222 60 Eysenck, H.J. (1995) Can we study intelligence using the experimental method? Intelligence 20, 217–228 61 Carroll, J.B. (1998). Human cognitive abilities: a critique. In Human Cognitive Abilities in Theory and Practice (McArdle, J.J. and Woodcock, R.W., eds), pp. 5–23, Erlbaum 62 Gustafsson, J-E. (1999) Measuring and understanding G: experimental and correlational approaches. In Learning and Individual Differences: Process, Trait, and Content Determinants (Ackerman, P.L. et al. eds), pp. 275–289, American Psychological Association 63 Deary, I.J. et al. (1997) Visual information processing and intelligence. Intelligence 24, 461–479