This is the current version of my previous "Beyond CHC theory" module. It presents my current thinking [based on extensive exploratory and confirmatory analysis of multiple data sets (esp. the WJ III norm data and WJ III joint cross-battery data sets) plus the integration of contemporary cognitive, neurocognitive, intelligence and neuropsychological research] re: potential future evolutions of the Cattell-Horn-Carroll (CHC) model of human cognitive abilities. This current presentation was last presented at the CNN (neuropsych) conference the first week of October, 2010, in Fremantle Australia
WJ IV Battery: Select Technical and Psychometric Information Overview
Pushing the edge of the contemporary cognitive (CHC) theory: New directions for psychologists
1. Pushing the edge of the contemporary cognitive (CHC) theory: New directions for psychologists Kevin S. McGrew, PhD Woodcock-Muñoz Foundation 16th Annual APS College of Clinical Neuropsychologists Conference From East to West: New directions in Neuropsychology 30 September - 2 October 2010 Notre Dame University, Fremantle, Western Australia
2. Or…..what an inquisitive applied intelligence scholar/psychometrician constructed/discovered from playing almost a decade in his data, literature, and theoretical sandbox
3. “Intelligent” testing and interpretation requires…knowing thy instruments Neuropsych. interpretation Error variance (reliability) External criterion relevance Uniqueness (specificity) g loading Degree of cognitive complexity CHC Ability factor classifications Degree of cultural loading Degree of linguistic demand Metric scale Information processing & stimulus/response characteristics Ability domain cohesion
4. “ If you give a monkey a stradivarius violin and you get bad music……..you don’t blame the violin” McGrew (circa 1986)
5. Three things (or major steps) completed that have resulted in the intelligence model(s) to be presented today
6. Things 1 and 2: Will be covered quickly to provide context and background for primary content of today – Thing 3
7. Psychometric vs. neuropsychological conception/model assessment gap “It is notable that there is a gap between neuropsychological measures and evolving conceptualizations of intelligence. That is, for as seemingly related as the instruments and concepts are, they have strikingly different historical backgrounds.” (Hoelzle, 2008)
8.
9. NP assessment has been traditionally non-theoretical---popular models of intelligence and cognitive abilities have been derived via statistical procedures
10.
11. Horizontalmultiple regression (aptitude/functional/pragmatic) model Criterion DVs Gf Gc Glr G.. Gsm Gv etc Attn TBI ? Brain Area/function Neuropsychological approaches have had primary (but not sole) focus/goal on external/predictive (Dx) validity – Horizontal models Result has been many NP measures are mixture measures of multiple CHC domain abilities (which abilities and in what amount [weighting] best predict criterion variables?)
12. My primary goal Present a different (yet compatible value-added) psychometric theory of intelligence perspective for thinking about testing cognitive abilities
13. Importance Of Classification Taxonomies In All Sciences Classification is arguably one of the most central and generic of all our conceptual exercises…without classification, there could be no advanced conceptualization, reasoning, language, data analysis, or for that matter, social science research (K.D. Bailey, 1994). A specialized science of classification of empirical entities known astaxonomy(Bailey, 1994; Prentky, 1994) is ubiquitous in all fields of study because it guides our search for information or truth.
14. ? Reliable variance (reliability) Error variance -individual/situational variables (e.g., distractibility) -item variables (e.g., item sampling and item gradients; test floor and ceiling) -examiner variables (e.g., rapport, scoring and administration errors) -testing environment variables (e.g., noise, comfort) Unique abilitiesnot shared in common with other CHC factor indicators (specificity) We have been searching for an empirically/theoretically-based cognitive taxonomyto interpret the reliable variance of cognitive tests
15. The Cattell-Horn-Carroll (CHC) theory of cognitive abilities is the contemporary consensus psychometric model of the structure of human intelligence The CHC Timeline Project (and detailed information re: CHC theory/model)can be found at IQ’s Corner blog www.iqscorner.com
16. g T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 PMA1 PMA2 PMA3 PMA4 PMA1 PMA2 PMA3 PMA4 PMA1 PMA2 PMA3 PMA4 PMA1 PMA2 PMA3 PMA4 …etc (1b) Thurston’s Multiple Factor (Primary Mental Abilities) Model …etc (1a) Spearman’s general Factor model G1 G2 G3 …etc …etc …etc g ? …etc G1 G2 G3 …etc …etc (1e) Consensus Cattell-Horn-Carroll Hierarchical Three-Stratum Model Arrows from g to each test (rectangle) have been omitted for readability Stratum III g G1 Stratum II G2 …etc Stratum I …etc …etc (1d) Carroll’s Schmid-Leiman Hierarchical Three-Stratum Model (1c) Cattell-Horn Gf-Gc Hierarchical Model Stratum III Note: Circles represent latent factors. Squares represent manifest measures (tests; T1..). Single-headed path arrows designate factor loadings. Double headed arrows designate latent factor correlations Stratum II Stratum I Figure 1: Major stages in the evolution of psychometric theories from Spearman’s g to Cattell-Horn-Carroll (CHC) theory
17. CHC theory has entered the mainstream neuropsychological assessment literature
18. A landmark event in understanding the structure of human cognitive abilities - 1993
19. THE SCOPE OF CARROLL’S FACTOR ANALYTIC REVIEW Reviewed factor analytic research of the past 50-60 years Includes nearly all of the more important and classic factor analytic investigations Started with 1,500 references Final pool of 461 data sets that meet specific criteria Reanalyzed all or nearly all of the data sets Used exploratory methods in order to “let the data speak for themselves”
20. The verdict is unanimous re: the importance of Carroll’s (1993) work Richard Snow (1993): “John Carroll has done a magnificent thing. He has reviewed and reanalyzed the world’s literature on individual differences in cognitive abilities…no one else could have done it… it defines the taxonomy of cognitive differential psychology for many years to come.” Burns (1994): Carroll’s book “is simply the finest work of research and scholarship I have read and is destined to be the classic study and referencework on human abilities for decades to come” (p. 35). John Horn (1998): A “tour de force summary and integration” that is the “definitive foundation for current theory” (p. 58). Horn compared Carroll’s summary to “Mendelyev’s first presentation of a periodic table of elements in chemistry” (p. 58). Arthur Jensen (2004): “…on my first reading this tome, in 1993, I was reminded of the conductor Hans von Bülow’s exclamation on first reading the full orchestral score of Wagner’s Die Meistersinger, ‘‘It’s impossible, but there it is!’’ “Carroll’s magnum opus thus distills and synthesizes the results of a century of factor analyses of mental tests. It is virtually the grand finale of the era of psychometric description and taxonomy of human cognitive abilities. It is unlikely that his monumental feat will ever be attempted again by anyone, or that it could be much improved on. It will long be the key reference point and a solid foundation for the explanatory era of differential psychology that we now see burgeoning in genetics and the brain sciences” (p. 5).
21. Contemporary psychometric research has converged on the Cattell-Horn-Carroll (CHC) theory of cognitive abilities as the consensusworking taxonomy of human intelligence McGrew, K. (2009). Editorial: CHC theory and the human cognitive abilities project: Standing on the shoulders of the giants of psychometric intelligence research, Intelligence, 37, 1-10.
22. T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 PMA1 PMA2 PMA3 PMA4 g ? …etc G1 G2 G3 …etc …etc (1e) Consensus Cattell-Horn-Carroll Hierarchical Three-Stratum Model CHC as the consensus psychometric model of intelligence Because the Carroll model is largely consistent with the model originally proposed by Cattell (1971), McGrew (2009) has proposed an integration of the two models which he calls the Cattell-Horn-Carroll (C-H-C) Integration model….Because of the inclusiveness of this model, it is becoming the standard typology for human ability. It is certainly the culmination of exploratory factor analysis. The Science of Intelligence (Doug Detterman, 2010; book manuscript in preparation)
23. T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 PMA1 PMA2 PMA3 PMA4 g ? …etc G1 G2 G3 …etc …etc (1e) Consensus Cattell-Horn-Carroll Hierarchical Three-Stratum Model CHC as the consensus psychometric model of intelligence “The Cattell–Horn–Carroll (CHC) theory of cognitive abilities is the best validated model of human cognitive abilities” [Ackerman, P. L. & Lohman D. F. (2006). Individual differences in cognitive functions. In P. A. Alexander, P. Winne (Eds.), Handbook of educational psychology, 2nd edition (pp. 139-161). Mahwah, NJ: Erlbaum.]
24. T2 T3 T4 T5 T6 T7 T8 T9 T1 T12 T10 T11 PMA1 PMA2 PMA3 PMA4 g ? …etc G1 G2 G3 …etc …etc (1e) Consensus Cattell-Horn-Carroll Hierarchical Three-Stratum Model CHC as the consensus psychometric model of intelligence A significant number of Australian intelligence scholars have framed (and/or continue to frame) their research as per the extended Gf-Gc (aka. CHC) model of intelligence. Many have made foundational contributions to building the model. N. R. Burns T. Nettlebeck L. Stankov R. Roberts S. Bowden
25. Importance Of Classification Taxonomies In All Sciences Classification is arguably one of the most central and generic of all our conceptual exercises…without classification, there could be no advanced conceptualization, reasoning, language, data analysis, or for that matter, social science research (K.D. Bailey, 1994). A specialized science of classification of empirical entities known astaxonomy(Bailey, 1994; Prentky, 1994) is ubiquitous in all fields of study because it guides our search for information or truth.
26. Gf Broad RG RP Narrow I RQ RE CHC theory classifies abilities according to three levels or strata g All CHC narrow abilities and their definitions can be found at www.IAPsych.com General RG = Gen Sequential (deductive) Reasoning I = Induction RQ = Quantitative Reasoning RP = Piagetian Reasoning RE = Speed of Reasoning
27.
28. ...most disciplines have a common set of terms and definitions (i.e., a standard nomenclature) that facilitates communication among professionals and guards against misinterpretations. In chemistry, this standard nomenclature is reflected in the ‘Table of Periodic Elements’. Carroll (1993a) has provided an analogous table for intelligence….. (Flanagan & McGrew, 1998)
29. g Gv Gf Glr Gs Gsm Gc Ga Reliable variance (reliability) Error variance -individual/situational variables (e.g., distractibility) -item variables (e.g., item sampling and item gradients; test floor and ceiling) -examiner variables (e.g., rapport, scoring and administration errors) -testing environment variables (e.g., noise, comfort) Unique abilitiesnot shared in common with other CHC factor indicators (specificity) CHC Theory is the best available empirically and theoretically sound cognitive ability taxonomy available today
30. This is where the field of psychometric intellectual assessment is at..and a bandwagon has formed g Gf Gc Induction (I) Lang. Develpmt (LD) (LD) General Seq. Reasoning (RG) Quantitative Reasoning (RQ) Listening Ability (LS) General Information (K0) Lexical Knowledge (VL) Speed of Reasoning (RE) Primary ability Reliable variance (reliability) Secondary ability Error variance -individual/situational variables (e.g., distractibility) -item variables (e.g., item sampling and item gradients; test floor and ceiling) -examiner variables (e.g., rapport, scoring and administration errors) -testing environment variables (e.g., noise, comfort) Unique abilitiesnot shared in common with other CHC factor indicators (specificity)
31. Published WJ III CHC model (McGrew & Woodcock, 2001 CFA analysis of 50+ cognitive and achievement tests
32. Starting point Ages 6-adult CFA Broad CHC Model in WJ III Technical Manual (McGrew & Woodcock, 2001) g .55 .91 .88 .73 .87 .88 .79 .82 .93 Gs Gsm Gv Gc Gq Ga Glr Gf Grw First order measurement model omitted for readability purposes
33. Deconstruction: The validated/published WJ III CHC structure was “torn down” Psychologists need a healthy degree of positive skepticism
34. Reconstruction: New structural models specified based on insights from large variety of statistical analysis of the WJ III norm data since 2001.
35. Stage (Thing 2) approach Theoretical considerations (Berlin BIS model; dual-processing cognitive models; etc.) also served as guides during exploratory model specification. Important caution: The final models demonstrated near identical model fit statistics (e.g., some equivalent models). Also, the large amount of exploratory model specification employed has the potential to capitalize on "random chance factors"- thus rendering statistical model evaluation comparisons useless. Thegoal of these analyses were to "push the edge of the envelope" of the WJ three data via SEM-based model generation procedures. Thelaw of parsimony was deliberately discarded. Cross validation of proposed final models in independent samples is needed.
42. Cluster Analysis Cluster analysis is an set of exploratory (structure discovering) data analysis tools for solving classification problems. Sometimes it has been called a “poor mans” factor analysis. Its object is to sort cases (people, things, events, tests, etc) into groups, or clusters, so that the degree of association is strong between members of the same cluster and weak between members of different clusters. Each cluster thus describes, in terms of the data collected, the class to which its members belong; and this description may be abstracted through use from the particular to the general class or type. CA often helps confirm EFA results and similar to MDS (multidimensional scaling), can spatially represent the degree of similarity of tests measuring a common dimension (dimension cohesion). Its hierarchical sequential structure is often useful in suggesting higher-order dimensions/factors.
43. Cluster Analysis The strength of cluster analysis (discovering structure in data with more relaxed statistical assumptions and mathematics than data reduction methods such as exploratory factor analysis) is also one of its major limitations. CA will find groups or clusters in random data. The algorithms are designed to find any structure, even if structure is not present. As a result, the later clusters in a hierarchical approach are often “necessary evils or by products”--CA must end with one grand cluster. Thus, often in CA a point is reached where the further collapsing of meaningful groupings ceases to make substantive sense. It is important to recognize this in the resultant cluster dendogram. Also, given the above, tests (objects, etc.) that share little in common with other measures need to be assigned to some grouping and cluster. Thus, often “loner” type tests will appear in very meaningful clusters but will not be consistent with the underlying interpretation of the grouping/cluster. Sometimes this suggests new insights regarding the test. Other times these “I’ve got to be grouped with some cluster somewhere in the process” tests are best ignored and should not interpreted as discounting the strong communality of a grouping or clustering of tests
44. AS CF SA UD Gf (language-based) AP Clusters beyond this point not easily interpretable – see limitations of CA method NS Complex lang. processing/ reasoning? QC NM Gf CAL RQ MF Gf (numeric-based) RDF WF ? = no apparent current CHC ability classification Red font = CHC factors Blue font = possible new abilities at different strata to consider PSC RV Gs (achievement) ED ? LW SP RC WS SOS Orthographic processing? Grw (words, sent, con. disc) WA PV VC Grw GI Grw (phonemes) AK OC LD/VL STR DRS K0 RPN ? REF Gc DS LS PC VM NA/R4 (RAN?) CO Gs (cognitive) VCL R9 PR BR Ppr SPR Pc SNP PLN MV NR Gv Cluster analysis (Wards method) of 50 WJ III cognitive and achievement tests (ages 6-18; NU norms) Kevin McGrew 11-13-09 AWM SR/Vz MS ? MW SR/Vz+ IW SB MW AA Gsm DRV MS VAL Temporal Processing or Tracking / Aud. Sequential Processing MN PC DRM Ga Glr-MA Distances 2.5 0.5 1.0 1.5 2.0 0.0
46. WAIS-IV test Cluster Tree (Wards method) of WAIS-IV subtest intercorrelations Verbal know & comp (Gc) IN CO VC Level (unspeeded) cognitive abilities SI Short-term & working memory (Gsm) LN DS AR Fluid Reasoning (Gf) FW MR Visual-Spatial Proc.(Gv) BD VP General Intelligence (g) as per WAIS-IV ? PCM CD Processing Speed (Gs) (rate cognitive abilities) SS CA 0.0 0.5 1.0 1.5 Distances
58. Differences in type of processesUsesMDS (multidimensional scaling)
59. Example of MDS (Radex Model) The closer a test is to the center of the figure, the more it is related to the underlying general dimension of the battery. Also, the center represents the most cognitively complex (i.e., have the largest number of performance components) tests. Tests that group together are interpreted as sharing common stimulus content or cognitive processing characteristics
60. MDS (Guttman Radex model) of WAIS-IV subtest intercorrelations 3 Short-term memory /working memory (Gsm) 1 Processing speed (Gs) LN DS CD Verbal know & comp (Gc) VC CO Dimension-2 Fluid reasoning (Gf) AR MR CA SS SI IN FW BD VP -1 PCM Visual-spatial processing (Gv) -3 -3 -1 1 3 Dimension-1
61. MDS (Guttman Radex model) of WAIS-IV subtest intercorrelations It is a common practice in MDS analysis to visually partition the MDS spatial configuration into broader dimensions and consider interpretation at a higher-order level. The current WAIS-IV MDS revealed the following hypothesized higher-order structure Note – similar to hand rotation of factors in early days of EFA, K. McGrew took the cross-hair lines and hand rotated them (simultaneosly) until a meaningful pattern emerged. The four-broad dimensions are interpreted as being very similar to the four cognitive domains of Woodcock’s Cognitive Performance Model (CPM) – see next two slides Short-term memory /working memory (Gsm) – Cognitive Efficiency unspeeded/memory 3 Verbal know & comp (Gc) – Acquired Knowledge or “Product” dominant abilities? 1 LN DS CD VC AR CA CO MR SS Processing speed (Gs) - Cognitive Efficiency speeded SI IN FW BD VP -1 PCM Fluid Reasoning (Gf) and Visual-spatial processing (Gv) –Thinking or “Process” dominant abilities? -3 -3 -1 1 3
71. “Intelligent” testing and interpretation requires…knowing thy instruments Neuropsych. interpretation Error variance (reliability) External criterion relevance Uniqueness (specificity) g loading Degree of cognitive complexity CHC Ability factor classifications Degree of cultural loading Degree of linguistic demand Metric scale Information processing & stimulus/response characteristics Ability domain cohesion
72. Food for thought: Are the MDS quadrants or partitions reflecting content “facets” or a combination of content“facets and “operations” as per the BIS model of intelligence….see next slide
73. BIS: Berlin Model of Intelligence Structure Gs Gsm + Glr (level abilities) Carroll’s Gy Glr(fluency abilities) Gf Note difference in term in different versions: Processing capacity defined as complex reasoning
75. Alternative Model 1 g .39 Gs (Cognitive speed) .82 .88 .71 .87 .86 .79 .84 1.0 .64 .55 .62 .59 .36 .49 Gs (Grw) .54 Gsm Gv Gs (Gv) Gs (Gq) Gs (Gc) .62 Gc Gq Ga Glr Gf Grw First order measurement model and other lower-order latent factors (below smallest oval latent factors) omitted for readability purposes. Thicker path arrow with bold font 1.0 parameter designates path that had to be constrained (fixed) to 1.0
76. Alternative Model 2 .86 .99 .93 g .36 1.0 Cognitive efficiency (More automatic & effortless) Cog. knowledge domains/systems (product/content abilities) Lang/linguistic./symbolic abilities Cognitive operations (process/operations/analytic/rule-based abilities) figural-spatial, lower-linguistic abilities Gs (Cognitive speed) .89 .76 .91 .85 1.0 .83 .82 .64 .52 .60 .58 .41 .52 Gs (Grw) .58 Gsm Gv Gs (Gv) Gs (Gq) Gs (Gc) .67 Gc Gq Ga Glr Gf Grw First order measurement model and other lower-order latent factors (below smallest oval latent factors) omitted for readability purposes. Thicker path arrow with bold font 1.0 parameter designates path that had to be constrained (fixed) to 1.0
77.
78.
79.
80. Close inspection of the evidence suggests that generic dual-system theory is currently oversimplified andmisleading We might be better off talking about type 1 and type 2 processes since all theories seem to contrast fast, automatic, or unconscious processes with those that are slow, effortful, and conscious (Samuels 2006). Such terminology does not commit use to a two-system view. However, it would then be helpful to have some clear basis for this distinction My suggestion is that type 2 processes are those that require access to a single, capacity-limited central working memory, while type 1 processes do notrequire such access. This implies that the core features of type 2 processes are that they are slow, sequential, and capacity limited. The last feature implies also that their functioning will correlate with individual differences in cognitive capacity and be disrupted by concurrent working memory load. Depending upon what else is assumed about working memory, there may be a rationale for describing such type 2 processes as registering in consciousness and having properties associated with executive processes and intentional, higher-order control.
81. Alternative Model 2b .93 .99 1.0 .86 g .36 1.0 Type I cognitive processing (Cognitive efficiency): More automatic & effortless Cog. knowledge domains/systems (product/content abilities) Lang/linguistic./symbolic abilities Type II cognitive processing: More cognitively controlled & deliberate Cognitive operations (process/operations/analytic/rule-based abilities) figural-spatial, lower-linguistic abilities Gs (Cognitive speed) .89 .76 .91 .85 1.0 .83 .82 .64 .50 .60 .58 .41 .52 Gs (Grw) .58 Gsm Gv Gs (Gv) Gs (Gq) Gs (Gc) .67 Gc Gq Ga Glr Gf Grw First order measurement model and other lower-order latent factors (below smallest oval latent factors) omitted for readability purposes. Thicker path arrow with bold font 1.0 parameter designates path that had to be constrained (fixed) to 1.0
82. Alternative Model 3 .95 .99 .94 .21 g Auditory temporal (serial) Proc. Visual/figural (parallel?) Proc. Cog. knowledge domains/systems Gs (Cognitive speed) .89 .77 .91 .85 1.0 .82 .86 .90 .63 .45 .60 .54 .48 .65 Gs (Grw) .64 Gsm Ga Gs (Gv) Gs (Gq) Gs (Gc) .73 Gc Gq Gv Glr Gf Grw First order measurement model and other lower-order latent factors (below smallest oval latent factors) omitted for readability purposes. Thicker path arrow with bold font 1.0 parameter designates path that had to be constrained (fixed) to 1.0
83. Alternative Model 3b 1.0 .94 .21 Cognitive operations (process/operations/ analytic/rule-based abilities) g .95 1.0 Auditory temporal (serial) Proc. Visual/figural (parallel?) Proc. Cog. knowledge domains/systems Gs (Cognitive speed) .89 .77 .91 .85 1.0 .82 .86 .90 .63 .45 .60 .54 .48 .66 Gs (Grw) .64 Gsm Ga Gs (Gv) Gs (Gq) Gs (Gc) .74 Gc Gq Gv Glr Gf Grw First order measurement model and other lower-order latent factors (below smallest oval latent factors) omitted for readability purposes. Thicker path arrow with bold font 1.0 parameter designates path that had to be constrained (fixed) to 1.0
84. Pushing the edge of the envelope of CHC theory and the WJ III measurement model: Part IIThe first-order measurement model and implications for interpretation of WJ III tests
85. Glr and Gsm measurement models were similar to those originally reported by McGrew & Woodcock (2001)
86. See next slide for other indicators Vis. Clos. (.41) Blk. Rot. (.52) Spat. Rel. (.66) Pic. Rec (.43) Planning (.43) Alternative Models: WJ III Measurement model for speed factors Gq Gv .54 Gs(Gq) .36 .62 Gs(Gv) GGs (Cog Spd) .64 .55 Gs(Gc) .59 Gc .49 Gs(Grw) See next slide for other indicators Grw .62 Wrd. Atk. (.78) Edit. (.78) Psg. Cmp.* (.55) Wrt. Smp, (.76) Rdg. Voc.* (.34) Spelling (.86) LWrdID (.89) * Dual loading on Gc on next slide
87. Alternative Models: WJ III Measurement model for possible new Gf factor structure Calculation (.75) Gq .34 .51 .27 Gf (RQ) .66 .17 Gf .99 Gf * .70 Gc Gen. Info .(.89) Acd. Knw. (.89) Orl. Cmp. (.77) Psg. Cmp. (.30) (.55-Grw) Rdg. Voc. (.54) (.34-Grw) Mem. Sen. (.36) (.38- Gsm) Story Rec. (.29) (.39-Glr) Sound Awareness and Understanding Directions did not load on any other factors Gf * = complex language-based working memory and reasoning?
88. Iteration 1: CHC-based Intelligence model of WJ III battery Kevin McGrew 8-18-2010 See handouts for clear copy
90. It is time to look at some non-CHC/Gf-Gc research on reasoning (Gf): Alternative lenses
91. The distinction between inductive and deductive reasoning (i.e., CHC/Gf-Gc Carroll-type model) may be outdated (Wilhelm, 2005) Most established reasoning tests confound the direction of inference with deductive and inductive reasoning task (Whilhelm, 2005)
92. Whilhelm tested Gf model’s as per CHC (I, RQ, RG) and BIS (verbal, quant, figural) structures, and various model interactions. The following was the best fitting model
93.
94. CFA using dual indicators (split-half—odd/even item sets) for each test: Conclusion: WJ III RG, I, RQ tests are highly correlated but do measure different aspects of Gf
95. WJ III CHC Gf model Fit for this and prior model (prior slide) more-or-less equivalent
98. Important Reminder: All statistical methods, such as factor analysis (EFA or CFA) have limitations and constraints. It only provides evidence of structural/internal validity and typically nothing about external, developmental, heritability, neurocognitive validity evidence Need to examine other sources of evidence and use other methods – looking/thinking outside the factor analysis box
99.
100. Under. Directions.#/quant. Lang (aud-verbal) 1 Calc ApPrb AcdKn NumMat PicVoc GenInf NumSer SndAwr Reasoning (procedural/Gf)-------Recall (declarative/Gc/Gq)) OrlVoc UndDir AnlSyn VrbAnl -1 Visual-figural ConFrm -3 -3 -1 1 3 Language (verbal/aud.)--------Nonverbal (#’s,visual) Guttman Radex MDS model of WJ III Gf, Gc, and Gq test indicators
101. Thing 3 – attempt to integrate Thing 1 and Thing 2 with neuropsychological assessment models
102. The First Commandment of Neuropsychological Assessment "If one writes a book on neuropsychological assessment, thou shall not write a book that is less than 3 inches thick or less than 3 lbs in weight“ (McGrew, August 13, 2010)
103. Lets look at the pieces one by one – blow them up
104. Arm-chair factor analysis of neuropsych. assessment domains [and CHC construct mapping] (K. McGrew; 8-18-10) [I of 3] g Gf Gc Grw Gq
105. Arm-chair factor analysis of neuropsych. assessment domains [and CHC construct mapping] (K. McGrew; 8-18-10) [I of 3] Gv Ga Gsm Glr
106. Arm-chair factor analysis of neuropsych. assessment domains [and CHC construct mapping] (K. McGrew; 8-18-10) [I of 3] Gs Gsm AC ?? Gp Gps Go Gh Gk
108. Mapping of current CHC domains with hypothesized new CHC-based intelligence model Kevin McGrew 8-18-2010 Lets look at the pieces one by one – blow them up Motor functions (including motor speed) - Expressive across domains?
132. This is research/work in progress: Suggested research that needs to be explored and integrated. Go from here to……………..
133.
134. “Intelligent” testing and interpretation requires…knowing thy instruments Neuropsych. interpretation Error variance (reliability) External criterion relevance Uniqueness (specificity) g loading Degree of cognitive complexity CHC Ability factor classifications Degree of cultural loading Degree of linguistic demand Metric scale Information processing & stimulus/response characteristics Ability domain cohesion
135. This is NOT a model of human functioning – it is a “working” heuristic of Kevin McGrew’s current hypothesized thinking (iteration 3?) regarding the important dimensions that may be important in the development and interpretation of measures of human abilities …………. (not a Guilford SOI model where all cells are believed to exist) Content/stimulus dimension Language (aud.-verb.) Numerical/quant. Somatasensory Visual-figural Olfactory ?: Is the low-how cog. complexity continuum simply a continuous representation of the Type 1/I processing distinction ? Cognitive knowledge domains/systems Cognitive operations Type II Processing Cognitive control High Abilty domain dimension Cognitive efficiency Sensory functions Low Type I Processing Motor functions Cognitve complexity dimension Note: CHC taxonomy is embedded in the ability domain dimension (see prior slides)
157. Aiming (AI)[ Narrow P abilities suggested by Ackerman et al. (2002) ] * Carroll classified P and R9 as narrow abilities under Gs/Gv and Gt, respectively ** Classified as speed and level (Gf) ability by Carroll *** Classified as a speed and level (Gc) ability by Carroll Also classified under Grw by the current author **** Classified as Psychomotor Ability by Carroll. Also classified under Grw by current author Figure 2: Hypothesized speed hierarchy based on integration of Carroll (1993) speed abilities with recent research (Ackerman, Beier & Boyle, 2002; O’Connor & Burns, 2003; McGrew & Woodcock, 2001; Roberts & Stankov, 1998; Stankov, 2000; Stankov & Roberts, 1997) Integrate proposed g-speed hierarchy (McGrew & Evans, 2004; McGrew, 2005)
158. Alternative Model 2b g – speed ? .93 .99 1.0 .86 g .36 1.0 Type I cognitive processing (Cognitive efficiency): More automatic & effortless Cog. knowledge domains/systems (product/content abilities) Lang/linguistic./symbolic abilities Type II cognitive processing: More cognitively controlled & deliberate Cognitive operations (process/operations/analytic/rule-based abilities) figural-spatial, lower-linguistic abilities Gs (Cognitive speed) .89 .76 .91 .85 1.0 .83 .82 .64 .50 .60 .58 .41 .52 Gs (Grw) .58 Gsm Gv Gs (Gv) Gs (Gq) Gs (Gc) .67 Gc Gq Ga Glr Gf Grw First order measurement model and other lower-order latent factors (below smallest oval latent factors) omitted for readability purposes. Thicker path arrow with bold font 1.0 parameter designates path that had to be constrained (fixed) to 1.0
170. Timescales of temporal processing (Mauk & Buonomano, 2004) Humans process temporal information over scales of at least 10-12 orders of magnitude that have been categorized into 3-4 major timescale groups
181. Integrate working model with temporal g (brain clock) research Temporal information processing models (Creelman, 1962; Gibbon, 1991; Rammsayer & Ulrich, 2001; Treisman et al., 1990; see Grondin, 2001 for review) are based on the central assumption ofneural oscilliations(note – same central feature of Jensen’s neural efficiency theory of g) as a major determinant of timing performance. The higher the frequency (higher speed) of neural oscillations the finer the temporal resolution of the internal clock = greater timing accuracy (Rammsayer & Brandler; 2007)
182. Temporal g ? Analyses suggested a unitary timing mechanism, referred to as temporal g. Performance on temporal information processing provided a more valid predictor of psychometric g than traditional reaction time measures r (with psychometric g) = .56 (temporal g) vs .34 (reaction time g) Findings suggest that temporal resolution capacity of the brain (as assessed with psychophysical temporal tasks) reflects aspects of neural efficiency associated with general intelligence. Rammsayer & Brandler (2007)