University of Southern California Dissertations and Theses

Cognitive aging: Expertise and fluid intelligence

(USC Thesis Other)

Cognitive aging: Expertise and fluid intelligence

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. UMI A Bell & Howell Information Company 300 North Zed) Road, Ann Arbor MI 48106-1346 USA 313/761-4700 800/521-0600 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. COGNITIVE AGING: EXPERTISE AND FLUID INTELLIGENCE ty Hiromi Takagi A Dissertation Presented to the Faculty of The Graduate School UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Education—Educational Psychology) May 1997 Copyright 1997 Hiromi Takagi R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. UMI Number: 9733148 Copyright 1997 by Takagi, Hiromi All rights reserved. UMI Microform 9733148 Copyright 1997, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation, written by ................................................................... under the direction of h er Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean of Graduate Studies Date I........ DISSERTATION COMMITTEE C o- Chairperson 'Co- ton R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. A cknow ledgem ent I have been extremely fortunate to have received the am ount of support I did during my four years at graduate school. I would first like to thank my committee members. I thank Dr. David Peterson for his continuing encouragement, for his thoughtfulness and for giving me abundant opportunities. Through my years as his assistant at the Leonard Davis School of Gerontology, my interest in aging has been cultivated and intensified. I thank Dr. Dennis Hocevar for his critical comments and precise advice on my dissertation. I enjoyed w orking as a teaching assistant for his statistics courses; I have learned a great deal from that invaluable experience. I thank Dr. Gretchen Guiton for her time, heartfelt care and unerring guidance in my personal life as well as in my academic pursuit. W ithout her loving support, I w ould have been lost while living in this country, which was so foreign to me at first. I am deeply thankful to my co-chairs, Dr. Richard Clark and Dr. John Horn. I thank Dr. Richard Clark for his continuing support and quick, accurate advice. As I attended his classes and personal conferences, his deep insight into hum an learning and cognition has kept inspiring me. His accuracy and efficient organization are a m odel by which to mold my life. I am deeply thankful to Dr. John Horn for his profound expertise in this field, which he has always willingly m ade accessible to me, his time, critical advice and comments, accurate guidance, continuing support, warm encouragement, understanding.... I do not even know how to put all that he has given me and done for me into words. His zest for life has taught me a valuable lesson on how deeply one could savor one's life. My thanks also go to the many people in Japan whom I have met through my studies and who have helped me realize my goals. 1 thank all of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. the staff members at the Japanese GO Association. W ithout their understanding for this study and both their physical and m ental support, I would still be wondering how I could obtain data for my research. I am especially grateful to one particular professional GO player, Mr. Nobuta, for his strong interest in this study and his assistance in making it so that this study would be supported by the Japanese GO Association. My special thanks also go to Mr. Sakamaki at the Japanese GO Association. His understanding, ability to access to a num ber of GO players and people in the news media, creative ways of thinking and efficacious adm inistration skills provided a solid foundation during the testing in the sum m er of 1996. In addition, I am grateful to a num ber of personal friends in the United Stated and Japan. In the United States, the creation and pilot studies on developed m easures were supported by a num ber of friends at school, in Japanese towns, and at GO clubs in Los Angeles. For the testing in Japan, Mr. Thomas Guiton helped me prepare beautiful slides of GO diagram s to be projected. In Japan, a friend gave me a great chance by introducing me to Mr. Nobuta. Six friends came to the headquarters of the Japanese GO Association on a hot day in August just to give assistance to my study. Mr. Kawahito and Mr. Mihashi at CBN (a slide projector company) not only gave me technical advice on slide projection, but also volunteered to make several trips to the testing place to deliver screens. A lot of GO players in my home town of Hannou gave me a chance to conduct further pilot studies on the newly developed measures, and a num ber of GO players from all over Japan volunteered to participate in this study and allowed me to realize this study. The Mental and physical support of my friends in both countries has continuously encouraged me. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. My deep thanks also go to my loving family in Japan. W ithout their consistent support, belief in me and continuous encouragem ent, I could not have been strong enough to keep confronting the big issue of pursuing my interests in this country. Although a great physical distance has separated them from me, I have always felt that I am surrounded by their warmth. When I close my eyes, I always see their warm smiling faces. And lastly, but not in the least, I would like to thank the United States of America. I do love the invaluable opportunities and challenges it has given me and the incredibly caring people-m entors and friends— it has brought to me. This study was partly supported by a grant from a Japanese foundation, the Organization of Planning for Seniors (Senior Plan Kaihatsu Kikou). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Table of Contents Acknowledgem ent List of Tables List of Figures Abstract Chapter 1: Review of the Literature Introduction Hum an Cognitive Abilities Fluid Intelligence (Gf) and Crystallized Intelligence (Gc) Nine Factors of Gf-Gc Theory Life-Span Developm ent of H um an Intelligence Within an Individual Interindividual Differences in Cognitive Change Vulnerable Cognitive Abilities SAR and Gs A ttentiveness Causes of Declines in Vulnerable Cognitive Abilities Expertise and Individual Differences in Cognitive Aging Development of Expertise Expertise and Short-term W orking Memory (Apprehension and Retrieval, SAR) Expertise and Cognitive Speed (Gs) Expertise and Attentiveness Sum m ary Overall Design of the Study Research Questions to be A ddressed Specific M ultiple-Group Design and Hypotheses A. Invariance of M easurem ent B. Averages between Groups C. Relationship between Age and Gf-Components Importance of the Study Chapter II: Method Subjects Variables R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. vi Page Within GO M easures 41 RECGO COMGO IDENTGO REPGO RESGO Measures from Previous Research 46 RECNO COMNO IDENTNO REPNO RESNO BackSpan PowLet Topology Short Q uestionnaire Procedure 51 Analyses 53 Chapter III: Results Part 1 Descriptive Statistics on Test Measures 56 Four Expertise Subgroups 57 Correlations 59 ANOVAs 65 Chapter IV: Results Part 2 Analyses with the Structural Equation Modeling using a Lisrel 76 SEM Stepl: Invariance of M easurement 78 SEM Step2: Averages between Groups 84 SEM Step3: Relationship between Age and Gf-Components 93 Chapter V: Conclusions 103 References 113 Appendix A Examples of Test Measures 121 Appendix B Correlations among Test Measures 146 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Tables List of Tables 1. GO Rank D istribution w ithin Whole Sample of Subjects 2. Descriptives of Test m easures am ong Whole Subjects 3. Mean Age and Mean GO Rank within Four Expertise Subgroups 4. Correlations between Age and GO Rank within Four Expertise Subgroups 5. Correlations of GO Tasks with GO Rank within Four Expertise Subgroups 6. Correlations of Non-GO Tasks w ith GO Rank within Four Expertise Subgroups 7. Correlations of Gf Measures w ith age within Four Expertise Subgroups 8. Intercorrelations among Gf M easures within Four Expertise Subgroups 9. Mean Scores and Standard Deviations of GO Tasks within Four Expertise Subgroups 10. ANOVA on M ean Differences of GO Tasks across Four Expertise Subgroups 11. Mean Scores and Standard Deviations of Non-GO Tasks within Four Expertise Subgroups 12. ANOVA on M ean Differences of Non-GO Tasks across Four Expertise Subgroups 13. Means of Residuals (after Variance of Age is Controlled for) of GO Tasks 14. ANOVA on Residuals (after Variance of Age is Controlled for) of GO Tasks R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. viii Tables Page 15. Means of Residuals (after Variance of Age is Controlled 72 for) of Non-GO T a s li 16. ANOVA on Residuals (after Variance of Age is Controlled 73 for) of Non-GO T aste 17. Descriptive Information (Means, Standard Deviations, 77 Skewness & Kurtosis) on 17 Measured Variables within Three Expertise Subgroups 18. Factor Loadings of M easurem ent Model 80 19. Variances/Covariances of Error-Uniquenesses of 81 M easurem ent M odel - Experts - Intermediates - Beginners 20. Factor V ariances/Covariances of M easurement Model 82 - Experts - Intermediates - Beginners 21. Scaled Means of M easured Variables 84 22. Scaled Standard Deviations of measured Variables 85 23. Factor Loadings and Variable Intercepts of Measurement 88 Model w ith Means Estim ated 24. Factor M eans and Significance of Mean Differences from 88 Beginners 25. Variance/Covariances of Error-Uniquenesses of 89 M easurem ent Model w ith Means Estimated - Experts - Intermediates - Beginners 26. Variances /Covariances of Factors of Measurement Model 90 with Means Estimated - Experts - Intermediates - Beginners 27. Significant Factor M ean Differences between Experts and 92 Interm ediates R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Tables 28. Significant Chi-square Change with 1 Degree of Freedom when a Path is Released to be Freely Estimated - Within the Domain of GO Expertise - Outside the Domain of GO Expertise 29. Total Effects of Age on Factors 30. Factor Means after Total Effects of Age are Considered and the Significance of Mean Differences from Beginners 31. Significant Factor Mean Differences between Experts and Intermediates after the Total Effects of Age are Controlled for from Factors 32. Intercorrelations of GO Tasks within Four Expertise Subgroups 33. Intercorrelations of Non-GO-Tasks w ithin Four Expertise Subgroups 34. Correlations between GO Tasks and Non-GO Tasks within Four Expertise Subgroups 35. Correlations of GO Tasks w ith Age within Four Expertise Subgroups 36. Correlations of Non-GO Tasks with Age within Four Expertise Subgroups R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. X List of Figures Figure Page 1. Schematic Structural Model to be Analyzed w ith Means 32 Estimated (but to avoid clutter m ean effects are not shown here) over Expertise-Specified G roups 2. Schematic M easurement Model w ith Means Estimated 34 (to avoid clutter correlations am ong factors are not show n here) over Expertise-Specified Groups 3. Age Distribution within Whole Subjects (N=263) 39 4. M easurement Model with Means Estimated (to avoid 87 clutter correlations among factors are not show n here) over Expertise-Specified Groups 5. Structural Model w ith Means Estimated (but to avoid 97 clutter mean effects are not show n here) over Expertise- Specified Groups R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. xi Abstract Age-related declines in short-term apprehension and retrieval (SAR), processing speed (Gs) and attentiveness (ATT) account for some of the age- related declines in fluid reasoning (Gf). One prom inent explanation of the causes of aging declines in cognitive abilities is the theory of disuse or lack of practice. This study investigates w hether or not frequent and intensive practice of the age-vulnerable cognitive abilities required in the developm ent of expertise has positive effects over the course of cognitive aging. Expertise- specified subgroups, each with comparable wide age range, are constructed in the dom ain of GO (an ancient chess-like game). Measures of ATT, Gs and SAR (elementary Gf processes) both w ithin and outside the GO dom ain and established measures of Gf, as such, are obtained. Structural models specifying how the elementary processes account for age differences in Gf are compared across the three levels of practice (i.e., expertise). Results indicate that higher GO expertise is associated w ith higher means for the GO- em bedded measures but not for m easures outside the domain. In particular, high levels of expertise are associated w ith expanded working memory capacities within the dom ain of expertise. With very high levels of expertise this expanded (long-term working) memory is substantially larger than the 4+1 or 7+2 elements that have characterized (short-term) working memory. For m easures obtained within as well as not within the dom ain of GO expertise, there is aging decline in all Gf components and Gf itself. This occurs at all levels of practice. Within the subgroup of highly practiced Experts this decline is most pronounced for ATT, where the decline is progressively and significantly sm aller for the less practiced Intermediates R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. xii and Beginners. When all aging effects indirectly determ ined through the processes of ATT, Gs and SAR are controlled both within and outside the dom ain of expertise, the rem aining direct effect of age on Gf is significantly larger for the least practiced Beginner subgroup than for the more practiced subgroups of Intermediates and Experts. Thus, the results do not support the hypothesis that high levels of intensive practice prevent the decline of major components of Gf reasoning. The results suggest that attentiveness, in particular capacity for dividing attention, is most vulnerable to the factors that produce age-related loss of Gf. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Chapter I: Review of the Literature Introduction It is well known that populations worldwide have been aging during the 20th century, although the rate of aging has varied considerably from one nation to another (the Organization for Economic Cooperation and Development, 1988). Current demographic projections have led the OECD Secretariat to estimate that the aging of the population will continue well into the next century. Indeed, in the United States, for example, there will be an 'elderly boom' after 2010 when the baby-boom cohort begins to reach the age of 65. It has been estimated that by 2030 the U.S. population at age 65 and over will be "at least 2.5 times larger than it was in 1980" (Siegle, 1993, p. 2). Although the aging of the population places a w ide range of influences on societies as to, for example, care giving, housing, social security, and retirement policies, one of the biggest concerns am ong aging individuals themselves is about the cognitive changes that m ay occur w ith advancing age. Aging people are especially concerned about losing memory and eventually getting "senile" (Folger & Stem, 1994). H um an Cognitive C apabilities The term intelligence is pervasively used to refer a person's memory and other cognitive powers and abilities. It is often argued that intelligence is "the most distinguishing feature of our species, hom o sapiens" (Anderson, 1990, p. 432). Although there is widespread belief that intelligence is important, there is no consensus about what it is or how it should be defined. Intelligence can only be inferred from behavior. Many behaviors are regarded as indicative of intelligence, and there is no definitive consensus am ong R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. scientists about precisely w hich behaviors are most indicative of intelligence, although there is substantial agreem ent about a variety of behaviors that together are indicative of intelligence (as sum m arized in Horn & Noll, 1997; see also Humphreys, 1985; Sternberg, 1994). A large num ber of different cognitive products of intellectual processes are m easured— hence observed— with tests. The abstract construct is thus m ade concrete. Often measures are grouped together in what are called intelligence tests, said to measure on an intelligence quotient or IQ. The abbreviation IQ is often used interchangeably with intelligence (Kail & Pellegrion, 1985; Richardson, 1900). This is an unfortunate practice because different so-called IQ tests measure different collections of cognitive capabilities (H orn 1986, H orn & Hofer, 1992). It confuses scientific discourse to call different things by the same name. Fluid Intelligence (Gf) and Crystallized Intelligence (Gc) There have been attem pts to organize understanding of the phenomena of intelligence under the rubric of an all encompassing scientific theory. None of these theories of intelligence has been embraced by all scientists of the field, but a theory that has been accepted more or less by many scientists is the theory of fluid intelligence (Gf) and crystallized intelligence (Gc) (Cattell, 1963; Horn, 1967; 1968), particularly the extended form of this theory (Horn, 1982,1997 in press). One major concept of this theory, Gc, is "a set of abilities that people in general and psychologists in particular often refer to in their attempts to specify what they mean by intelligence" (Horn, 1982, p. 858). It is the collection of skills and knowledge that best represents the intelligence of the culture. Through acculturation, individuals acquire segments of this intelligence. Tests designed to sam ple the intelligence of the culture are R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 indicative of this broad capability. While Gc is indeed indicative of w hat people m ean by intelligence, it is not sufficient to represent all of what is referred to as human intelligence. Also im portant are abilities for dealing with novel problems. Such abilities enable one to gain entirely new understandings that are im portant for adjustm ent and adaptation. Abilities characterized as enabling one to cope w ith novel circumstances are those of Gf. This is "a collection of abilities in reasoning, comprehending relationships, draw ing sound inferences, problem solving, concept formation, abstracting, and in general, thinking intelligently" (Horn, 1982, p. 856). According to Horn, Gf best-represents a capacity first-described by Spearman (1927) as "g" and regarded as the sine qua non of hum an intelligence. "Gc is based on, and reflects, individual differences in acculturation learning; Gf is based on, and reflects, individual differences in . . . casual learning— learning that is not heavily shaped by acculturation" (Horn, 1985, p. 289-290). Gc is more strongly correlated with indicators of acculturation experiences than Gf: "Gf correlates about .35 with social class and years of education,. . . whereas the comparable correlation for Gc is .55" (Horn & Hofer, 1992, p. 60). Nine Factors of Gf-Gc Theory The extended Gf-Gc theory specifies nine major cognitive capabilities. These nine factors represent most of the m easured features of human intellectual abilities. In sum m ary form these nine factors can be described as follows (after Horn & Hofer, 1992): Fluid reasoning (Gf), m easured in tasks requiring inductive, deductive, conjunctive, and disjunctive reasoning to arrive at understanding relations among stimuli, comprehend implications, and draw inferences R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 4 A cculturation knowledge (Gc), m easured in tasks indicating breadth and depth of the knowledge of the dominant culture Q uantitative knowledge (Gq), measured in tasks requiring understanding and application of the concepts and skills of m athem atics Short-term apprehension-retrieval (SAR), also called short term memory (Gsm), measured in a variety of tasks that mainly require one to maintain awareness of, and be able to recall, elements of immediate situation-i.e., events of the last m inute or so Fluency of retrieval from long-term storage (Glr), also called long-term memory, measured in tasks that indicate consolidation for storage and mainly require retrieval, through association, of information stored minutes, hours, weeks, and years before Visual processing (Gv), m easured in tasks involving visual closure and constancy, and fluency in "image-ing" the way objects appear in space as they are rotated and flip-flopped in various ways Auditory processing (Ga), m easured in tasks that involve perception of sound patterns under distraction or distortion, m aintaining awareness of order and rhythm am ong sounds, and comprehending elements of groups of sounds, such as chords and the relations among such groups Processing speed (Gs), although involved in almost all intellectual tasks (Hertzog, 1989), m easured most purely in rapid scanning and responding in intellectually simple tasks (in which almost all people would get the right answ er if the task were not highly speeded) Correct decision speed (CDS), m easured in quickness in providing answers in tasks that require one to think (pp. 56-57) Life-Span Developm ent of H um an Intelligence W ithin an Individual As stated earlier, aging individuals share a major concern about the age-related cognitive changes that m ight make them "senile." Speaking to this concern is evidence accumulated over more than 30 years. This w ork R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 5 suggests that some cognitive capabilities decline with advancing age, while others do not or even improve throughout most of adulthood (Horn & Noll, 1997, for recent review). Declines have been most clearly demonstrated for Gf, SAR and Gs. Declines (and improvements) that have been dem onstrated are in averages computed over a num ber of individuals at each age. The averages are interpreted as representing individuals. The difference in averages are interpreted as changes occurring w ithin individuals (H orn & Hofer, 1992). Questions arise about this evidence of decline. Often the questions are premised on hope that the evidence is somehow incorrect or misleading. One reply to the evidence is that the abilities of the research findings are not important, not abilities that really m atter for the work, enjoyment, adaptations, and adjustm ents of hum ans. Another reply is from research evidence showing that the averages for Gc increase over a major period of adulthood or at least do not decrease until very late in life. This evidence is comforting, but unfortunately it does not discount well-replicated evidence of age-related declines in abilities that appear to be important. In the same studies in w hich aging increase has been demonstrated, it has been found that the abilities of Gf, SAR and Gs steadily decline over most of the period of adulthood (Horn, 1997 in press; Horn & Hofer, 1992; H orn & Noll, 1997 for recent review). These findings have prom oted understanding that cognitive capabilities are not unidim ensional (only one factor of g) and not unidirectional (only regression). Instead, there is more than one intelligence, and aging involves both enhancem ent and deterioration. Horn (1997 in press) notes that when two types of abilities, one having a positive relation to R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 6 age and the other having a negative relation to age, are combined in a conglomerate measure, as they are in measures of IQ, an important developmental distinction is lost. Since both enhancem ent and deterioration proceed in the course of aging, cognitive aging can be described as the process in which the discrepancy between the functional levels of Gc and Gf widens "within" an individual w ith advancing age. Researchers have m easured the incremental function of Gc in the second half of life w ith tests of vocabulary, verbal ability and general information. This incremental function of Gc appears to be ascribable to acculturation experience that increases with age. Nearly all types of one's experiential history is reflected in Gc (Baltes, 1987; H orn & Cattell, 1967; Horn & Hofer, 1992; Rabbitt & Abson, 1991). The evidence for decline in Gf has been obtained w ith inductive reasoning tasks, measures of concept formation w ith novel materials, indicators of working memory, information-processing speed, and paced learning in a simple laboratory tasks (Baltes, 1987; Hertzog, 1991; Horn, 1985; 1987; Horn & Cattell, 1967; Horn, Donaldson, & Engstrom, 1981; Nettlebeck & Rabbitt, 1992; Rabbitt, 1993a; Schaie, 1990; 1994). The decline of Gf during adulthood, from approximately 30 years of age onw ard, is between 3 and 7 IQ points per decade (Horn et al., 1981). One attempt to explain why age is so differently related to Gc and Gf has supposed that Gc is "automated" and Gf is "controlled" in the following ways: The 'fluid' capabilities of a network w ould correspond to its total available resources for providing the maximum num ber of alternative pathways and inform ation processing control during learning of novel tasks. The 'crystallized' properties of the network would correspond to its acquired R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. stock of special-purpose pathways, each developed by practice to carry out a particular task with maximum efficiency, and with m inim al dem ands on the information processing resources of the network as a whole. (Rabbitt, 1990, p. 240) This m etaphor is sim ilar to Snow/s (1981) suggestion that Gc reflects a long-term assembly of performance processes required for transfer to familiar situations and Gf reflects a short-term assembly of performance processes required for transfer to unfam iliar situations. These hypotheses suggest that "practice"— repeated use— makes one's cognitive skills resilient to the effect of aging. The abilities of Gc are practiced daily. These abilities are sufficient to enable one to get by adequately in most of the coping required in every day living. In consequence, the "raw" information processing abilities of Gf are not often needed, are not used, and thus are likely to decline from lack of use as age progresses. This hypothesis thus makes Gf of particular concern for this study. It represents a flexibility of thinking based on a reasoning ability, or an efficiency of information processing that is relatively independent of cultural determinants. It enables one to educe relations and correlates and to arrive at conclusions in the pursuit of resolving complexities. It involves abstract systems of procedural rules for solving problems. These systems are defined operationally w ith tests of abstract reasoning, problem solving, memory functioning, and decision making. These capabilities are very much needed in the formation of the intelligence of Gc, but they may become increasingly less needed as Gc takes over the intellectual functioning of every day life. At least this is a plausible hypothesis that has yet to be given adequate test. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Interindividual Differences in Cognitive Change In addition to the average "within" individual cognitive changes described in the previous section, attention has been paid to "interindividual" differences in the changes (Nesselroade, 1987; 1991). Rabbitt and his colleagues (1990; 1991; 1993a; 1993b) have suggested that w ith advancing age, the difference between cognitive abilities of the most able individual and the least able individual steadily increases, and this, rather than inevitable differences between young and old, m ay account for findings of decline. This hypothesis requires that only a few individuals bring the averages down. In a study in which choice reaction tim e was cross-sectionally compared, Rabbitt (1993a) found that age accounted for only 9% of the variance of the performance by subjects betw een 20 and 79 years of age. Salthouse (1991a) dem onstrated that for some m easures of cognition the variance associated w ith aging is smaller than th at of individual differences within separate ages. For several abilities, the variance within age groupings increases, as the average age of the group gets higher (Christensen, Mackinnon, Jorm, Henderson, Scott & Korten, 1994; Horn, 1988; Rabbitt, 1993a). The evidence pertaining to this hypothesis has been mixed. In the 1960 standardization sample of the Wechsler A dult Intelligence Scale (WAIS) it was found that up to 75 years of age individual differences in Gc increased with age, but Gf variance remained relatively constant across age (Horn & Hofer, 1992). Christensen et al. (1994), on the other hand, found that individual differences increased in Gf and m em ory, but not in Gc. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. The difference betw een these findings may reflect differences in the gender and age composition of the samples. Peris (1995) has suggested that because men, on average, die at younger ages than women, samples of older m en are selected for survival relative to women of the same age. He contended that for males in their late nineties, for example, the average is more m entally intact than the average female in their late nineties. If this survival was prem ised primarily on Gc, but not on Gf, the finding of the WAIS w ould result. Interpreting results from a recent study of McArdle (1995) show ing that Gc predicts maintenance of Gf over age but Gf does not predict maintenance of Gc in adulthood, H orn suggested that, indeed, Gc is a repository of inform ation that tends to enable one to retain health and vigor. Thus, Gc w ould be selective for survival, and increases in variability of Gc with age could be expected. By this reasoning (based on findings) Gf would not be selective for survival, and thus Gf variance would not be expected to increase w ith age. Results in accordance w ith this reasoning would not be expected in samples selected for Gc abilities. Such a sample would leave out those low on Gc— i.e., individuals in the lower segments of the Gc distribution, who would, had they been included in the sample, increased the variance on Gc. These appear to be the conditions in the Christensen et al. (1994) study that obtained the finding of age-related increase in Gc variance w ith a sample of 897 comm unity-dwelling males and females aged 70 years of age and over. The sample was draw n in proportion to the actual num ber of individuals in a particular age group. The "average" for different ages may hide the reality of changes within individuals, since "the change of one individual is averaged with the lack of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 10 changes in another individual" (Horn & Hofer, 1992, p. 75). Similarly, Chamess and Bosman (1990) note that: When confronted w ith the wearing array of findings that show that older groups do less well than younger ones on nearly every task invented by experimental psychologists, the tendency is to assume that all older adults perform less effectively than all younger adults. Means tend to hide the extensive individual differences in perform ance in both groups, (p. 348) Thus, to sum up, the evidence is clear in indicating that with advancing age there is increase in intraindividual discrepancy between the average levels of Gc, which is m aintained w ith age, and the average levels of Gf, which declines w ith age. There is also suggestion of increase in interindividual differences in some cognitive abilities with advancing age. In the next section, the evidence on the nature of age-vulnerable Gf will be reviewed and evaluated w ith an aim at obtaining a clear understanding of w hat is presently known about interindividual differences in this important class of capabilities. Age-related changes in Gf need to be studied carefully for, as Perfect and Rabbitt (1993a) note: many changes in cognitive functioning ascribed to ageing can be described as being due to changes in current fluid intelligence level, which covaries with age, but which does not necessarily decline w ith increasing age for all individuals at the same rate. (p. 133) V ulnerable Cognitive Abilities The reasoning abilities of Gf, and thus the aging decline of Gf, can be described partly in terms of elementary cognitive processes— that is, narrower, more nearly nuclear or basic abilities (Horn, 1997 in press; Horn & Hofer, 1992; Horn & Noll, 1997; Horn et al., 1981; Noll & Horn, 1997). Prominent R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 11 among these basic abilities are those of short-term apprehension and retrieval (SAR), often referred to as short-term memory or, more fundam entally, working memory (Baddeley, 1992; Carpenter & Just, 1989). Also related are processes that are identified w ith measures of cognitive speed (Gs), attention and concentration (McDowd & Craik, 1988; Horn, 1997 in press; H orn et al., 1981; Rabbitt, 1993b; Salthouse, 1992). SAR and Gs Horn and H ofer (1992) sum m arized the evidence on age relationship for nine common factors of Gf-Gc theory. They concluded that Gf, SAR, and Gs are vulnerable abilities because they decline both with age and with known brain damage. They described Gc and TSR(Glr), on the other hand, as maintained abilities: they do not decline with age over most of adulthood and are not irreversibly decreased by brain damage in adulthood. Horn (1997 in press) characterized the relations among the age- vulnerable functions as follows: Gf, SAR and Gs appear to be interconnected processes of reasoning. In order to comprehend relationships and make decisions about them (i.e., reason effectively, as in GO it is necessary to apprehend information and hold it in the span of im m ediate awareness (as in SAR) and cycle through various possible relationships quickly (as in Gs). A process of concentration is also necessary. W hen requirements for reasoning of a cognitive ability test are small, answers are provided quickly, automatically. In difficult reasoning tasks, on the other hand, it is necessary to sustain concentration. Individual differences in capacity for concentration become important. Horn et al. (1981) and subsequently Noll and Horn (1997 in press) investigated how age-related declines in SAR and Gs account for the age- related decline in Gf. Using part correlation techniques, they dem onstrated R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 12 that a capacity to apprehend and retain information over short periods of time (SAR) and a capacity to think quickly (Gs) are both involved in Gf reasoning abilities and the aging decline of these abilities. They evidenced this w ith the following findings: 1. The averages for both Gs and SAR, as well as Gf decrease systematically w ith age, 2. The negative correlation between age and Gf is significantly reduced w hen either of SAR or Gs is controlled in the estimate of Gf, and 3. This reduction in correlation is significantly larger when both SAR and Gs are sim ultaneously rem oved from the estimate of Gf. This suggests that both SAR and Gs are part of the age-related decline in Gf. A ttentiveness A great deal of research has focused on functions of attentiveness, or attention (Anooshina, 1989; H orn et al., 1981; H orn & Noll, 1997; Kausler & Kleim, 1978; Kirasic, Allen & Haggerty, 1992; Logan & Etherton, 1994; M adden, 1985; McDowd & Craik, 1988; McDowd & Birren, 1990; Perfect & Rabbitt, 1993a; 1993b; Plude, Enns, & Brodeur, 1994; Plude & Hoyer, 1985; Shiffrin, Dumais, & Schneider, 1981; Sugar & McDowd, 1992). Attentiveness is needed to process incoming information, promptly select what to process, and stay aware of the information so that additional processing of the information can be enabled. Attentiveness is a "mechanism by which incoming information is assigned priority and passed on for further processing" (Sugar & McDowd, 1992, p. 319). According to McDowd and Birren (1990), there are four capacities of attention: 1. divided attention required to pursue tw o simultaneous tasks 2. attention switching required to alternately monitor multiple sources of input R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 13 3. sustained attention required to maintain performance of a task over extended time 4. selective attention required to focus on goal-relevant information Research to date indicates that a capacity of attentiveness, or a capacity to activate and m aintain a proper type of attention is age-vulnerable at least for divided attention and selective attention (M adden, 1985, McDowd & Birren, 1990, Plude & Hoyer, 1985, Rabbitt, 1965). McDowd and Craik (1988) found that divided attention performance declined with increasing age, and that the differences between young and older adults were more salient in concurrent tasks of higher difficulty. H orn et al. (1981) examined the relationship between Gs, Gf, SAR, and attentiveness, particularly that which McDowd & Birren (1990) characterized as sustained attention. Relatively high correlations (r=.30~.40) w ere found among these capabilities. More important, they found that a capacity to maintain close attention while performing cognitive tasks appear to be a component of SAR, Gs and Gf. Moreover, this capacity of attentiveness declines w ith age during adulthood and appear to account for the aging decline in Gs, SAR and Gf. These findings suggest that one of the significant causes of the age-related decline of Gf is the age-related decline of attentiveness. "Loss of abilities in m aintaining and dividing attention reduces ability to encode, which is registered in loss of short-term memory and speed of apprehension, as well as reasoning" (Horn & Hofer, 1992, p. 92). Attentiveness was found to be related not only to Gf, Gs and SAR, but also to measures of concentration, encoding organization, incidental memory, eschewing irrelevancies, dividing attention, working memory, hypothesizing, and inspection speediness (Horn, 1987). R eproduced with perm ission o f the copyright owner. Further reproduction prohibited without perm ission. 14 Causes of Declines in Vulnerable Cognitive Abilities Research, thus, indicates that age-related declines in the elementary processes of SAR, Gs and attentiveness are associated with the age-related declines in Gf. However, this information alone does not indicate how individual differences in the process of cognitive aging are caused. It is still not clear w hat causes interindividual variances in age-related changes in the elementary processes of the Gf reasoning ability, and there is little consensus in what causes individual differences in cognitive aging among same-age adult populations (Salthouse, 1988; 1991a). "Little is known precisely about w hat produces declines and enhancements in cognitive capabilities in adulthood" (Horn & Hofer, 1992, p. 93). There are two prom inent alternative theories about the causes of aging declines in vulnerable abilities. Discussed most often is the brain damage theory. W hen it is know n that brain damage has occurred, there is decline in cognitive capabilities, m ost prom inent and most persistent for the vulnerable abilities, particularly Gf. Autopsy evidence suggests that it is more likely that evidence of brain dam age will be found with advancing age. From these findings, it is reasoned that as age progresses, the likelihood that one will suffer brain dam age increases and this will register in loss of vulnerable abilities. Blows to the head, breathing carbon monoxide, abusive use of alcohol or anoxia produce brain damage. This damage can accumulate. Small losses thus can m ount to larger losses in brain function. Ultimately the loss can become large enough to be detected w ith measures of Gf, SAR and Gs (Horn & Hofer, 1992). Different life styles probably lead to different losses of processes of cognitive abilities with advancing age (Horn & Noll, 1997). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 15 A second explanation for age-related declines in cognitive capacities has grown up from the interest in the modifiability of older adults' cognitive functioning (Shaie, 1994; Shaie & Willis, 1986; Willis & Shaie, 1986; Willis, 1990). Positive results from the cognitive training studies on such abilities as inductive reasoning and spatial orientation am ong older adult subjects have led researchers to argue that the age-related decline of vulnerable abilities might be ascribable to "disuse" of the particular abilities (Shaie, 1994). The basic notion underlying this argum ent is that "older adults, relative to younger adults, infrequently use Gf reasoning and other vulnerable abilities— have little daily practice in these abilities— and this is why the abilities decline" (Horn & Hofer, 1992, p. 88). Findings from studies on expertise add to the theory that lack of use results in loss of vulnerable abilities. These results indicate how individual differences in daily practice relate to maintenance and increases in im portant abilities (Ericsson, 1996; Ericsson & Chamess, 1994; Ericsson & Kintsch, 1995; Morrow, Leier, Altieri, & Fitzsimmons, 1994). Theory about the development of expertise thus provides an approach to research on whether or not frequent use (daily practice) of the age-vulnerable cognitive abilities results in maintenance and enhancement of these abilities over the course of aging. If developm ent and maintenance of expertise m aintain the otherwise vulnerable abilities required for exercise of the expertise, then this is evidence in support of the "disuse" explanation of age-related declines in vulnerable cognitive abilities. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 16 Expertise and Individual Differences in Cognitive Aging D evelopm ent of Expertise Advanced expertise in chess, crossword puzzle solving, medical diagnosis, piano playing and tennis results from a long-term (usually more than a decade of) deliberate and well structured practice (Anderson, 1990; Ericsson, Krampe, & Tesch-Romer, 1993; Ericsson & Chamess, 1994). Deliberate and well structured practice is directed at improvement, and usually is supervised by teachers and coaches who provide the person with immediate feedback that is well designed to eliminate "bad" habits and force advance to ever more difficult skills (Ericsson et al., 1993). With such practice in developing and using knowledge in a particular context, the structure of knowledge changes. Practice facilitates knowledge compilation in which knowledge becomes more and more compiled into production rules, or condition-action pairs, and compiled knowledge enables experts to automate or proceduralize their performance in their dom ain (Anderson, 1990; 1993; Anderson & Fincham, 1994). This autom atization (or proceduralization) of knowledge m akes the experts' application of knowledge more rapid and reliable (Anderson, 1993; Anderson & Fincham, 1994). Major characteristics of expert performance can be summarized as follows: 1. Experts perceive large meaningful patterns in their domain. 2. Experts are fast; they are faster than novices at performing the skills of their dom ain, and they quickly solve problems with little error. 3. Experts see and represent a problem in their domain at a deeper (m ore principled) level than novices; novices tend to represent a problem at a superficial level. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 17 4. Experts have strong self-monitoring skills (Glaser & Chi, 1988). Expertise is thus characterized by extremely high levels of autom ated skill in a given activity domain. As emphasized, it is acquired only w ith deliberate, well-structured practice (Chamess & Bosman, 1990; Ericsson et al., 1993; Ericsson & Chamess, 1994; Glaser & Chi, 1988; M orrow et al., 1994; Salthouse, 1987). Salthouse (1985; 1990; 1991b) dem onstrated that expertise and aging m ust be related. This is logically necessary because both occur over time: one must get older as one engages in deliberate practice. According to Salthouse (1985), the development of a skill within a given dom ain m ight provide a hint to understand a nature of changes in information processing with advancing age, since behavioral capacities of an individual go through a dynam ic change in the process of the acquisition of expertise. Results from the studies on the nature and developm ent of expert performance suggest that w ithin some domains of expertise, individuals develop and maintain basic abilities comparable to the basic abilities and elementary processes of Gf. For example, successfully playing the game of chess requires reasoning, and this reasoning appears to be sim ilar to the reasoning required in test m easures of Gf. Related to aging, it has been found that older chess experts are capable of looking as far ahead as younger experts and search for the next move as deeply as their younger counterparts (Chamess, 1981a; 1991). This suggests that inductive reasoning abilities in the dom ain of chess are m aintained among chess experts until old age. Similarly, in studies in which novices and experts in the area of crossw ord puzzles were compared, it was found that among novices crossword-solving ability was positively correlated with test scores on fluid intelligence (r=.719) and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 18 negatively correlated w ith subjects' ages (r= 248), while among experts it was positively associated w ith age (r=.241) and barely associated w ith test scores on fluid intelligence (Rabbitt, 1993b). These findings thus suggest that if the basic abilities of Gf reasoning were assessed w ithin an area of expertise, those abilities, and hence Gf, w ould not decline w ith age— i.e., at least they would not decline if expertise itself did not decline. This suggests the major hypothesis guiding the design of this study. Before presenting research questions and hypotheses in detail, it is necessary to build up a rationale for the study from review of the evidence indicating the nature of three basic processes of Gf that appear to be related to advanced expertise; the processes of short-term w orking memory, cognitive speed, and attentiveness. Expertise an d Short-term W orking M emory (A pprehension and Retrieval, SAR) Ericsson and Kintsch (1995) pointed out that the current theory of working m em ory does not describe the large am ount of information that is brought into awareness during expert performance. Deep mental planning on a move sequence for a chess position requires awareness of more than merely seven plus or m inus two alternatives— the capacity limit show n to obtain in studies of short-term working memory. For a search of the best possible move to be effective, chess experts have to be capable of keeping track of envisioned moves while pushing their search deeply into tree of move possibilities (Chamess & Bosman, 1990). Ericsson and Kintsch (1995) argued that in order for a theory of working memory to explain such high dem ands on imm ediate awareness, it is necessary to extend the concept of tem porary storage in short-term working R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 19 memory to a concept that includes the storage of long-term memory (LTM) and allows quick, reliable retrieval of that stored inform ation in the working memory space. Such retrieval of information would be triggered by a cue in the working memory that matches the retrieval structure in LTM. This new theory thus requires a concept, long-term working m em ory (LT-WM), that is distinct from the established concept of short-term w orking memory, ST- WM. The Ericsson-Kintsch theory thus suggests that experts develop an increased working memory capacity in (exclusively in) their dom ain of expertise. The concept of LT-WM provides a basis for description of phenomena that are not w ell explained by resort to ST-WM. For example, experts are able to m anipulate large am ounts of inform ation in the im m ediate situation. They are able to successfully resum e activities after experiencing interruptions m aintaining such large am ounts of information. Such successful resum ption of expert performance after interruptions can be explained with a concept specifying access to inform ation in LT-WM while solving problems. According to Ericsson and Kintsch (1995), in the course of accumulating domain-specific knowledge, procedures and perceptual-m otor skills, individuals also attain domain-specific capability for expanding working memory capacity. Ordinary working memory is expanded by developing methods for storing information in LTM in accessible forms. This storage of information in LTM is possible only for skilled experts who have become able to "foresee retrieval dem ands and develop memory skills to index relevant information with retrieval structures" (Ericsson & Kintsch, 1995, p. 239). This comes about only through an extensive experience at meeting the dem ands for stably completing the task. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 20 The most frequently studied activity that demonstrates the exceptional capabilities of expert performance is memory for meaningful stimuli from a task dom ain (Ericsson & Smith, 1991). Superior performance of experts over novices in their dom ains of expertise has been found w ith measures that are essentially the sam e as those that indicate the SAR capacity in the studies of H orn and coworkers (e.g., H orn & Noll, 1997). The tasks used in research w ithin the expert dom ain are measures of recall (reproduction) and recognition (Chamess, 1981a; 1991; Ericsson & Smith, 1991; Reitman, 1976). Experts recall and recognize significantly larger am ounts of information in their dom ain of expertise than novices even when stimuli are presented for a very short duration. These findings support the Ericsson-Kintsch theory of LT-WM. Researchers have also explored how age relates to experts' memory performance for m eaningful stimuli w ithin the dom ain of expertise. Som ewhat threatening to the major working hypothesis of this study, the findings suggest that there is aging decline within the dom ain of expertise. Cham ess and Bosman (1990) provided a review of the findings on this issue. They concluded that "the recall advantage accruing from great skill declined with increasing age" (p. 356). Chamess (1991) noted that "older chess players do not recall briefly presented chess positions as effectively as do equally skilled younger players. This disadvantage increases as exposure duration increase from 1 to 2 to 4 seconds" (p. 46). Chamess and Bosman (1990) hypothesized that this age-related decline in recalling chess information indicates that older players are slower in the pattern recognition and chunking processes. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 21 Research findings thus indicate that memory functioning declines with age even among experts. The m easures that indicate this decline are indicative of ST-VVM. Is this decline as large as the decline in SAR outside a dom ain of expertise? No evidence of direct relevance for answering this question was found in the research literature. Results from the sam e studies that indicate decline in ST-WM also indicate an LT-WM that is more or less maintained among older experts— i.e., LT-WM was not literally m easured on a scale that would perm it test of the m agnitude of difference; it was simply described in terms that indicated maintenance. Depth of search is a type of working memory measure that defines the point at which a player can no longer retain accurate information about the projected changes to a presented chess position (Chamess & Bosman, 1990). Chamess (1981a) found that not only do the older chess experts search for the next move as deeply as their younger counterparts, age has little relation to the quality of the move that directly reflects depth of search (Chamess, 1981a; 1991). This indicates that advanced expertise helps older persons m aintain an im portant function of working memory. Expertise and Cognitive Speed (Gs) Experts are faster as well as more accurate than novices at performing the skills of the dom ain of expertise (Glaser & Chi, 1981). The extensive knowledge that is organized within the retrieval structure of experts enable them not only to efficiently access relevant information, but also to rapidly reject inconsistent hypotheses while solving problems (Ericsson & Smith, 1991). Chamess and Bosman (1990) described how high speed of expert performance is m aintained am ong older chess players: R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 22 In brief, age seemed to have little impact on the quality of the move selected, though it did exert an influence on the processes responsible for the move selections. In fact, the curious picture arises that older players were more efficient than younger players, choosing an equally good move w ith less extensive search. O lder players took significantly less time, on average, to choose their move. (p. 355) Expertise development thus appears to be associated with increased Gs speed within the domain of expertise— at least to the degree of speed that is rewarded in completing real-life expert performance. Indeed, Charness and Bosman (1990) maintained that "the input flow rate m ust be under the person's control, otherwise slowing in basic perceptual operations may limit performance" (p. 377). Expertise and Attentiveness The following statement by Salthouse (1985) indicates the nature of attentiveness am ong experts: Increased efficiency and automaticity of tasks w ithin the dom ain of expertise may reduce the attention dem ands of the relevant activities and thus free more for the performance of concurrent tasks, but it seems unreasonable to suggest that experience increased an individual's general attentional capacity, (p. 106-107) Plude and Hoyer (1985) add that practice and experience diminishes attention dem and. In autom ated processing based on practice, "the identity and sequencing of processes is predeterm ined, thereby minimizing the demand on attentional capacity" (p. 52). The distinction between autom atic and controlled information processing and the relationship betw een these functions and attentional requirements have been studied rather in depth by Logan (1988), Shiffrin, Dumais and Schneider (1981), Schneider (1985) and Strayer & Kramer (1990). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 23 Schneider (1985) found that after a four-phase transition from controlled to automatic processing a person reaches a state wherein a very quick memory search is performed in the form of "pure automatic processing". With such processing, an autom ated task can be performed perfectly, while sharing time with another task which dem ands attentional resources. This also is a process that is not described or explained w ith ST-WM in inform ation processing theory. Age-related maintenance of attention in expert performance has also been described in Morrow et al. (1994), studying younger and older pilots on real-life tasks. They found that when a task was highly dom ain-relevant within a domain of expertise, word recognition, parsing, and the process of integrating the message information w ith context was descriptively sim ilar in older and younger experts. That is to say, "expertise often aids in identifying task-relevant information" (Morrow et al., 1994, p. 142). Results from several studies thus indicate that the development of expertise is associated with the attainm ent and maintenance of high levels of functioning of Gs, attentiveness, and a LT-WM portion of working memory. This suggests that elementary processes of Gf may be m aintained within domains of expertise. Experts appear to retain Gf reasoning ability in their dom ain throughout the course of aging. There have been studies in which ordinary measures of SAR, Gs, and attentiveness have been measured to provide age-related changes in Gf among experts. There have been no studies of Gf processes, as such, m easured in a particular dom ain of expertise. If the basic abilities of Gf reasoning were assessed w ithin a dom ain of expertise, would those abilities, and hence Gf, decline with age? If so, would R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 24 the decline be of a m agnitude comparable to that found outside a dom ain of expertise? These questions will be addressed in this study. Sum m ary Review of the literature indicates that: 1. Short-term apprehension and retrieval (SAR), cognitive speed (Gs) and attentiveness are basic processes of an important fluid reasoning (Gf) ability that declines with age in adulthood, 2. These processes facilitate aging declines of Gf, and are themselves age- vulnerable, 3. Aging declines in attentiveness are associated w ith age-related declines in Gs, SAR and Gf, 4. Advanced expertise appears to alter functioning of these three processes w ithin the dom ain of expertise; the processes are, at least to some extent, m aintained with aging, 5. The relationships between SAR, Gs and attentiveness w ithin and outside a dom ain of expertise have not been described, 6. There is no concrete evidence from comparisons of experts and novices over the range of adulthood on m easures of attentiveness, Gs, SAR, and Gf w ithin and outside the dom ain of expertise, 7. There is no concrete evidence of age-related changes in Gf m easured conventionally and measured through measures of the elementary processes of Gf within a dom ain of expertise. These conclusions from literature review indicate the major questions to be examined in this study. Overall Design of the Study Previous findings and current theory of cognitive development in adulthood thus suggest that it is im portant for advancem ent in understanding of hum an cognitive function to obtain information about how practice of cognitive skills relates to maintenance of these skills through the adulthood period of development. It is possible that some, or perhaps R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 25 even all, of the decline in Gf reasoning and the components of this reasoning (attention capability, working memory, cognitive speed), is a reflection of decreasing practice of these capabilities. If m easures of these capabilities can be obtained in domains wherein there is a range of practice of the abilities- from very high to very low— that cross-cuts a range of age in adulthood-from young to old, it will be possible to estimate the relationships between practice and aging declines of cognitive capabilities. It will be also possible to compare these relationships w ith those obtained for comparable m easures on which the evidence of age-related decline has been based an d which are believed, in accordance with hypothesis, not to be practiced. The design for this study derives from this rationale. There exists in Japan (particularly, but elsewhere in the world, too) a w ide range of expertise, which is to say practice, in playing the ancient game of GO. This expertise (practice) is distributed across a substantial range of age in adulthood. The abilities developed, practiced and expressed in playing GO can be seen to have m any of the essential features of high-level cognitive functioning-abilities similar in form to those of Gf reasoning. In this respect, the game of GO is sim ilar to the game of chess, w ith which those in the Western world are more familiar. Reitman (1976) has studied the game of GO and found it to be every bit as dem anding of cognitive reasoning as the game of chess. Major characteristics of GO have been described as follows: In Go, two players compete to surround m axim um territory on a board, a 19 x 19 grid, with walls consisting of single pieces, called stones. .. . Players move in turns, at each turn placing one stone on an unoccupied grid point. Isolated stones are incorporated into higher order units called strings. Two or more stones of the same color imm ediately adjacent R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 26 to each other on a grid line constitute a string. Neighboring sets of friendly strings form a group, units of still higher order. It is these higher order units that constitute the walls that surround final territory. Go is thus a game of accretion rather than displacem ent. (Reitman, 1976, p. 339) The game is very complex in that the possible placements, the strategies, and the contingencies are very large. It is very difficult to become expert in the game. It takes years of intensive, well-structured study and practice to reach the highest levels of expertise in GO. Reitman (1976) examined reproduction and recall of GO patterns by a GO Master and a GO beginner, and essentially replicated findings from studies of chess expertise, dem onstrating that experts are superior in recall skills for m eaningful patterns, but not for random patterns, within the dom ain of expertise. In the present study, tests were constructed within the dom ain of GO to measure abilities that are fundam ental to Gf reasoning and that have been shown in the past to account for some of the aging decline of Gf. Tests previously found to be indicative of Gf abilities but not w ithin a dom ain of expertise w ere also used to measure these fundamental abilities. A sample of subjects was draw n in which expertise is to a large extent developed over the adulthood period of developm ent and is maintained into old age. Analyses were directed at exam ining the working hypothesis that age-related decline will not be found for Gf abilities that are measured with tests that are within one's dom ain of expertise, but, in accordance with previous findings, such decline will be found for Gf abilities measured in the ways they have been measured in previous research (i.e., not within a dom ain of expertise). The present study is thus designed to examine whether or not and how components of Gf and Gf itself ages within people who are at different levels R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 27 of expertise in GO. Variables that have been shown to be related to, or be components of Gf (attentiveness, Gs and SAR) and Gf itself are to be analyzed in this study. Research Questions to be Addressed Thus, the sam pling of subjects and measures of this study will provide a basis for comparing levels of expertise (practice) w ithin the dom ain of playing GO for cross-sectional samples of persons of different ages in adulthood. The broad w orking hypothesis to be examined is of the form: Individual difference in the aging declines of components of Gf are positively related to individual differences in am ount of practice of these abilities, as indicated by level of expertise in expression of the abilities; the aging decline of these abilities is less than the decline of comparable abilities producing notable individual differences in a dom ain of which there is no structure for practice in adulthood. The following basic questions are to be addressed: 1. Can the Power Letter Series test that has been a staple of measurement of Gf in the research of H orn and coworkers (and m any others in research on aging) be adapted to provide a comparable measure of Gf with the Japanese language? 2. Will the Japanese adaption of the Power Letter Series test provide evidence of aging decline in cognitive capabilities w ithin a sam ple that is diverse in respect to expertise in GO? Will the evidence be comparable to the evidence provided in studies of samples not selected for levels of expertise? 3. Can subtest m easures of attentiveness, Gs and SAR components (processes) of Gf be m easured within the domain of expertise in GO? R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 28 Given that m easures having the features indicated in items 1-3 are developed, the following questions of structure (relationships among measures) will be considered. 4. Within different levels of expertise (e.g., Experts, Intermediates, Beginners), are the relationships am ong attentiveness Gs, SAR and Gf comparable for both within expertise (GO) tasks (i.e., developed within the dom ain of expertise) and non-expertise (Non-GO) tasks (i.e., tasks developed within previous research)? 5. Are the age-related declines for expertise (GO, highly practiced) skills smaller than the declines for comparable non-expertise (non GO, lowly practiced) skills? 6. Specifically considering memory, is there evidence of a long-term working memory (LT-WM) separate from short-term working memory (ST-WM) that: (a) relates to level of expertise; and (b) supports maintenance throughout adulthood in practiced (expertise) skills? Objective tests and ratings have been developed over the centuries to establish a system of rankings that represent som ewhat more than 40 levels of expertise in GO: Go players are ranked on a single ordinal scale ranging in increasing ability from 35-kyu to 1-kyu to 1-dan to 9-dan. Though progress through the high kyu rank is rapid, later increases in ranking are accompanied by m uch larger amounts of time playing against better players. (Reitman, 1976, p. 339) Each GO player in Japan has a rating in accordance w ith this scale. The Japanese GO Association sets a standard for this rating and guides the practices on the aw ard of these ratings. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 29 These ratings provide a basis for specifying individual differences in well-structured practice of the abilities of GO. The principal hypothesis of the research stipulates that w ith increase along this scale (increase in practice) there is decrease in the relationship between age and m easures of components of Gf that are m easured w ithin the dom ain of expertise. Multiple group analysis, classifying levels of practice, can be used to test the basic working hypothesis. For such analysis the sam ple is divided into groups representing different levels (here levels of practice) and there is a test of (the significance of) the difference between relationships for the variable of interest (here age and the components of Gf). A major problem in such a design is how to divide the sample into groups. This also can be a disadvantage of design if, as in the present case, the division is m ade w ith a m easure that has more levels than the num ber of levels formed by the dividing into subgroups. In this case variance along the scale of measurement is reduced, which means that usually the reliability, and hence discriminability and validity are reduced. Exceptions to this generalization occur when there are non-replicable outliers, other such effects of unit of measurement, or non-monotonic nonlinearity along the scale used for dividing the sample. The divide cut-points are to be well-chosen to avoid these difficulties. Also, and more im portant for the present study, the multiple-group analyses permit analysis of the extent to which the structural and measurement model relationships rem ain stable or change as development proceeds (age increases). For example, it is an implicit hypothesis of this research that the basic scales (components of Gf, Gf) measure the same things all along the range of levels of practice. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 30 M ultiple-group analyses can be particularly informative if the partitioning into groups can be done in a non-arbitrary meaningful way. Fortunately, in the present study, this is the case. This is true because the ratings of GO that are based on centuries of experience, provide a division of players into categories that represent the equivalent of movement by large steps from one stage to another. Ratings below— classify beginners. The kyus within this range represent different levels of learning on the rudim ents of the game. On the other hand, ratings above— indicate top-level, accomplished experts. Not only are all of the rudim ents of the game fully mastered by people in this group, these persons have dem onstrated through tournam ent play that they can play w ith the very best and, on occasion at least, beat them. In between the extremes of Beginners and Experts is a meaningful classification of Interm ediates— people who have a firm grasp of the rudiments of the game, and in all respects play the game well. Given a meaningful classification into categories of different levels of practice, the principal working hypothesis stipulates that the relationship between age and indicators of Gf developed within the relevant dom ain of expertise will decrease systematically as level of expertise increases; specifically the hypothesis is that the relationship between age and m easures in the dom ain of GO will be smaller in the class of Intermediates than in the class of Beginners and the relationship will be smaller in the class of Experts than in either the class of Intermediates or the class of Beginners. This will be true for all GO-related m easures— that is, all indicators of components of Gf that are reliably and validly measured w ithin the area of expertise. A contrast hypothesis stipulates that for indicators of components of Gf that are not measured within the area of expertise, the relationships between R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 31 age and measures that are measured outside the domain of expertise will not differ significantly. With thorough m ultiple-group m odeling analyses (McArdle & Hamagami, 1991), these hypotheses can be tested even under an assum ption that level of expertise in GO is partially determined by prior levels of Gf a n d /o r the components of Gf. This is true because in thorough multiple- group analyses, the m eans and variances as well as the covariances (correlations) are analyzed. If the level of expertise reached in GO is partly a function of Gf (or its components), which seems at least plausible if not likely, it w ould m ean that prim arily only those individuals in the higher ranks of Gf will reach the higher ranks of expertise in GO. This w ould be an effect on the mean, which would not affect the level of relationship w ithin groups. Specific M ultiple-G roup Design and Hypotheses All subjects will be m easured on three components of Gf both within the dom ain of expertise in GO and using measures of previous research not couched w ithin any particular dom ain of expertise. The three components of Gf will be attentiveness (ATT), short-term apprehension and retrieval (SAR) and cognitive speed (Gs). A n adaptation of Power Letter Series for the Japanese language (PowLet) will be used to estimate Gf. To check that this m easure represents the common factor, three established indicators of Gf— Topology, Mazes and Backward Span— will be analyzed. The total sam ple will be divided into three groups for the m ultiple-group analyses: Beginners, Intermediates, and Experts. The m easurem ent and structural model design for the multiple-group analyses with these three groups is described in sum m ary form in Figure 1. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 32 Figure 1 Schematic Structural M odel to be A nalyzed w ith M eans Estimated (but to avoid clutter m ean effects are not show n here) over Expertise-Specified Groups RECNOl G O ATT FI RECN02 1 RECN03 IDENTGO IDENTNO Non-GO Gs F4 COMGO COMNO Non-GO SAR F6 Topology PowLet REPNO f RESNO BackSpan E16 E17 Note: Expected Path Coefficients for Age -> GO Factors (E= -. 2,1= -. 3, B= -. 4) Expected Path Coefficients for GO Factor -> GO Factor (E=. 2,1=. 3, B=. 4) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 33 In the term of Figure 1, the following are the hypotheses of this study: A. Invariance of M easurem ent The pattern for the m easurement model specified in Figure 2 will be invariant. That is, a model in which the pattern relationships are the same for the Beginner (B), Intermediate (I) and Expert (E) group will fit the data. As shown by H orn (1991), H orn & McArdle (1995), and M eredith (1964; 1993), invariance of the pattern in a m easurem ent model of the form depicted in Figure 2 is necessary and sufficient to support an hypothesis that the common factors of the model are m easurem ent invariant. Such invariance is a necessary prerequisite for meaningful comparisons and analyses of group means, variances and covariances (as show n in detail by Horn, 1991). M easurement invariance does not require invariance of the factor means, variances or covariances. M easurement evidence of factors is not sufficient to support an hypothesis of m easurem ent invariance of the manifest variables, as such, although it is indicative of this (Meredith, 1993). B. Averages Between G roups Fixing an invariant m easurem ent model across groups, it is hypothesized that there will be significant difference between the B, I, and E groups in common factor means on GO-ATT, GO-Gs, GO-SAR, and GO. The order of the means will be E-Mean > I-Mean > B-Mean for all three components. This will indicate that the measures of Gf components w ithin the dom ain of expertise are indeed indicative of abilities within that domain. It is not specifically hypothesized that the B, I and E groups will differ on the Non-GO-ATT, Non-GO-Gs, and Non-GO-SAR measures of Gf components outside the dom ain of expertise. There will be these differences if these Gf-components are predictive of the level of expertise reached. Under R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 34 Figure 2 Schematic M easurement M odel w ith M eans Estim ated (to avoid clutter correlation ^ m n n g ^ rfn rs arp n o t show n here) over Expertise-Specified G roups G O A Non-G ATT RECNOl RECNQ2 RECN03 GO Gs IDENTGO IDENTNO COMNO C onstant REPNO BackSoan PowLet lopology R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 35 an assum ption that such differences should obtain, it is expected also that the means for Gf should be in the same order as for GO-tasks~i.e., E-Mean > I-Mean > B-Mean, as specified above. If such conditions obtain in the data, controlling for these differences in Non-GO tasks— a selection effect— in test of the between-group difference in means for the GO components will provide a som ewhat more precise test of the hypothesized effects. C Relationship Between Age and Gf-Components With a m ultiple-group analysis, the significance of the difference in the structural part of the m odel across the three expertise-specified subgroups will be tested. The chi-square change will be tested for paths from age to the common factor by releasing each to be freely estim ated. Each such release provides a one-degree-of-freedom test of an hypothesis of an aging effect. Summing the direct and indirect effects provides a basis for estim ating the negative effects of age on each factor within and outside the dom ain of GO expertise. For measures obtained w ithin the dom ain of GO, the hypothesis is that the B group will have the largest degree of total aging effects, and the E group, the sm allest degree of it. On the other hand, total negative effects of age on factors outside the dom ain of GO will not differ significantly across the three expertise-specified subgroups. In consequence, the B groups w ill experience the largest degree of total negative aging effects on G f-sum m ed over all the com ponents-and the E group, the smallest total aging effect. After the total effects of age on the factors are controlled for, factor means in the dom ain of GO will maintain the same order as hypothesized in the analyses with the m easurem ent model— that is, E-Mean > I-Mean > B- R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 36 Mean. On the other hand, the differences in factor mean of Gf across the subgroups will become wider (or significant). These are the principal hypotheses of the study. It can be seen that there are num ber corollaries and qualifications associated w ith each hypothesis. Rather than spell these out in detail here, it will be best for communication to describe these in the course of the analyses. Importance of the Study Spearman (1927) described the essence of intelligence as the eduction of relations and correlates~in other words, reasoning. These kinds of thoughts lead to a supposition that Gf is the sine qua non of intelligence. It is not pleasant to contemplate that this capability declines w ith age in adulthood. (Horn & Hofer, 1992, p. 60) The expert-novice paradigm of this study provides a valuable basis for examining an im portant hypothesis of a major theory pertaining to the development of abilities that are highly valued. If evidence is adduced that indicates that Gf is m aintained in a domain of expertise, it implies that the decline heretofore found (a) is not an inevitable aspect of loss of neural substrata, and (b) is a function of lack of practice (use). The study is important because it can provide support for an hypothesis that is central to theory of cognitive developm ent. But just as important, the study can provide evidence that contradicts or leads to rejection of this major hypothesis. If it is found that the aging decline of components of Gf w ithin a domain in which these basic abilities are highly practiced is as large as the decline found outside the dom ain of expertise, this evidence also is critical to understanding the nature of cognitive development. It suggests that explanation of Gf decline in terms of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 37 lack of practice is not sufficient to describe the established findings of age- related loss of vulnerable abilities. Findings in between these two extremes of support and no support for the working hypothesis also are im portant in suggesting that previously found evidence of decline may exaggerate the m agnitude of loss, but belief that all the decline is due to lack of practice is overly simplified, too. If the findings indicate support for the major hypothesis, they provide an im portant basis for intervention to prevent loss of im portant abilities. Such results will suggest that high level reasoning is acquired through long, deliberate practice. This knowledge can help guide the search for intervention practices that can ameliorate cognitive decline— one of the greatest concerns among people in this aging society. For instance, if it is found that basic components of reasoning are m aintained through a process of transition from a novice to an expert and these components moderate the age-related decline in Gf itself, there will be evidence that educators can help ameliorate Gf decline through program s designed to develop expertise. Findings indicating that practice to develop expertise w ard off declines in vulnerable cognitive abilities will suggest that frequent reasoning under the threatening conditions of dealing w ith novelty will help prevent age- related declines of the capabilities that people value. On the other hand, findings indicating lack of support for the major hypotheses suggest that developm ent of expertise is not likely to prevent losses of processes that are basic to reasoning. Such findings will help to point the search for prevention of Gf decline in other directions. Also im portant is the evidence this study can provide for or against the hypothesis specifying long-term working memory (LT-WM). The concept of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 38 LT-WM has emerged from findings suggesting that w orking memory as described on the basis of results from studies of short-term apprehension and retrieval is not adequate to explain the large span and other access features of the mem ory of experts in their dom ain of expertise. LT-WM has not been hypothesized, a prior, to be an outcome of study in which measures of ST- WM are obtained with tasks that are typical in indicating its features. This study is thus important in providing a basis for comparing, empirically, LT- WM and ST-WM. Thus, in several w ays this study can be of value. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Chapter II: M ethod Subject A sample of 269 Japanese male GO players, all volunteers, was obtained with the assistance of the GO Association of Japan. In order to obtain this sample, description of the study and requests for volunteers were published in newspapers and GO journals, presented on TV and sent to members of the Association. Am ong 269 participants, 263 provided usable data. The subjects ranged between 15 and 83 years of age: m ean age was 54.97 (SD: 13.71). The age distribution of the subjects is presented in Figure 3, and the GO rank distribution is show n in Table 1. The sam ple represented the full range of 46 ratings of expertise in GO, although the representation among the very lowest (beginning) ranks w as lean. The m ean of this variable was at 4 dan (SD: 4.47). Twenty players not only had the highest ranking in GO, but were also professionals who taught and had earned w orld recognition for their expert play. Figure 3 Age Distribution within Whole Sample of Subjects 701------------------ - - - ----------------------------------------------- Std. Dev = 13.71 Mean = 55.0 N = 263.00 15.0 25.0 35.0 45.0 55.0 65.0 75.0 85.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 40 Table 1 GO Rank Distribution within Whole Subjects GO Rank (Dan, Kyu) N Percent Professional 9 dan+ 5 1.9 Professional 8 dan 3 1.1 Professional 7 dan 6 2.3 Professional 5 dan 4 1.5 Professional 3 dan 1 .4 Professional 1 dan 1 .4 Am ateur 7 dan 5 1.9 Am ateur 6 dan 38 14.4 Am ateur 5 dan 49 18.6 Amateur 4 dan 33 12.5 Amateur 3 dan 33 12.5 Amateur 2 dan 23 8.7 Amateur 1 dan 27 10.3 Am ateur 1 kyu 8 3.0 Amateur 2 kyu 4 1.5 Am ateur 3 kyu 8 3.0 Amateur 4 kyu 3 1.1 Amateur 5 kyu 3 1.1 Amateur 6 kyu 1 .4 Am ateur 7 kyu 3 1.1 Amateur 8 kyu 1 .4 Amateur 10 kyu 1 .4 Amateur 12 kyu 1 .4 Amateur 16 kyu 1 .4 Amateur 30 kyu 1 .4 Total 263 100.0 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 41 Variables In three pilot studies, tests within the GO dom ain of expertise were developed to measure nuclear process abilities that are fundam ental to Gf reasoning and, as indicated in prior research, account for some of the aging decline of this reasoning. Tests outside the GO dom ain of expertise were selected from previous research to m easure the same nuclear process abilities assessed in the domain of GO expertise. The processes on which the research focused were attentiveness (ATT), short-term apprehension and retrieval (SAR), cognitive speed (Gs), and fluid reasoning (Gf). The tests used to measure these processes are described in the following sections. W ithin GO Measures RECGO (Attentiveness). This test, consisting of 14 questions, is based on short-term recognition m em ory under a dual-task dem and. Subjects must simultaneously complete two different tasks. The test thus provides a measure of divided attention. In each of the first seven questions, subjects go through the following steps: 1. They view a m eaningful GO pattern that is projected on a screen for 11 seconds. 2. During this viewing, they count the num ber of either white stones, black stones, or white and black stones combined, whichever is required by instruction attached to the specific question. 3. They also attem pt to remember the pattern. 4. After 11 seconds the GO pattern is removed. Subjects write down the num ber of stones they counted and turn that page over. 5. On next page six GO patterns appear. Subjects encircle one pattern to indicate their recollection of the pattern that had been presented on the screen. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 42 This same procedure is followed in the last seven items of the RECGO test except the patterns are projected for only seven seconds (rather than 11 seconds). The 14 patterns differ in the num ber of stones. Level of difficulty of com prehending a pattern in terms of principles of playing GO increases as total num ber of stones in a pattern increases from 13 to 37. The num ber of stones in a pattern is equal for items 1 and 8, 2 and 9, 3 and 10, and so on. Examples of target GO patterns and six choices accompanying a pattern are show n in Appendix A. Each correct recognition of the GO diagram is assigned one point. Thus, the possible range of the scores is zero to 14. High scores indicate high levels of ability to divide attention between counting and remembering. If scores on this test increase with level of expertise, it suggests that experts successfully hold larger amounts of GO-related information in their w orking memory space and have higher levels of tolerance for distraction. This is an indication that experts develop LT-WM in the dom ain of GO expertise. COMGO (Gs). This test is designed to m easure cognitive (perceptual) speed in comparing patterns that represent different decision problem s in playing GO. It consists of 25 pairs of GO patterns of the lower right quadrant of a GO grid, five pairs printed per page. Two GO patterns (90 degrees apart) are presented side by side. Under highly speeded conditions, subjects are required to decide if the patterns are the same or different. Examples of GO pair comparisons are presented in Appendix A. A 40-second trial of the first 10 items and a 55-second trial of the next 15 item s are separately timed. The num ber of correct comparisons w ithin the R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 43 entire 95 seconds is the measure. Possible scores range between zero and 25. High scores suggest high levels of Gs abilities w ithin the dom ain of GO. IDENTGO (Gs). This test also is intended to m easure cognitive (perceptual) speed in identifying patterns. It consists of 100 GO diagram s of the lower right quadrant of the 19-by-19 GO grid. O n each page, 25 patterns are printed. Some of the GO patterns depict what are called "Atari" configurations. Others do not represent this condition. Examples of test items are show n in Appendix A. "Atari" is a fundam ental and im portant pattern of stone placement. GO player becomes familiar with the "Atari" pattern as they become expert in the game. The patterns representing "Atari" and those not including this condition are random ly placed on a page. Subjects are required to as quickly as possible identify the GO patterns w ith "Atari" by circling them. The trials for the first 50 diagram s and for the second 50 diagram s are separately timed, for 45 seconds each. The num ber of GO diagram s correctly identified within 90 seconds is the score on this test. The possible score on this test ranges from zero to 56. High scores indicate high levels of Gs abilities. REPGO (SAR). This test, consisting of 10 items, measures short-term recall memory. Subjects go through the following steps: 1. They are show n a meaningful pattern of stones of the low er right quadrant in a grid. 2. Subjects are asked to rem em ber as many stone positions (including the color of the stones) as possible. 3. After eight seconds (for the first five items) or five seconds (for the last five items) the slide is removed. Subjects attem pt to reproduce the GO patterns on a page in a test booklet that includes only grids R eproduced with perm ission o f the copyright owner. Further reproduction prohibited without perm ission. 44 of the lower right quadrant, but no stones. They are told to indicate a white stone with a circle draw n in pencil and a black stone w ith a blackened circle in pencil. Some exam ples of GO diagram s to be reproduced are shown in Appendix A. The num ber of stones to be reproduced in each pattern ranged from 12 to 16. The num ber of stones correctly reproduced is the score. Possible scores range between zero and 142. H igh scores on this test indicate high levels of either ST-WM (for those of low expertise) or LT-WM (as expertise develops) within the dom ain of GO expertise. LT-WM capabilities be indicated if experts recall notably more than non-experts and m ore than the 7 plus or minus 2 that has been established as the limits for ST-WM. RESGO (SAR/Gf). This test consists of two types of items, both of which are solved using forward inductive reasoning. The test is modeled on tests used in chess studies in which the player must search for the best move from a given configuration of pieces. In RESGO, subjects are required to indicate the best stone placements to follow a particular mid-game stone configuration. As in chess, performance on this test indicates the player's depth of search— the point at which the player can no longer anticipate and retain accurate inform ation about projected changes (Charness & Bosman, 1990). It assesses the capacity of w orking m em ory-probably a form of LT- WM— in the dom ain of GO expertise. The cue of a particular configuration of stones gives rise to a large num ber of possible moves that must be held in immediate awareness and must be compared in order to choose the one move that is best. As is the case in REPGO, significantly higher scores of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 45 experts than other GO players m ight indicate that experts have acquired LT- WM capabilities. The first type of item used in RESGO has four novel GO patterns constructed within 9 x 9 grids. Players are familiar with this small size GO board, since beginners often practice w ith this or slightly larger (i.e., 13 x 13 grids) boards. Subjects are required to find three weak points in each pattern for the player with black stones and to indicate how that player can eliminate or best reduce the danger. Specifically, subjects draw three best placements of black stones— one for each weak point— in a pattern. The weak points are a threat in the sense that, for instance, the opponent has a high chance of eliminating a black stone or several stones by surrounding it (them) with white stones w ith future placements. Professional 6-dan GO players determined these best placements and also identified less desirable (but acceptable) placements for the 12 w eak points in the four GO patterns. In scoring, each of the most desirable stone placements was assigned two points, each less desirable but reasonable placem ent was assigned one point and other placements were scored zero. The possible range of scores on the four items is zero to 24. The second type of item in RESGO consists of eight novel GO patterns arranged in the lower left quadrant of the standard GO board. In each pattern, there is a configuration from a partially-played game. Subjects are required to demonstrate the best stone placem ents that should follow the presented configuration by indicating the placements w ith consecutive num bers up to a point where no other stone placement is necessary. Subjects are told that it is the turn of the player using black stones. Answering with the number one indicates a placement of a black stone, the second placement is for a white R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 46 stone, the third for a black stone, and so on. To successfully deal with the task, a subject m ust envision several stone placements ahead, while maintaining, evaluating and modifying possible and made changes. Again a professional 6-dan GO player indicated the best answers to the eight items. O n this basis, using all answers provided by the subjects, the GO expert assigned a m inim um of zero points up to a maximum of four points, reflecting the desirability of the sequence of the stone placements and the consequence caused by the stone placements. W hen the expert had difficulty deciding on the rating for a particular stone configuration, he consulted with another 9-dan professional GO player. The scoring thus indicates how very high level GO experts w ould anticipate and make moves. The possible range of the scores on the eight questions of this type is zero to 32. Subjects worked on the 12 questions for eight minutes over both the first and second type of RESGO items. The possible range of scores on both sets of items is zero to 56. High scores over both types of items indicate high levels of understanding and retaining pattern and sequence information in the dom ain of expertise. Examples of the first and the second types of m id game GO patterns are show n in Appendix A. M easures from Previous Research RECNO (Attentiveness). With items that do not tap expertise in GO, this test, consisting of two pages of figures, requires short-term recognition memory under a dual-task demand. Subjects are required to complete two different tasks simultaneously. This test thus provides a measure on divided attention. The first page has 40 figures, the second page, 50 figures. Those figures are shown in Appendix A. While studying the figures on the first page, R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 47 attempting to remember as many figures as possible, subjects are required to count the num ber of squares which construct, or are part of, figures on the page. After one minute and 15 seconds, subjects are told to write down the number of squares they counted, at the bottom of the first page, and then turn to the second page. Of 50 figures on the second page, 20 are figures presented on the first page, 30 are not. By encircling a figure, subjects are required to indicate the figures seen on the first page. Fifty seconds was found to be sufficient to enable them to do this. N um ber of correct recognitions is the score. The possible range of scores is zero to 20. High scores indicate retention under conditions of divided attention. COMNO (Gs). This test consists of 50 pairs of strings of letters. The task is to decide whether or not the two strings of letters are the same or different. The following are two examples of such letter strings: Sam e Different vlkq vklq_______ ________ ________ tnpr tn p r ________ ________ Subjects were given 30 seconds to make as many com parisons as possible for each of two sets of 25 pairs. The num ber of correct comparisons over the entire 60 seconds is the measure. The possible score range is zero to 50. High scores indicate high levels of cognitive speed. IDENTNO (Gs). In each of three pages of m ore than 600 Japanese letters scattered over a page, subjects were required to identify a particular letter by crossing out in each instance. The letter to be crossed out differs from page to page. Examples of test items are shown in A ppendix A. Subjects were given 50 seconds for each page. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 48 The num ber of letters correctly crossed out over all three pages is the measure. The largest total score over three pages was 135. High scores indicate high levels of Gs. REPNO (SAR). In this test, stimuli are pictures of nine cubes for which the three sides that are show ing contain either a num ber or a simple figure, such as a blackened circle. The picture of nine cubes is presented on the screen for one minute. Subjects are asked to memorize the numbers and figures, and their exact places on the cubes. After the slide is removed, subjects are given 40 seconds to reproduce the numbers and the figures on a page of the test booklet that includes only frame pictures of nine cubes— i.e., no numbers or figures are in the cubes. The num ber of correct reproductions is the measure. Possible scores range between zero and 27. High scores indicate high levels of short-term apprehension and retrieval (working memory) ability. RESNO (SAR/Gf). On each of 10 pages there is a maze problem that can be solved by means of forward inductive reasoning. Examples of mazes are presented in A ppendix A. Subjects are given 10 minutes to work through 10 mazes. In each case they draw a line to indicate the best path through a maze while satisfying the following requirements: 1. Start at the very bottom of the maze. 2. Keep the draw n line on a dotted line. 3. Keep the line going upward. 4. Go through as m any dots as possible in the path through the maze. Subjects m ust continuously acquire, evaluate, change and m aintain the envisioned inform ation in their working memory while deciding on the best path way. The test reflects both reasoning and capacity to retain relational information. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 49 In scoring, a path that includes a maximum possible number of dots is assigned three points, paths that include one dot less than the maximum are accorded two point, and paths that include two dots less than the maximum are given one point; other responses are scored zero. Possible scores range from zero to 30. High scores indicate high levels of both Gf reasoning and the SAR function. BackSpan (Gf/SAR). Subjects are required to listen to a set of numbers and recall them in the reverse of the order of presentation. The test consists of 12 sets of numbers, the smallest set being three, the largest, eight. To keep the timing of the presentation equivalent across different testing occasions, the numbers are recorded in advance and played w ith a tape recorder. Score on the test is the num ber of correct recalls. Possible scores range from zero to 12. Horn et al. (1981) found that while backward recall was indicative of SAR, it was most characteristic of Gf abilities. Forw ard recall (i.e., recall in the same order of presentation) was more purely indicative of SAR. PowLet (G f). This test consists of 33 items m odeled on the Power Letter Series test that has been most often used to m ark Gf in the research of Horn and his coworkers. The major difference betw een the original test, based on letters of alphabets of English language, and this newly developed version of the test is that the new test uses characters in written Japanese. These characters are in major respects equivalent to letters of the alphabet of English language, but 51 phonetic characters constitute the Japanese alphabet, compared to only 26 letters in the alphabet of English. Thus, the test can not be a direct transformation of the original test. Four pilot studies, based on 114 subjects, were done in the United States and japan to develop the test, shape it R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 50 to be as similar to the original test as possible, make it equivalent in terms of the difficulty of items, and insure that it has adequate reliability. The developed test item s are show n in Appendix A. The test requires one to discern the pattern and order of a series in a string of letters and choose the letter that continues the series. In the instructions and examples introducing the test, subjects are taught that some problems have no good solution. This is done by introducing trial questions in which there is no solution. Subjects are advised that when a problem of this kind is encountered in the test, they should select a "no answer" (NA) option and move on to the next item. Thus, the subject learn that if after working on a problem for a reasonable am ount of time no good answer can be found, then the problem m ay be one for which there is no solution. This enables the person to abandon a problem that has a good solution w hen that solution is beyond the person's level of comprehension. This condition, plus the fact that the subjects m ust produce an answer— not select an answer from am ong several choices-decreases the likelihood that subjects will guess or can guess a correct answer. These testing conditions also discourage perseveration on particular items at the expense of attem pting others. (Noll & Horn, 1997 in press) At the beginning of the test, three items at a very low level of difficulty are presented to provide a warm -up. They are followed by six sets of five items, each item representing one of five levels of difficulty. Although the test is timed by giving subjects 20 minutes to work on it, the measure obtained by it does not reflect the speed of responding to the questions. The score is obtained by sum m ing up the number of the difficulty level of each item over the sets completed and by dividing the sum by the num ber of completed sets, the maximum of which is six. The sum for a partially R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 51 completed set is prorated on the basis of a set rule. Possible scores range from zero to 16. High scores are indicative of high levels of reasoning. Topology (G f). This test of 20 questions is a subtest of the Culture-Fair Intelligence test (Cattell, 1979). Examples of items are show n in Appendix A. Subjects are required to select a pattern from among five choices that perm its the same placement of a dot within and outside several geometric forms of a figure. It is not a speeded task; subjects are given eight minutes to w ork on 20 questions. The num ber of correct selections is the score. The possible scores range from zero to 20. High scores are indicative of high levels of Gf reasoning. Short questionnaire. A questionnaire was designed to obtain information on: 1. gender, 2. current GO rank, 3. age, 4. the age when GO playing was initiated, 5. hours of weekly practice for each year after starting practicing of GO, 6. experience in individually guided practice, 7. levels of education achieved, 8. employm ent status, 9. major disease within the past five years, and 10. game other than GO that are played w ith some frequency. Only the information on gender, current GO rank, and age was used in the present study. Procedure Two and one-half hours of testing was conducted on three separate occasions in Tokyo, Japan on August 1,11 and 31 of 1996 at the headquarters of the GO Association of Japan. A total of 300 GO players took the test: 25 at the first session, 253 at the second session, and 22 at the third session. At the second testing, which included a large num ber of subjects, six assistants helped the author with proctoring and administration throughout the session. Among these 300 participants, 269 were male. Only the data from R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 52 these male players were analyzed for the present study. (The data for the 31 female players will be analyzed for another study). The num bers of male GO players at the three testing occasions were 20, 229 and 20. The mean ages for these groupings are 47.15, 58.04 and 32.25, respectively. The participants at the first session were staff members of the GO Association of Japan. The participants of the second session were GO players who responded to a request for volunteers published in newspapers and journals, broadcasted on TV, and introduced in GO classes. Considerably more volunteers came forw ard than were actually tested. The capacity of the room available for testing did not allow all volunteers to participate. After the second session, a specific request for younger volunteers was again published in newspapers and journals. The first 20 volunteers to respond to these advertisem ents provided the third subsample. Test items were adm inistered in two booklets. The first booklet and two pencils were distributed to each subject at the beginning of the testing. Subjects were instructed not to open the booklet until asked to do so and to write dow n their own 5-digit-identification num ber on the front cover of the booklet. The second booklet was distributed after completion of the tests in the first booklet and after a 10-minute break. This occurred m idway through 2.5 hours of testing. The 2-page questionnaire was filled out just prior to the mid-testing break. The first booklet included the tests of BackSpan, Topology, RESNO, RESGO, IDENTNO and IDENTGO. The second booklet included COMGO, COMNO, REPGO, RECGO, REPNO, RECNO and PowLet. Every task was accompanied with an instruction page. Most tests also had examples. All the tasks were initiated and ended with a signal by the R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 53 author, who timed each task. Some tasks required slide projector presentation of stimuli; the testing room was equipped with two slide projectors, both located near the m iddle of the room. The two screens (1.8 m x 1.8 m) were located at the front and back of the room. The same slide projectors, operated by the same assistants, w ere used in all the testing. As a rew ard for their participation in this study, the subjects of the second session received lunch and a folding fan. Those of the third session received a portable GO game— a magnetized board and stones. Also, at second session, 140 subjects won a chance to play GO w ith professional GO players, each of whom played 10 people at the same time. The gifts given to the subjects (i.e., the fan and the GO set) were contributed by the GO Association of Japan. After the testing, all test booklets were placed in boxes and shipped to Los Angeles in September, 1996, and scored by the author. Analyses As scoring w as completed, item analyses were directed at determining the basic characteristics of the individual scales— the distribution characteristics (means, standard deviations, skewness, kurtosis) and reliability. Subgroups were formed on the basis of their levels of GO expertise, as determ ined by the ratings of the Japanese GO Association. Preliminary correlation analyses were carried out to provide a first, rough description of how expertise, age and test m easures were intercorrelated. Analysis of variance, ANOVA, was used to obtain basic information of the test performance of the subjects and to provide information about how GO expertise is related to the levels of performance on the tests within and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 54 outside the dom ain of GO expertise. Mean scores of test m easures were compared across subgroups. The Lisrel program (Joreskog & Sorbom, 1996) was used to perform multiple-group modeling analyses of an eight-factor m easurem ent model. As indicated in the statement of rationale for this study, the factors considered are 1. GO-Attentiveness (GO-ATT), 2. Non-GO-Attentiveness (Non-GO-ATT), 3. GO-Gs, 4. Non-GO-Gs, 5. GO-SAR, 6. Non-GO-SAR, 7. Gf, and 8. GO. The eighth factor, GO, was first specified to account for the variability over all GO measures. It represents GO ability itself, as manifested in the ATT, Gs and SAR components. Three m om ent matrices were analyzed to examine m easurem ent invariance for GO Experts (N=92), Intermediates (N=89) and Beginners (N=62). In all, 17 measured variables were analyzed sim ultaneously and the invariance of the measurement model was tested across the subgroups. The importance of m easurem ent invariance in aging research has been emphasized by H orn and McArdle (1992). If the invariance of the measurement m odel holds— requiring equivalent factor loadings across all subgroups— there is evidence that the same constructs are m easured in all subgroups. If it does not hold, there is a need to treat the subgroups as arising from different populations. Figure 2 ( in Chapter 1) indicates the nature of these analyses. (SEM Stepl) Another multi-sample analysis was done, fixing the m easurem ent model, to examine the structural relationships across the subsam ples of Experts, Intermediates and Beginners. In these analyses, differences between groups on the factor means were examined. This analysis indicates how GO R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 55 expertise is related to differences between factor means, within the dom ain of GO expertise and outside this domain. (SEM Step2) M ultiple-group analyses were also conducted in which the invariance of the causal structures was tested and factor means were compared across the three expertise subgroups. The nature of these analyses is indicated in Figure 1 (in Chapter 1). The aim of these analyses was to describe for different levels of GO expertise the interrelationship among factors and relationships between age and factors. In general, it was expected that higher GO expertise would be associated w ith lower influence of age on GO factors. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 56 Chapter III: Results Part 1 The results section consists of two chapters. In this C hapter III— Part 1, there is a review of the results of descriptive statistics and analyses of correlations and ANOVAs. In Chapter IV— Part 2, there is a review of the results obtained from the Lisrel analyses. Descriptive Statistics on Test Measures Table 2 provides a summary of the descriptive analyses over the entire sample of 263 m ale subjects. For four m easures of Gs (i.e., IDENTGO, COMGO, IDENTNO and COMNO), lower boundary of alpha reliability coefficients were computed by intercorrelating total scores from separately timed subparts. For the other tests, alpha reliability coefficients were obtained by intercorrelating total scores of all test items. Analyses of PowLet indicated that Set 5 and Set 6 were completed by only 183 and 128 subjects, respectively, am ong the 263 subjects. Since the majority of the 237 subjects finished all sets up to Set 4, pow er scores could be obtained over the warm-ups and for Set 1 through Set 4 w ithout prorata scoring, which m ay add invalid variance. Accordingly, the PowLet measure was obtained over Sets 1 through 4. It can be seen in Table 2 that most of the measures have reliabilities in the .70 to .89 range; only COMGO has a reliability as low as .64, and this is probably an under-estimate of the true reliability. Similarly, the skewness indices are in all cases within a range acceptable for discrim inating measures, the largest (absolute value) being 1.34 (for RESGO). The kurtosis of this measure is also on the large side, but not out of bounds for m easures with reasonable discrimination. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 57 TABLE 2 Descriptives of Test M easures am ong Whole Subjects (N=263) Mean SD Skewness Kurtosis Reliability (Alpha) GO Tasks RECGO 4.90 2.97 .71 .30 .70 IDENTGO 37.02 10.17 -.22 -.25 .85* COMGO 15.50 4.58 -.15 -.17 .64* REPGO 54.51 23.32 1.12 1.04 .89 RESGO 40.91 9.89 -.53 -.27 .85 Non-GO Tasks RECNO 8.20 3.71 .26 -.44 .72 IDENTNO 65.68 13.09 .35 .22 .82* COMNO 26.98 7.03 .19 -.36 .85* REPNO 8.46 3.81 .18 -.01 .71 RESNO 23.20 4.38 -1.34 2.13 .77 BackSpan 6.73 2.76 -.02 -.49 .81 PowLet 4.05 2.74 1.11 .99 .79 Topology 12.83 4.33 -.68 -.13 .84 * - The reliability is for separately timed subsets of the test. Standard errors for skewness is .15, and for kurtosis, .30. Four Expertise Subgroups On the basis of the subject's GO rank~an objective m easure of GO expertise and an indicator of the accumulated am ount of practice on GO— the subjects were divided into four subgroups. These subgroups are called, from the highest level of expertise to the lowest: 1. Professionals (N=20), 2. Experts (N=92), 3, Intermediates (N=89), and 4. Beginners (N=62). Professionals include all 20 professional GO players. Experts include GO players between amateur 5 dan and 7 dan, Interm ediates are between 2 dan and 4 dan, and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 58 Beginners are between 30 kyu and 1 dan. In reality, the players in the group of Beginners are more experienced than the term beginners m ight imply. Descriptive information of these subgroups are presented in Table 3. TABLE 3 M ean Age and M ean GO Rank w ithin Four Expertise Subgroups Age Avg. GO Rank N Professionals 34.10 (SD: 9.34) Pro 7 dan (SD: 2.24) 20 Experts 56.61 (SD: 11.18) 6 dan (SD: .60) 92 Interm ediates 57.70 (SD: 12.00) 3 dan (SD: .79) 89 Beginners 55.35 (SD: 15.09) 2 kyu (SD: 3.26) 62 Total 54.97 (SD: 13.71) 4 dan (SD: 4.47) 263 The N for group w ith the highest GO expertise, the Professionals, is small and the mean age for this group is low. The younger mean age reflects a restricted range for age, only 19 and 48 years. Age ranges of other groups are between 21 and 77 for Experts and Intermediates and between 15 and 83 for Beginners. Although the possibility of combining the Professionals with the Experts was considered, this classification was not deem ed wise because the age range for this group was so restricted and because these players would be at the top level of expertise for different reasons (in GO playing history) than the other experts. To attain the exalted position of professional, the players would have accumulated over 10 years of concentrated practice during which they devoted themselves almost exclusively to GO. Even the strongest R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 59 players in the group of Experts w ould have not experienced such years of intensive practice (Kasai, 1986). For these reasons it seemed best to analyze the Professional group in more detail in a separate study. The Professional experts are omitted in the main analyses of the present study (i.e., SEM analyses). They, as the female subsample, w ill be compared in studies that will follow the present study. Correlations Correlations were computed to obtain information for the total group and the subgroups. These correlations are provided in Table 4. TABLE 4 Correlations between Age and GO Rank within Four Expertise Subgroups Professionals .53* Experts .09 Interm ediates .05 Beginners -.01 Total (N=263) - .26** Amateur Players (N=243) .04 Note: **- Correlations are significant at the .01 level. * - Correlations are significant at the .05 level. Over the entire sample, the correlation between age and GO rank is -.26 (p<.05). This indicates that as age of the subjects gets higher, the level of GO expertise becomes lower, but this occurs only in the Professional subgroup. When this subgroup is omitted (N=243), the correlation is only .04. This R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 60 m eans that when the younger professionals are not entered in the sample, there is virtually no relationship between age and GO rank. The correlation betw een age and GO rating was also close to zero in the subgroups of Experts (r=.09), Intermediates (r=.05) and Beginners (r=-.01). Both age and the level of GO expertise are im portant variables in this study; they are to be entered separately and simultaneously in analyses of interindividual difference in other variables. For purposes of this study, it is desirable that age and GO rank be correlated near zero in order to have a full distribution of GO expertise (practice) across all levels of age. Table 5 provides the correlations betw een GO rank and scores on the GO tasks over the entire sample and within the subgroups. In Table 6, the comparable correlations for the Non-GO tasks are provided. TABLE 5 Correlations of GO Tasks w ith GO Rank w ith in Four Expertise Subgroups RECGO IDENTGO COMGO REPGO RESGO Professionals -.15 -.03 .10 -.07 .30 Experts .08 .11 .07 -.03 .22* Interm ediates .16 .03 -.14 .18 .27** Beginners .21 .46** -.06 .11 .53** T otal (N=263) .51** .55** .20** .57** .71** * - Correlation is significant at the .05 level (2 tailed). ** - Correlation is significant at the .01 level (2 tailed). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 61 In Table 5 it is seen that the correlations between GO ratings and GO tasks are positive but usually low within groups, but substantial across groups. This indicates that tests designed to measure aspects of GO expertise are m easuring basic process abilities w ithin the GO domain. The highest correlation of .71 was obtained for RESGO. This supports the observation of professional GO players that this test directly reflects levels of GO playing skills. For RESGO within all groups and for IDENTGO in the Beginners group, the correlation w ith GO ratings are significantly larger than zero. These results indicate that the relatively small differences in expertise within groups are most sensitively detected w ith RESGO and IDENTGO. RESGO is a measure of a crucial part of playing GO~making a stone placement that best anticipates possible contingencies in subsequent stone placements of the opponent and oneself. It is reasonable that this m easure of a crucial aspect of playing the game should be most sensitive to the small differences in GO ratings w ithin groups. Similarly, it is reasonable that IDENTGO, a m easure of quickness in spotting Atari, which also is crucial to playing the game, should be sensitive to differences in GO ratings. It is reasonable, too, that this should be true primarily only among Beginners, for spotting the condition of Atari is a skill that m ust be learned early in developing expertise— rather like learning to not put one's pieces in position to be forked in chess. Beyond the beginning stage of learning the game, individual differences in these basic skills are not indicative of ratings of overall playing ability. It can be seen in Table 6 that the GO rank regulated by the Japanese GO Association correlate near zero with measures of attentional, speed and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 62 working memory that are not couched within the game of GO. This is as expected. The exceptions are the - .45 correlation between RECNO and GO ratings within the sm all group of Professionals and the - .23 correlation between RESNO and GO ratings w ithin the Intermediate subgroup. Neither correlation indicates an hypothesis likely to be confirmed. TABLE 6 Correlations of Non-GO Tasks with GO Rank within Four Expertise Subgroups RECNO IDENTNO COMNO REPNO RESNO Professionals - .45* -.05 -.13 -.05 -.29 Experts -.01 .04 .05 .03 .03 Interm ediates -.08 -.03 -.12 -.06 - .23* Beginners -.18 -.05 -.01 -.07 -.02 Total (N=263) -.10 -.02 -.07 i o w .05 BackSpan PowLet Topology Professionals -.31 -.25 -.17 Experts .16 -.09 .02 Interm ediates -.07 -.03 .09 Beginners .02 .04 -.14 Total (N=263) .14* .07 .10 * - Correlation is significant at the .05 level (2-tailed). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 63 Tables 7 provides correlations between age and PowLet and other established Gf measures (i.e., RESNO, BackSpan, Topology). Table 8 presents intercorrelations among RESNO, BackSpan, Topology and PowLet. These are provided here to provide information pertaining to w hether PowLet developed for this study (within the Japanese language) is measuring Gf reasoning in the manner of previous studies. It can be seen that, indeed, the pattern of correlations for PowLet is sim ilar to that seen in previous studies. Across the subgroups, age correlations for PowLet are ordered almost identically with Topology, which is the most salient Gf m arker among the established Gf measures presented here. In general, in these measures higher GO expertise is associated with smaller degrees of negative correlations with age. In Noll and Horn (1997 in press), the correlation between age and the Power Letter Series test was found to be -.47 for an adult sample between 22 and 92 years of age. The correlation of -.57 between age and PowLet over the total sam ple in this study is som ewhat higher, but w ithin a range of w hat can be expected under different sampling. TABLE 7 Correlations of Gf M easures w ith Age w ithin Four Expertise Subgroups RESNO BackSpan Topology PowLet Professionals - .56* -.44 -.07 - .47* Experts - .37** - .26* - .26* - .64** Interm ediates - .33** - .28** - .37** - .50** Beginners - .47** - .52** - .51** - .75** Total - .38** - .36** - .34** - .57** * - Correlations are significant at the .05 level (2-tailed). ** - Correlations are significant at the .01 level (2-tailed). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 64 TABLE 8 Intercorrelations among Gf Measures within Four Expertise Subgroups RESNO BackSpan Topology BackSpan Professionals Experts Interm ediates Beginners Total .16 .29** .28** .35** .30** Topology Professionals -.02 .15 Experts .26* .43** Interm ediates .23* .21 Beginners .48** .42** Total .31** .34** PowLet Professionals .20 .57** .42 Experts .41** .50** .54** Interm ediates .39** .41** .43** Beginners .44** .56** .55** Total .41** .49** .51** * - Correlations are significant at the .05 level (2-tailed). * * - Correlations are significant at the .01 level (2-tailed). Table 8 indicates that PowLet has relatively high positive correlations (.41-.57) with BackSpan and Topology across the four subgroups. In general, the results of Table 7 and Table 8 suggest that the PowLet measure, developed within the Japanese language, is a good measure of Gf reasoning. Further evidence on this issue will be provided by the factoring of Gf measures in the section on the structural equation m odeling (SEM) analyses. Intercorrelations among all GO measures and am ong all Non-GO measures, and the correlations between age and all GO and Non-GO measures across the expertise-subgroups as well as over the total sam ple are R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 65 shown in Appendix B. Detailed analyses of these correlations are important and will be carried out in separate studies to provide information on the relationships between age, expertise, and Gf abilities m easured w ith test performance. The focus of the present study is on analyses w ith the SEM. It should be noted at this point, however, that some of the correlational information on Professionals (show n in A ppendix B) may not be generalizable to a larger population: the age range is highly restricted and the N for the group is small. ANOVAs Tables 9 and 11 present the means and standard deviations for GO tasks and Non-GO tasks within the four expertise subgroups. An ANOVA and a post hoc test were conducted to compare these means across the subgroups (See Tables 10 & 12). These analyses provide evidence of how GO expertise is associated w ith the levels of test perform ance w ithin and outside the dom ain of GO expertise (hypotheses 1 & 2) and indicate whether there is evidence of LT-WM associated w ith developm ent of expertise. The following summarize the principal hypotheses: 1. The means for GO-relevant tasks are monotonically related to level of expertise: Professionals > Experts > Intermediates > Beginners, and the differences betw een means are significantly different overall (weak hypothesis) and pair-by-pair in post hoc analysis (strong hypothesis). 2. The means of GO-irrelevant tests are not significantly different across the expertise subgroups. 3. As level of expertise increases from Beginners to Intermediates to Experts to Professionals, the am ounts retained in the GO-relevant R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 66 reproduction tasks will exceed the amounts typically found in studies of short-term working memory. 4. The average amount retained in GO-irrelevant tasks will not vary with expertise and will not exceed the am ounts typically found in studies of ST-WM. 5. In general, on all m easures indicating working memory w ithin the dom ain of GO expertise, the means will increase m onotonically with level of expertise and exceed the am ounts retained in working m emory for tasks not w ithin the dom ain of expertise. F scores and their significance and results from post hoc tests are provided in Tables 10 and 12. The Tukey was perform ed as a Post Hoc test. Each subgroup is represented by a num ber in the sum m ary table: 1. Professionals, 2. Experts, 3. Intermediates, and 4. Beginners. The designation 12 stands for the comparison between group 1 and group 2, that is, between Professionals and Experts. If the m ean difference for such a pairing is significant by post hoc analysis, an asterisk * is placed in the column under that heading. For all GO tasks, with one exception, means systematically decrease from highest am ong Professionals to lowest among Beginners: the exception is for COMGO betw een Intermediates (14.61) and Beginners (14.89). The overall ANOVA F is significant for all tasks. Post hoc tests show that the means for Professionals are significantly larger than the means for any other group on all five GO tasks, and that the means for Experts are significantly larger than the means for Intermediates on RECGO and RESGO and for Beginners on four out of five GO tasks (RECGO, IDENTGO, REPGO and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 67 TABLE 9 Mean Scores and Standard Deviations of GO Tasks within Four Expertise Subgroups RECGO IDENTGO COMGO REPGO RESGO Professionals 10.30 51.90 19.40 107.00 54.15 (N=20) (2.54) (4.09) (4.20) (10.20) (2.52) Experts 5.27 38.60 15.92 55.52 45.35 (N=92) (2.64) (8.87) (4.63) (20.01) (6.28) Interm ediates 426 36.66 14.61 49.54 40.81 (N=89) (2.32) (9.09) (4.32) (17.11) (7.58) Beginners 3.55 30.39 14.89 43.23 30.19 (N=62) (2.33) (9.01) (4.35) (14.63) (8.54) Total 4.90 37.02 15.50 54.51 40.91 (N=263) (2.97) (10.17) (458) (23.32) (9.89) Note: N um ber in parentheses ( ) indicates Standard Deviation. TABLE 10 ANOVA on Mean Differences of GO Tasks across Four Expertise Subgroups F(df) Significance Post Hoc Test 12 13 14 23 24 34 RECGO 41.31 (3,259) .00 ★ * * * * IDENTGO 32.41 (3,259) .00 ★ * * * * COMGO 7.05 (3,259) .00 * * * REPGO 72.98 (3,259) .00 * * * * RESGO 81.28 (3,259) .00 * * * * * * * - The m ean difference is significant at the .05 level. 1 - Professionals, 2 - Experts, 3 - Intermediates, 4 - Beginners R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 68 TABLE 11 Mean Scores and Standard Deviations of Non-GO Tasks within Four Expertise Subgroups RECNO IDENTNO COMNO REPNO RESNO Professionals 8.25 67.35 26.20 8.25 24.25 (N=20) (3.32) (12.55) (6.54) (3.91) (3.77) Experts 7.79 64.38 26.14 8.18 23.12 (N=92) (3.8 7) (12.51) (6.72) (3.64) (4.09) Interm ediates 8.31 66.34 27.76 8.96 23.47 (N=89) (3.53) (11.96) (6.81) (3.97) (4.31) Beginners 8.61 66.11 27.37 8.24 22.58 (N=62) (3.86) (15.59) (7.90) (3.79) (5.02) Total 8.20 65.68 26.98 8.46 23.20 (N=263) (3.71) (13.09) (7.03) (3.81) (4.38) BackSpan PowLet Topology Professionals 7.95 4.45 13.95 (N=20) (2.21) (1.72) (3.41) Experts 7.04 4.09 12.91 (N=92) (2.8 7) (2.77) (4.2 7) Interm ediates 6.44 4.23 13.43 (N=89) (2.75) (2.62) (4.38) Beginners 6.27 3.60 11.47 (N=62) (2.65) (3.10) (4.3 7) Total 6.73 4.05 12.83 (N=263) (2.76) (2.74) (4.33) Note: Number in parentheses ( ) indicates Standard Deviation. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 69 TABLE 12 ANOVA on Mean Differences of Non-GO Tasks across Four Expertise Subgroups F(df) Significance Post Hoc Test 12 13 14 23 24 34 RECNO .65 (3,259) .58 IDENTNO .51 (3,259) .68 COMNO .95 (3,259) .42 REPNO .75 (3,259) .53 RESNO .92 (3,259) .43 BackSpan 164 (3,259) .05 PowLet .83 (3,259) .48 Topology 3.15 (3,259) .03 * * - The m ean difference is significant at the .05 level. 1 - Professionals, 2 - Experts, 3 - Intermediates, 4 - Beginners RESGO): the exception in this case is again COMGO. These results provide supporting evidence for hypothesis 1. On Non-GO tasks, the differences between the means are neither systematically monotonic nor, with one exception, statistically significant. The exception is Topology, for which the order of mean differences is Professionals, Intermediates, Experts and Beginners, and the only pairwise significant difference is that between Intermediates (13.43) and Beginners (11.47). These results do not indicate that the null hypothesis is true, of course, but they are consistent with hypothesis 2. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 70 As described in Table 3, the m ean for age is smaller for the group of Professionals than for the other groups. This raises the question of w hether the high average performance of this group on GO-inbedded tasks is an age effect rather than an effect associated w ith development of expertise. To examine this possibility, the m eans were compared across the four subgroups after the variance of age was controlled. Specifically, the means of the residuals were compared across the subgroups after the linear effect associated with age was regressed out of the scores of all 13 measures. The results of these analyses are sum m arized in Tables 13,14,15, and 16. In Table 13, it is seen that for RECGO, IDENTGO, REPGO, and RESGO (again COMGO is the exception) the means of residuals are ordered from highest for Professionals to lowest for Beginners. For COMGO, Experts have a higher mean than Professionals, although the mean differences are not significant for any pair of subgroups for this task. Table 14 shows that for RECGO, IDENTGO, REPGO and RESGO, means for Professionals are significantly higher than the m eans for Intermediates and Beginners, and for RECGO and REPGO the means for Professionals are larger than for Experts. Also the Expert means are larger than the Beginner means on RECGO, IDENTGO, REPGO and RESGO, and larger than the Intermediate means on RECGO and RESGO. These results thus again provide supporting evidence for hypothesis 1. The results for Non-GO tasks (See Tables 15 & 16) indicate that w ith correction for age differences the Professionals have the lowest means for six of the eight tests (RECNO, IDENTNO, COMNO, REPNO, RESNO and PowLet). The post hoc analyses indicate that these differences are pairwise significant for COMNO and PowLet. Thus, when the positive influence of R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. TABLE 13 Means of Residuals (after Variance of Age is Controlled for) of GO Tasks RECGO IDENTGO COMGO REPGO RESGO Professionals 2.97 6.85 .51 29.05 8.41 (2.57) (4.53) (4.20) (14.55) (2.89) Experts .56 2.21 .69 2.85 4.82 (2.20) (7.55) (4.17) (15.74) (6.02) Interm ediates -.33 .69 -.45 -1.91 .53 (2.30) (7.72) (3.93) (13.96) (7.49) Beginners -1.31 -6.48 -.55 -10.85 -10.63 (2.18) (9.00) (3.70) (13.74) (8.14) Total - 1.1E-15 - 4.3E-16 - 5.4E-17 1.7E-15 - 8.6E-16 (2.50) (8.69) (4.00) (17.53) (9.37) Note: Num ber in parentheses ( ) indicates Standard Deviation. TABLE 14 ANOVA on Residuals (after Variance of Age is C ontrolled for) of GO Tasks F (df) Significance Post Hoc Test 12 13 14 23 24 34 RECGO 21.04 (3,259) .00 * * * * * * IDENTGO 22.10 (3,259) .00 * * * * COMGO 1.80 (3,259) .15 REPGO 39.49 (3,259) .00 * * * * * RESGO 73.42 (3,259) .00 * * * * * * - The mean difference is significant at the .05 level. 1 - Professionals, 2 - Experts, 3 - Intermediates, 4 - Beginners R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 72 TABLE 15 Means of Residuals (after Variance of Age is Controlled for) of Non-GO Tasks RECNO IDENTNO COMNO REPNO RESNO Professionals -1.12 -5.79 -4.98 -2.29 -1.49 (3.38) (13.16) (6.38) (3.83) (3.28) Experts -.31 -.71 -.51 -.12 .12 (3.82) (11.52) (6.21) (3.33) (3.80) Interm ediates .27 1.64 1.33 .76 .61 (3.39) (10.40) (5.95) (3.67) (4.07) Beginners .44 .57 .46 -.18 -.57 (3.71) (14.46) (6.88) (3.32) (4.46) Total -4.7E-16 - 8.8E-15 - 3.9E-15 - 9.1E-16 2.3E-15 (3.63) (12.14) (6.47) (3.55) (4.05) BackSpan PowLet Topology Professionals -.28 -1.96 -1.10 (2.01) (1.54) (3.48) Experts .44 .23 .26 (2.78) (2.19) (4.13) Interm ediates -.09 .49 .89 (2.64) (2.28) (4.08) Beginners -.42 -.40 -1.32 (2.29) (2.15) (3.82) Total - 2.2E-16 2.4E-16 1.1E-16 (2.58) (2.26) (4.07) Note: Num ber in parentheses ( ) indicates Standard Deviation. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 73 TABLE 16 ANOVA on Residuals (after Variance of Age is Controlled for) of Non-GO Tasks F (df) Significance Post Hoc Test 12 13 14 23 24 34 RECNO 1.33 (3,259) .27 IDENTNO 2.24 (3,259) .08 COMNO 5.79 (3,259) .00 * * * REPNO 4.39 (3,259) .01 * RESNO 2.03 (3,259) .11 BackSpan 1.56 (3,259) .20 PowLet 7.96 (3,259) .00 * * * Topology 4.36 (3,259) .01 * * - The mean difference is significant at the .05 level. 1 - Professionals, 2 - Experts, 3 - Intermediates, 4 - Beginners young mean age is rem oved, Professionals are found to score lower than others on the basic m easures of Gf and the component ability of comparing patterns under highly speeded conditions. With regard to acquisition of LT-WM w ith development of expertise, the mean for the 10 items of REPGO is 107 for Professionals (Table 9). This means that on average Professionals remembered both color and positions of 10.7 stones, in a target GO pattern that was displayed for five or eight seconds. This is considerably more than the four plus or m inus one that would be typical for recall of visual m aterials under these time restraints. It is even R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 74 larger than the seven plus or m inus retrieval expected for sym bolic materials such as numbers: this result is thus indicative of acquisition of LT-WM. In order to realize the level of retrieval realized by the Professionals, it is necessary to remember stone positions on a GO board of 11 x 11 grids, providing 121 possible positions. The Professionals' retrieval of an average of 10.7 placements exceeds the average of 5.6 placements by Experts by almost double. This indicates truly outstanding performance by the Professionals on a very dem anding memory task in the immediate situation. It thus suggests that a very different quality of w orking memory capabilities has been acquired by the Professional players. As can be seen in Tables 9 and 11, both Professionals and Experts perform ed significantly superior to Intermediates and Beginners on RECGO and RESGO, but not for RECNO or RESNO. Also, both Professionals and Experts outperform ed Beginners for REPGO, but not for REPNO. High performance of Professionals and Experts on RECGO suggests that higher expertise in GO is associated w ith higher capabilities in recognizing GO- relevant information, while dealing w ith distractions of the kind that is not directly related to the practices of GO. H igh performance of Professionals and Experts on RESGO indicates the larger am ounts of information GO players can maintain, evaluate and m anipulate in the space of w orking memory. Higher expertise in GO is associated w ith larger amount of GO-relevant information retained and retrieved in the working memory. Thus, Hypotheses 4 and 5 are supported by data: as level of expertise increases, the functioning of the player's working memory expands, as suggested by the LT- WM theory by Ericsson and Kintsch. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 75 To summarize, hypotheses were supported w ith regard to the mean score differences across the subgroups. Exclusively within the dom ain of GO expertise, means of the test scores are ordered in the same m anner as mean levels of GO expertise (practice) of the subgroups. Professionals have the highest GO expertise and the highest means for the tests in the dom ain of GO expertise, while the Beginners have the lowest GO expertise and the lowest means. Even though the m ean comparison of the residuals controlling for age indicated that Professionals were disadvantaged in their Gf reasoning and abilities to compare patterns under highly speeded conditions, the means in the dom ain of GO expertise were in the sam e order as when age was not controlled. For the residuals, the order of the m ean scores rem ained the same for four out of the five GO-relevant tasks as the order of means for raw scores. Hypotheses were also supported w ith regard to the acquisition of LT- WM among advanced experts. Data indicated that Professionals and Experts are superior to Interm ediates and Beginners in their abilities to attend, select, maintain, evaluate, change, and retrieve GO-relevant inform ation in and from the space of w orking memory. The superior memory perform ance of Professionals and Experts was realized in immediate situations and under dem and for dividing attention. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 76 Chapter IV: Results Part 2 Analyses w ith the Structural E quation M odeling (SEM) using a Lisrel This Chapter reviews the results from the 3-step structural equation modeling (SEM) analyses using a Lisrel (Joreskog & Sorbom, 1996). As discussed earlier, because of the restricted age range and small N, Professionals were not included in these analyses. There are three distinct sets of analyses. First, invariance of the m easurem ent model was tested. Second, factor means were compared across the three subgroups representing levels of practice (expertise). Third, structural relationships and factor means of the structural model were compared across the three expertise-specified subgroups. Mean comparisons based on SEM are different from mean comparisons using ANOVA at least in that m easurem ent errors are removed from the factor means in the SEM analyses. Two submeasures were entered to define RECGO— RECGOl IS and RECG07S— based on the difference in the duration of the presentation of stimuli (i.e., 11 seconds and 7 seconds). Similarly, two submeasures were entered to define REPGO— REPG08S (8-second presentation) and REPG05S (5-second presentation). RECNOl, RECN02 and RECN03 were derived from RECNO, by random ly dividing the items of RECNO into the three subsets. In all, then, 17 variables were obtained to indicate eight common factors. These variables are described in Table 17, which provides the means and standard deviations for all variables in each of the three comparison groups. The indices of variances, skewness and kurtosis for the measured variables indicate whether or not the variables discriminate well and have R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 77 symmetric distributions. W ithin-group data in Table 17 suggest that the 17 variables are generally symmetrically distributed and do discriminate well, although w ithin the Expert group there is high kurtosis on REPG08S and in the Beginner group there is high kurtosis for REPG05S and RESGO. TABLE 17 Descriptive Inform ation (Means, Standard Deviations, Skewness & Kurtosis) on 17 M easured Variables w ithin Three Expertise Subgroups Experts Interm ediates Beginners M (SD) S K M (SD) S K M (SD) S K RECGOllS 2.60 (1.65) 3 4 - 3 7 222 (1.43) .67 .10 1.84 (1.46) .88 .46 RECG07S 2.67 (1.48) 2 9 -2 6 203 (134) .17 -5 6 1.71 (131) .70 28 IDENTGO 38.60 (8.87) -.06 -.62 36.66 (9.09) - 5 3 .70 3039 (9.01) - 2 3 - 2 9 COMGO 15.92 (4.63) 27 -.90 14.61 (432) -.72 .69 14.89 (435) - 2 3 -2 1 REPG08S 26.40(12.08) 1.18 235 2355 (9.83) 57 .40 2150 (9.10) .71 1.70 REPG05S 29.12 (9.70) .73 .90 25.99 (8.98) .42 -.01 21.73 (759) .90 220 RESGO 4535 (628) -1.01 1.74 40.81 (758) -.15 -.47 30.19 (854) 2 2 59 RECNOl 333 (1.66) .17 -.08 3.18 (1.63) .43 -3 7 3.61 (1.80) 24 -.78 RECN02 225 (151) .48 - 2 3 230 (127) .18 -5 7 234 (138) .41 -.78 RECN03 222 (1.49) 5 0 -.38 283 (1.40) .08 -2 9 2.66 (1.49) 2 4 - 2 9 IDENTNO 6438(1251) .41 - 2 7 6634(11.%) 52 .81 66.11(1559) 2 2 20 COMNO 26.14 (6.72) .08 33 27.76 (6.81) .09 -.66 2737 (7.90) 24 -.83 REPNO 8.18 (3.64) -.02 - 5 2 8.% (3.97) .44 .11 824 (3.79) 2 2 20 RESNO 23.12 (4.09) - 126 153 23.47 (431) -1 2 2 1.49 2258 (5.02) -1 5 2 2.86 BackSpan 7.04 (2.87) - 3 0 -.44 6.44 (2.75) .13 - 3 9 6.27 (2.65) 35 -.03 PowLet 4.09 (277) 120 1.44 423 (262) 1.03 .85 3.60 (3.10) 132 1.01 Topology 1291 (427) - 5 3 -.43 13.43 (438) -1 2 2 1.18 11.47 (437) -.12 -.70 Standard Errors 25 50 26 51 30 50 Note: M=Mean; (SD)=Standard deviation; S=Skewness; K=Kurtosis. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 78 SEM Step 1: Invariance of M easurem ent All m easured variables were standardized over the entire sample of 243 subjects. This scaling produced variables with standard deviations equal to one and means equal to zero. Differences in the means and standard deviations of the measured variables, which would result from different scoring patterns across the variables, were thus eliminated. The scaling thus provides a basis for comparison across subgroups and betw een different variables. It also ensures that each variable has a comparable degree of contribution to the m easurement model. The m easurem ent m odel was tested for invariance through multiple- group analyses over three subgroups— Experts, Intermediates, and Beginners (total N of 243 subjects). The m oment matrices of the three expertise subgroups were simultaneously analyzed. No uniquenesses (errors) of the 17 m easured variables were allowed to intercorrelate in any subgroup. The chi- square overall was 351.24 based on 293 degrees of freedom. The standardized root m ean square of the residuals was .08; the comparative fit index was .96. For an a priori model, not adjusted by post hoc m anipulations, this fit is rather good. Although the fit was good, standardized residuals and modification indices were examined in order to investigate whether the a priori m easurement m odel left some part of the multi-sample data poorly accounted for. The examination was directed at exploring if there were systematically large residuals of the same direction and systematically large modification indices across the three subgroups. These statistics indicated that relatively large residuals remained in the following four pairs of variables: 1. REPG08S and REPNO, 2. COMNO and RECNOl, 3. COMNO and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 79 RECN02, and 4. Topology and REPNO. A test on chi-square change with 1 degree of freedom was carried out to examine w hether or not the correlations between error-uniquenesses of these pairs w ould account for the residuals. The correlations between error-uniquenesses w ere in each case constrained to be equal across the three subgroups in order to ensure they would not affect the invariance of the m easurem ent model. This procedure is equivalent to specifying a small factor for each pair of correlated variables (while constraining the factor loadings to be equal across the subgroups). The tests on chi-square change w ith 1 degree of freedom suggested that the following three correlations of the uniquenesses are significant: 1. REPG08S and REPNO, 2. COMNO and RECN02, and 3. Topology and REPNO. With these uniquenesses correlated, the overall chi-square was reduced from 351.24 to 335.43 based on 290 degrees of freedom— a reduction of 15.8 chi-square for 3 df. The standardized root m ean square residual was .07; the comparative fit index was .97. These goodness of fit statistics indicate a good fit of the m easurem ent model. Table 18 provides the factor loadings, constrained to be equivalent across the three subgroups and Tables 19 contains variances and covariances for uniquenesses. Variances and covariances for the common factors are shown in Table 20 for the three subgroups. Factor loadings are usually significantly different from zero. The exceptions are Factor 8 (GO) loadings on RECGOllS, RECG07S, COMGO, and REPG05S. FI through F8 in the Tables stand for latent variables, or factors: FI is GO-Attentiveness, F2 is Non-GO- Attentiveness, F3 is GO-Gs, F4 is Non-GO-Gs, F5 is GO-SAR, F6 is Non-GO- SAR, F7 is Gf and F8 is GO. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 80 Factor 8, GO factor, was constructed to represent the variability on the overall GO expertise within subgroups; it was expected that this factor would significantly load on all seven GO measures. However, the results indicated that RECG011S, RECG07S, COMGO, and REPG08S did not have significant loadings. This result was examined by considering the differences betw een the GO measures. A picture arose. The GO factor had high loadings on (in order) RECG07S than RECGOllS and on REPG05S than REPG08S. These are the more dem anding tasks of the battery. This is because the subjects were given shorter periods of time for these measures. The highest loading for GO was for RESGO, which is probably the most difficult of the GO measures. Factor 8 can be characterized as a factor which represents an ability to accomplish difficult tasks in the dom ain of GO expertise. TABLE 18 Factor Loadings of M easurem ent M odel Variables FI F2 F3 F4 F5 F6 F7 F8 RECGOllS (VI) 1.0= (.00) .13 (.10) RECG07S (V2) .76 (.10) .16 (.10) RECNO! <V3) 1.0= (.00) RECN02 (V4) .79 (.09) RECN03 (V5) .76 (.09) IDENTGO (V6) 1.0= (.00) S3 (.12) COMGO (V7) .96 (.09) .05 (.09) IDENTNO (V8) .91 (.09) COMNO (V9) 1.0= (.00) REPG08S (V10) 1.0= (.00) .06 (.10) REPG05S (VI1) .89 (.08) .33 (.10) RESGO (V12) .45 (.09) 1.0= (.00) REPNO (V13) 1.0= (.00) RESNO (V14) .86 (.12) BackSpan (V15) .63 (.08) PowLet (V16) 1.0= (.00) Topology (V17) .76 (.08) Note: Maximum likelihood estimates w ith standard errors in parentheses; "1.0=" - fixed parameter p for identification purposes. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. TABLE 19 Variances/Covariances of Error-Uniquenesses of Measurement Model Experts VI V2 V3 V4 V5 V6 V7 V8 V9 V10 V ll V12 V13 V14 V15 V16 V I7 VI .60* (.14) V2 .70* (.12) V3 .22* (.08) V4 .66* (.11) V5 .62* (.11) V6 .19* (.07) V7 .56* (.10) V8 .49* (.09) V9 -.08* (.04) .41* (.09) V10 .36* (.10) VI1 .41* (.09) V12 .30* (.12) V13 .11* (.04) .56* (.11) V14 .64* (.11) V15 .73* (.12) V16 .21* (.08) V17 -.11* (.04) .57* (.10) Intermediates VI V2 V3 V4 V5 V6 V7 V8 V9 V10 V ll V12 V13 V14 V15 V16 V17 VI .38* (.11) V2 .63* (.11) V3 .29* (.08) V4 .48* (.09) V5 .64* (.11) V6 .33* (.08) V7 .44* (.09) V8 .37* (.08) V9 -.08* (.04) .37* (.08) V10 .23* (.07) V ll .37* (.07) VI2 .45* (.11) V13 .11* (.04) .93* (.16) V14 .93* (.16) V15 .81* (.13) V16 .28* (.08) V17 -.11* (.04) .73* (.12) (Table Continues) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 82 TABLE 19 (Continued) Beginners VI V2 V3 V4 V5 V6 V7 VS V9 V10 V ll V12 V13 V14 V15 V16 V17 VI .28* (.13) V2 .66* (.14) V3 .43* (.13) V4 .54* (.12) V5 .51* (.12) V6 .55* (.15) V7 .59* (.13) V8 .85* (.17) V9 -.08* (.04) .66* (.15) V10 .23* (.08) V ll .31* (.07) V12 .04 { 29 ) V13 .11* (.04) .42* (.14) V14 .71* (.15) V15 .60* (.11) V16 49* (.11) V17 -.11* (.04) .46* (.10) Note: M aximum likelihood estimates with standard errors in parentheses; * - Variances/Covariances are significant. TABLE 20 Factor Variances/Covariances of M easurem ent M odel Experts FI F2 F3 F4 F5 F6 F7 F8 FI .66* (.19) F2 .23* (.11) .78* (.15) F3 .46* (.11) .30* (.10) .63* (.14) F4 .35* (.10) .44* (.10) .57* (.10) .53* (.13) F5 .68* (.14) .16 (.11) .50* (.12) .35* (.11) .95* (.19) F6 .31* (.10) .29* (.09) .34* (.09) .28* (.08) .30* (.10) .30* (.11) F7 32* (.12) .27* (.10) .45* (.11) .37* (.10) .55* (.13) .37* (.09) .75* (.15) F8 .05 (.08) .15* (.06) .00 (.06) .03 (.07) .33* (.15) (Table Continues) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 83 TABLE 20 (Continued) Intermediates FI F2 F3 F4 F5 F6 F7 F8 FI .46* (.13) F2 .31* (.09) .62* (.14) F3 .30* (.09) .29* (.09) .56* (-13) F4 .24* (.09) .20* (.09) .48* (.10) .57* (.13) F5 .46* (.10) .38* (.10) .42* (-10) .32* (.09) .64* (.13) F6 .20* (.08) .35* (.09) .33* (.09) .38* (.09) .36* (.10) .14 (.13) F7 .35* (.09) .33* (.09) .26* (.09) .31* (.09) .47* (.10) .44* (.09) 57* (.13) F8 .07 (.06) .03 (.06) .07 (.06) .16* (.06) .08 (.11) Beginners FI F2 F3 F4 F5 F6 F7 F8 FI .61* (.18) F2 .12 (.12) .77* ( 20 ) F3 .23* (.10) .21* (.10) .32* (.13) F4 .14 (.12) .17 (.13) .55* (.13) .62* (20) F5 .43* (.11) .19 (.11) .28* (.10) .36* (.12) .51* (.14) F6 .27* (.12) .42* (.14) .34* (.11) .43* (-14) .28* (.12) .67* (20) F7 .37* (.13) 25 (.14) .39* (.12) .71* (.17) .53* (.13) .63* (-16) .86* (22) F8 .20 (.17) .01 (.17) .10 (.17) .28 (20) 1.68* (.45) Note: Maximum likelihood estim ates w ith standard errors in parentheses; * - Variances/Covariances are significant. Thus, an invariant m easurem ent model was obtained across the three expertise-specified subgroups. The analyses hereafter are consistently based on this invariant m easurem ent m odel supported by the multi-sample data. Significant factor loadings of Gf factor on BackSpan, Topology and PowLet indicate that these m easures effectively construct a Gf common factor. This suggests that PowLet, a developed measure, is one representative measure defining the Gf reasoning factor. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. SEM Step 2: Averages betw een Groups Next, the invariance of the latent m ean structures was examined, while introducing a constant variable to the m easurem ent model w ith which variable intercepts and factor means w ere estim ated. Three moment matrices for each of the three subgroups were sim ultaneously analyzed, while constraining all the factor loadings and variable intercepts to be equivalent across the subgroups. The three matrices were constructed from correlation matrices, the scaled m ean as well as the scaled standard deviations. As noted earlier, all the m easured variables were standardized to have a mean of zero and standard deviation of one within the entire group of 243 subjects. Means and standard deviations of all the m easured variables in each subgroup are, therefore, scaled in the forms of deviations from overall group (See Table 21 & 22). Although these scaled means and standard deviations were also used for moment matrices in the preceding analysis on m easurem ent invariance, they are presented here because means are of major concern in this analysis. Table 21 Scaled M eans of M easured Variables VI V2 V3 V4 V5 V6 V7 V8 V9 V10 V ll V12 V13 V14 V15 V16 V17 Groupl .213 334 -.012 -.030 -.229.294 .167-.088 -.128 .214 .324 383 -.078 .002 .150 .027 .041 Group2 -.028-.111-.098.008 .187 .091-.128.061 .101-.052-.010.104 .124 .082 -.067.075 .158 Group3 -277-337.158 .034 .072 -367 -.065.044 .045 -243 -.466-1.016-.063-.120-.126-.148-288 Total .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 .000 Note: Groupl=Experts, Group2=Intermediates, Group3=Beginners R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Table 22 Scaled Standard Deviations of M easured Variables 85 VI V2 V3 V4 V5 V6 V7 V8 V9 V10V11 V12 V13 V14 V15 V16V17 Groupl 1.0671.027 .9871.0871.008 .9311.037.951.949 1.1281.036.662 .957 .926 1.033.987 .973 Group 2 .923 .935 .965 .918 .948 .953 .966 .909 .962 .918 .960 .800 1.044.977 .988 .936 .998 Group 3 .945 .9121.064.9931.010 .946 .973 1.1851.116.849 .811 .901 .9% 1.135 .9541.104.997 Total 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.001.00 1.00 1.00 1.00 1.00 1.00 1.00 Note: Groupl=Experts, Group2=Intermediates, Group3=Beginners The intercepts associated w ith factors, or the factor means, w ere set to zero in the group of Beginners, although they w ere freely estimated for Experts and Intermediates. The factor mean of zero among Beginners provided a basis w ith which factor means am ong Experts and Intermediates were estimated in the forms of either positive or negative deviations from zero. The positive deviation indicated that the factor had a higher m ean than that among Beginners, and the negative deviation, a lower mean. T-values associated with a factor mean indicated how significantly the m ean of Experts and Intermediates deviated from the basis of Beginners. The aim of this analysis was to test the following hypotheses: 6. Experts have higher factor means w ithin the domain of GO (i.e., GO-ATT, GO-Gs, GO-SAR and GO) than Intermediates, who in turn have higher factor m eans within the dom ain of GO expertise than Beginners. 7. Factor means outside the domain of GO (i.e., Non-GO- ATT, Non-GO-Gs, Non-GO-SAR and Gf) are not significantly different across the three expertise subgroups. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 86 8. For both GO-ATT and GO-SAR, Experts have significantly higher factor means than Intermediates, who in turn have significantly higher factor m eans than Beginners. Yet, for Non-GO-ATT and Non-GO-SAR, the factor means am ong Experts, Intermediates and Beginners do not differ significantly. Hypothesis 8 is to test the acquisition of LT-WM among advanced experts. A Lisrel was run. A very good fit of the invariant mean structure model (See Figure 4) to the multi-sample data was indicated with the chi- square of 382.02 based on 308 degrees of freedom. The corresponding standardized root mean square residual was .07; the comparative fit index was .95. Table 23 provides Factor loadings and variable intercepts. Table 24 contains estimated means of the eight common factors and the significance of the factor m ean differences w hen compared to the basis of Beginners. Table 25 provides variances and covariances of uniquenesses of m easured variables and Table 26 contains variances and covariances of factors within each subgroup. In Table 24, it can be seen that for the four factors outside the dom ain of GO expertise, neither Experts nor Intermediates have factor means that are significantly different from those among Beginners. Thus, data support hypothesis 7. For factors w ithin the dom ain of GO expertise, on the other hand, higher GO expertise is positively associated with higher means. Thus, data support hypothesis 6, although there is an exception for this trend in the factor, GO-Gs. The factor m ean of -.01 for GO-Gs among Intermediates indicates that the mean of Intermediates and Beginners are almost equivalent for this factor. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 87 Figure 4 M easurem ent M odel w ith M eans Estimated (to avoid clutter correlation am ong factors f l r e j n f .sh ow n Hpi-p) o v er Exgertisg-SperifipH G roups * Chi-Square = 382.016 based on 308 df G O A 3 Non-G ATT RECNOl RECN02J IDENTGO IDENTNO COMNO Constant REPNO RESNO BackSpan PowLet lopology R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 88 TABLE 23 Factor Loadings and Variable Intercepts of M easurem ent M odel w ith Means Estimated FI F2 F3 F4 F5 F6 F7 F8 Intercept Recgolls 1.0= (.00) -.11 (.20) -29 (.12) Recgo7s .76 (.10) -.01 (-17) -30 (.12) Recnol 1.0= (.00) .12 (.13) Recno2 .78 (.08) .09 (.11) Recno3 .75 (.08) .08 (.10) Identgo 1.0= (.00) 59 (.19) -56 (.12) Congo .99 (.10) .03 (.18) -.08 (.12) Identno .89 (.09) .03 (.12) Comno 1.0= (.00) .04 (.13) Repgo8s 1.0= (.00) -.20 (21) -24 (.11) Repgc6s .88 (.08) .15 (-16) -.49 (.10) Resgo .40 (.08) 1.0= (-00) -1.02 (.11) Rep no 1.0= (.00) -.10 (.13) Resno .86 (.12) -.08 (.12) BackSpan .63 (.08) -.13 (.10) PowLet 1.0= (.00) -23 (.14) Topology .77 (.08) -.18 (.11) Constant .0= .0= .0= .0= .0= .0= .0= .0= 1.0=(.00) Note: Maximum likelihood estimates with standard errors in parentheses; "1.0=" and ".0=" - fixed parameter p for identification purposes. TABLE 24 Factor Means and Significance o f M ean Differences from Beginners Experts Intermediates Beginners Within the Domain of GO GO-Attentiveness .72* (.29) 35 (.24) .00= (.00) GO-Gs .12 (.28) -.01 (23) .00= (.00) GO-SAR .72* (.28) 37 (.23) .00= (.00) GO 129* (.18) 1.01 * (.16) .00= (.00) Outside the Domain of GO Non-GO-Attentiveness -.17 (.16) -.14 (.16) .00= (.00) Non-GO-Gs -.16 (.16) .05 (.16) .00= (.00) Non-GO-SAR .05 (.15) 22 (.15) .00= (.00) Gf 28 (.17) 31 (.16) .00= (.00) Note: * - Factor mean is significantly different from beginners; Number in parentheses ( ) indicates Standard Deviation. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 89 TABLE 25 Variances/Covariances of Error-Uniquenesses of M easurem ent M odel w ith Means Estim ated Experts VI V2 V3 V4 V5 V6 V7 V8 V9 V10 V II V12 V13 V14 V15 V16 V17 VI .62* (.14) V2 .69* (.12) V3 .20* (.08) V4 .66* (.11) V5 .63* (.11) V6 .20* (.06) V7 55* (.10) V8 .51* (.09) V9 -.09* (.04) .39* (.08) V I0 .36* (.10) V I1 .42* (.09) V12 .24* (.08) V13 .11* (.04) .56* (.11) V14 .65* (.11) V15 .73* (.12) V16 .21* (.08) VI7 -.12* (.04) .56* (.10) Interm ediates VI < to < C O V4 V5 V6 V7 V8 V9 V10 V11 V12 V13 V14 V15 V16 V17 VI .38* (.10) V2 .63* (.11) V3 .28* (.08) V4 .49* (.09) V5 .64* (.11) V6 .33* (.08) V7 .42* (.08) V8 .38* (.08) V9 -.09* (.04) .36* (.08) V10 .22* (.07) V I1 .37* (.07) V12 .47* (.10) V13 .11* (.04) .94* (.16) V14 .93* (.16) V15 .81* (.13) VI6 .28* (.08) VI7 -.12* (.04) .73* (.12) (Table Continues) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 90 TABLE 25 (Continued) Beginners VI V2 V3 V4 V5 V6 V7 V8 V9 V10 V ll V12 V13 V14 V15 V16 V17 VI .28* (.13) V2 .65* (.14) V3 .42* (.13) V4 .53* (.12) V5 .52* (.12) V6 .50* (.13) V7 .60* (.13) V8 .87* (.17) V9 -.09* (.04) .63* (.15) V I0 .22* (.08) V ll .31* (.07) V12 .18 (.17) V13 .11* (.04) .42* (.14) V14 .71* (.15) V15 .60* (.11) V16 .50* (.12) V I7 -.12* (.04) .46* (.10) Note: Maximum likelihood estimates with standard errors in parentheses; * - Variances/Covariances are significant. TABLE 26 Variances/covariances of Factors of M easurem ent M odel w ith M eans Estim ated Experts Fl F2 F3 F4 F5 F6 F7 F8 FI 5 7 (.17) F2 2 4 (.10) F3 .42 (.10) F4 3 7 (.10) F5 5 9 (.13) F6 3 3 (.09) F7 5 2 (.12) F8 (Table Continues) .78 (.15) 28 (.10) .43 (.10) .17 (.11) 2 9 (.09) 2 7 (.10) .01 (.06) 5 9 (.13) 5 5 (.10) .46 (.11) 3 2 (.09) .42 (.10) 51 (.12) 37 (.10) 28 (.08) 38 (.10) -.04 (.04) .85 (.17) 3 2 (.10) 55 (.12) 31 (.11) 38 (.09) .05 (.05) .76 (.15) .06 (.05) .13 (.08) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 91 TABLE 26 (Continued) Interm ediates FI F2 F3 F4 F5 F6 F7 F8 FI .46 (.13) F2 JO (.09) .64 (.14) F3 3 0 (.09) 30 (.09) 55 (.13) F4 26 (.09) 21 (.09) .49 (.10) 38 (.13) F5 .46 (.10) 3 8 (.10) .43 (.10) 34 (.09) .66 (.13) F6 2 2 (.08) 36 (.09) 3 4 (.09) 3 7 (.09) 3 8 (.10) .12 (.13) F7 3 7 (.09) 33 (.09) 2 6 (.09) 30 (.09) 3 0 (.10) .42 (.09) 5 4 (.12) F8 -.07 (.05) .01 (.05) .04 (.05) .12 (.05) .03 (.08) Beginners FI F2 F3 F4 F5 F6 F7 F8 FI 55 (.17) F2 .12 (.11) .76 (20) F3 2 0 (.09) 21 (.10) 2 9 (.12) F4 .16 (.12) .17 (.13) 5 5 (.13) .63 (21) F5 3 8 (.10) .19 (.10) 2 5 (.09) 37 (.11) .47 (.12) F6 2 5 (.12) .43 (.14) 3 2 (.11) .43 (.14) 2 7 (.11) .66 (20) F7 3 4 (.13) 28 (.14) 3 7 (.12) .72 (.17) 51 (.12) .61 (.15) .80 (21) F8 -.08 (.11) .04 (.12) -.00 (.11) .04 (.12) 5 2 (21) Note: Number in parentheses ( ) indicates Standard Deviation. Tests on chi-square changes were perform ed in order to examine whether or not the m ean difference between Experts and Intermediates are significant. As indicated in Table 27, the following factors were found to show significant chi-square change w ith 1 degree of freedom when factor means were constrained to be equal between Experts and Intermediates: 1. GO-ATT, 2. GO-SAR, and 3. GO. Hence, it was indicated that for these three factors, Experts have significantly higher means than Intermediates. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 92 TABLE 27 Significant Factor Mean Differences between Experts and Intermediates Factor Chi-Square Change Experts Means Interm ediates GO-ATT 5.72 (1 df) .72 (.29) > .35 (.24) GO-SAR 4.47 (1 df) .72 (.28) > .37 (.23) GO 8.06 (1 df) 1.29 (.18) > 1.01 (.16) Findings are: 1. Experts have significantly higher means than both Intermediates and Beginners on GO-ATT, GO-SAR, and GO, 2. Factor means w ithin the dom ain of GO expertise are ordered from the highest among Experts through Intermediates to the lowest am ong Beginners, 3. GO-Gs, in which no significant difference in the m eans across the subgroups are found, differs in nature from other GO-related factors and 4. No significant differences were found for the factor means outside the dom ain of GO expertise. These results suggest that the probability does not hold that level of Gf component abilities reached in the dom ain of GO expertise is the function of prior levels of Gf a n d /o r the components of Gf. Indeed, GO players with different levels of GO expertise do not differ in their levels of Gf abilities outside the dom ain of GO expertise. Hypothesis 8, regarding the acquisition of LT-WM among advanced experts, was tested by comparing means of ATT factors and SAR factors across the subgroups. These factors are defined w ith abilities on short-term recognition w ith a distraction, short-term recall m em ory in immediate situations, and deep inductive reasoning w hile attending, maintaining, evaluating and changing information in the working memory space. As R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 93 hypothesized, in Tables 24 and 27, it can be seen that higher GO expertise is associated with significantly higher factor means for GO-ATT and GO-SAR, but not for Non-GO-ATT or Non-GO-SAR. Even after m easurem ent errors are considered w ith the SEM, these results are in line w ith the results from ANOVA analyses. Experts have significantly higher factor means for GO-ATT and GO- SAR than Intermediates, who in tu rn have significantly higher m eans than Beginners for these factors. Im portantly, this superiority does not apply to the performance outside the dom ain of GO. No significant differences across the subgroups were found in factor m eans outside the dom ain of GO expertise. These results suggest that as GO players acquire GO-spedfic skills, they expand or enhance working memory capabilities within the dom ain of GO expertise. This notion of the "expansion" or "enhancem ent" of working memory capabilities in a task domain is in support of the basic argum ent of the LT- WM theory by Ericsson and Kintsch. SEM Step 3: Relationship between A ge and Gf-Com ponents A Lisrel was run with the three moment m atrices, while constraining the following param eters to be equal across the three subgroups: 1. factor loadings, 2. variable intercepts, and 3. path coefficients. Equivalent path coefficients across the groups forced the invariance of the structural relationships across the groups. Disturbances of factors were correlated betw een a GO factor and a corresponding Non-GO factor. The moment matrices were again constructed w ith three sets of correlation matrices, scaled means and standard deviations. Factor means were set to zero am ong Beginners and were estimated am ong Experts and Intermediates. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 94 A fit of the model to the multi-sample data was indicated with the chi- square of 519.57 based on 411 degrees of freedom. The standardized root mean square residual was .11; the comparative fit index was .94. In order to test hypotheses 9 through 13, additional analyses followed. These were pursued by releasing a path at a time and testing the significance of the chi-square change w ith 1 degree of freedom. For instance, a path from age to GO-ATT was released in the Expert group, while constraining all other paths to be equal across the groups. The significance of the chi-square change was examined between the overall chi-squares before the release and after the release of the parameter. All the paths included in the model was systematically tested in this manner. W ith these analyses, it can be examined w hether or not and how relationships between age and Gf- Components differ across the three expertise-specified groups. The hypotheses 9 through 13 to be tested were: 9. Age-related declines of the three basic abilities of Gf are the largest among Beginners, the second largest am ong Intermediates, and the smallest among Experts w hen tasks are GO-relevant. 10. Age-related declines of the three basic abilities of Gf do not differ across the expertise subgroups when tasks are GO-irrelevant. 11. Age-related declines of Gf in the domain of GO are the smallest among Experts, the second smallest among Intermediates, and the largest among Beginners. 12. Age-related declines of Gf outside the dom ain of GO do not differ across the expertise-specified subgroups. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 95 13. Age-related declines of overall Gf are the smallest among Experts, the second smallest among Intermediates, and the largest among Beginners. Table 28 provides the results from chi-square change tests, while listing the paths that resulted in significant chi-square changes with their releases. TABLE 28 Significant Chi-Square Changes w ith 1 Degree of Freedom w hen a Path is Released to be Freely Estimated W ithin the Dom ain of GO Expertise Significant Chi-Square Change Experts Interm ediates Beginners Param eter GO-ATT <- Age GO-ATT <- Age 759 (1 df) 5.01 (1 df) Outside the Domain of GO Expertise Significant Chi-Square Change Experts Interm ediates Beginners Parameter Gf <- Age 5.18 (1 df) Non-GO-Gs <- Non-GO-ATT 4.13 (1 df) When all these significant paths were released, a good fit of the structural model to the m ulti-sample data was indicated with the chi-square of 497.66 based on 407 degrees of freedom. The release of the four paths (i.e., the loss of 4 degrees of freedom) resulted in the improvement of the overall chi-square at 21.9 points. The standardized root mean square residual was .08; the comparative fit index was .95. These goodness of fit indices indicate that R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 96 the structural model fits the multi-sample data well. Figure 5 provides a structural model with estimated path coefficients and their significance. The number of param eters released are lim ited to two both inside and outside the domain of GO expertise. This indicates that the level of GO expertise does not place substantial influence on the relationships between age and the common factors and the interrelationships among factors. Released paths are of major concern of this study. They provide information on how different levels of GO expertise affect age-related changes in Gf-components. Data indicated that Experts have the highest degree of age- related decline of GO-ATT (- .56). Moreover, it in turn causes Experts to have the highest indirect negative effects from age to GO-Gs, GO-SAR, and Gf within the domain of GO expertise. Nam ely, data suggest that age-related declines of Gf-Components and Gf w ithin the dom ain of GO expertise are the highest among Experts, instead of Beginners. Thus, the results obtained are quite contrary to hypotheses 9 and 11. Concerning aging declines of Gf basic abilities outside the domain of GO expertise, hypotheses 10 and 12 are not supported by data, either. Since Experts and Intermediates have higher path coefficients from Non-GO-ATT to Non-GO-Gs, Experts and Intermediates experience higher degrees of indirect aging effects on Non-GO-Gs, Non-GO-SAR and Gf than Beginners. The total effects of age on factors are com puted as in Table 29. In Table 29, it can be seen that the degree of total negative effects of age on factors within the dom ain of GO expertise are ordered from the highest among Experts through Intermediates to the lowest among Beginners. This might be an indication that Experts have acquired more to lose, or som ething that is analogous to "regression to the mean" might be in progress w ith age among Experts. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 97 Figure 5 Structural Model w ith M eans Estim ated (but to avoid clutter m ean effects are not show n here) over Expertise-Specified G roups * Chi-Square = 497.664 based on 407 df Age E=-.56 (.09) I =-.28 (.09) B= -. 20 (.08)| E3 = .08 (.05)# E RECGOllS RECNOl GO ATT Non-GO ATT RECN02 RECG07! RECNQ3 .63(.13J 58 i .14) -.32/.0I - Jl7 (.(6) - (.07) IDENTNO IDENTGO Non-GO Gs GO Gs GO COMNO COMGO (.09) 22 (.10) E13 i r REPNO REPG08S GO SAR Non-GO SAR RESNO REPGQ5S Ell- RESGO E rU - ,Q7(.09) f =37(.13) B = 28(.09f .62 (.12)— Topology BackSoan PowLet El 5 E16 El 7 Note: Paths that are significantly different are shown with its coefficients in Italic. Disturbances of factors and correlations between disturbances (D1 & D2; D3 & D4; D5 & D6) are not shown in this model; # - Path coefficients are "not" signficant. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 98 TABLE 29 Total Effects of Age on Factors Experts Interm ediates Beginners GO-ATT -.56 (.09) -.28 (.09) -.20 (.09) Non-GO-ATT -.23 (.06) -.23 (.06) -.23 (.06) GO-Gs -.64 (.08) -.48 (.08) -.44 (.08) Non-GO-Gs -.49 (.05) -.49 (.05) -.42 (.05) GO-SAR -.66 (.08) -.45 (.08) -.39 (.08) Non-GO-SAR -.45 (.05) -.45 (.05) -.42 (.05) Gf -.64 (.07) -.51 (.07) -.68 (.07) GO .08 (.05) .08 (.05) .08 (.05) Note: N um ber in parentheses ( ) indicates Standard Deviation. However, a different picture arises when the direct effect of age to Gf is considered. Intriguingly enough, age places significantly higher direct effects on Gf among Beginners (-.28) than am ong Experts and Intermediates (-.07). In addition, the direct effect of age to Gf among Beginners (-.28) is significant, while the corresponding figure among Experts and Interm ediates (-.07) is not. Considering the fact that the direct effect of age on Gf is for the residual of Gf after the process indirect effects have been removed, the difference between the Beginner group and the other two groups are understood as follows: 1. Age-related changes in ATT, Gs and SAR account for aging decline of Gf reasoning to a less extent for Beginners than for Experts and Intermediates. Aging changes in ATT, Gs and SAR factors leave a significant portion of aging changes in Gf reasoning unaccounted for. 2. Age-related changes in ATT, Gs and SAR explain virtually all age-related changes in Gf reasoning among Experts and Intermediates. It might be indicative that higher GO expertise is positively associated w ith clearer functional segm entation of Gf reasoning into its three major subcomponents. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 99 Because of the stronger direct effect of age on Gf, Beginners experience the strongest degree of total aging declines of Gf (-.68) through both GO and Non-GO factors. Thus, w ith this regard, hypothesis 13 is supported by data, although the causes of the reverse order found between Experts and Intermediates in term s of total effects of age on Gf is yet to be investigated. In addition to the path from age to Gf among Beginners, the following paths are not significant in the structural model: 1. Age -> GO (for all groups) and 2. Non-GO-ATT -> Non-GO-Gs (for Beginners). Put differently, all other paths in the model are significant. Factor m ean comparisons add more information to these issues. Factor means obtained in the present analysis are for the portions of the factors that are not accounted for by age, or means of factors when age differences across the subgroups are held constant. Comparisons between the results from the preceding analysis (sum m arized in Table 24) and the results from this analysis would be m ost informative. These comparisons provide an opportunity to examine how differently the levels of GO expertise (practice) is associated with the levels of Gf abilities between when differences in age variabilities across the subgroups are not considered and when they are considered. Hypotheses to be tested are: 14. Experts have significantly higher factor means within the dom ain of GO (i.e., GO-ATT, GO-Gs, GO-SAR and GO) than Intermediates, who in turn have significantly higher factor means within the dom ain of GO expertise than Beginners. 15. Factor means of basic abilities of Gf outside the domain of GO (i.e., Non-GO-ATT, Non-GO-Gs, Non-GO-SAR) are not significantly different across the three expertise subgroups. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 100 16. Experts have a significantly higher factor mean of Gf than Intermediates, who in turn have a significantly higher factor mean of Gf than Beginners. These hypotheses can be read in summary form that after considering total aging effects in Gf abilities, the order of factor means in the dom ain of GO expertise remains the same as the order in the measurement model (i.e., E - Mean > I - Mean > B - Mean), while the degrees of differences are intensified as will be indicated with their significances. This is expected because aging decline of Gf abilities in the dom ain of GO expertise are hypothesized to be negatively associated with the level of expertise. Factor m ean differences of Gf across the subgroups, too, will become significant when total aging effects are controlled for in the factors. This is expected because it can be reasoned that different degrees of aging declines of Gf components in the dom ain of GO result in different degrees of total aging declines of Gf reasoning. Factor means and their significant difference from Beginners are provided in Table 30, and factor means that are significantly different between Experts and Intermediates are show n in Table 31. Tests on chi-square change w ith 1 degree of freedom were conducted to examine the difference between Experts and Intermediates. In Table 30, it can be seen that both Experts and Intermediates have significantly higher means for GO-ATT, GO-SAR and GO than Beginners. Data in Table 31 further indicate that the mean difference between Experts and Intermediates are significant for GO-Attentiveness and GO. Except for the nonsignificant differences across the subgroups on GO-Gs and nonsignificant difference between Experts and Intermediates for GO-SAR, data are in accordance w ith hypothesis 14. Of major importance in these R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 101 findings is that mean difference between Interm ediates and Beginners turned out to be significant for GO-ATT and GO-SAR com pared to the nonsignificant difference in the previous analyses (Table 24). This suggests that the difference in the levels of GO-ATT and GO-SAR w idens when age differences are held constant. TABLE 30 Factor Means after Total Effects of Age are Considered and the Significance of Mean Differences from Beginners Experts Intermediates Beginners Within the Domain of GO GO-Attentiveness 1.14* (-29) .72* (.29) .00= (.29) GO-Gs -.02 (-35) -.07 (36) .00= (.40) GO-SAR 1.13* (-23) .76* (.23) .00= (.26) GO .93* (-21) .74* (.21) .00= (.21) Outside the Domain of GO Non-GO-Attentiveness -.15 (.16) -.07 (.16) .00= (.16) Non-GO-Gs -.11 (.13) .13 (.13) .00= (-13) Non-GO-SAR .09 (.13) .29* (-13) .00= (.13) Gf 31*' (-12) 39* (.14) .00= (.20) Note: * - Factor mean is significantly different from beginners; ( ) includes standard errors. TABLE 31 Significant Factor Mean Differences between Experts and Intermediates after the Total Effects of Age are Controlled for from Factors Factor Chi-Square Change Experts Means Interm ediates GO-ATT 9.26 (1 df) 1.14 (.29) > .72 (.29) GO 5.16 (1 df) .93 (.21) > .74 (.21) Non-GO-Gs 3.98 (1 df) - .11 (.13) < .13 (.13) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 102 R egarding hypothesis 15, significant differences were found between Intermediates and Beginners for Non-GO-SAR and betw een Interm ediates and Experts for Non-GO-Gs. In both cases Intermediates possess higher factor means. These findings are not in support of hypothesis 15. It appears that factor m ean differences that already existed in Table 24 have become more salient w ith the control of aging effects. It is probable, although as yet to be further investigated, that subjects in the group of Intermediates are selected to form the m ost advantaged group in term s of Gf basic abilities outside the domain of GO expertise. The major finding in the factor m ean analyses is that when total effects of age were controlled for, the means of Gf for both Experts and Interm ediates became significantly higher than the m ean of Gf am ong Beginners. Both Experts and Interm ediates were found to have significantly higher m eans of Gf reasoning than Beginners. Thus, data support hypothesis 16. It is a consequential finding that higher GO expertise is positively associated with significantly higher levels of Gf reasoning after groups are set to the "same"age. This m ight be an indication that high GO expertise is of help for GO players to m aintain high levels of Gf in the course of aging. Yet, again, it is also still probable that this finding include some reflection of a priori differences in factor means of Gf betw een Experts and Beginners and Intermediates and Beginners which already existed in Table 24. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 103 Chapter V: Conclusions Analyses were perform ed across expertise-specified subgroups on mean differences of test m easures, mean differences of factors, effects of age on factor means, effects of age on relations am ong factors, and indication of the acquisition of LT-WM am ong GO players who have developed advanced expertise. The analyses w ere principally directed at testing the working hypothesis that if the basic abilities of Gf reasoning were assessed within a domain of expertise, those abilities, and hence Gf, would not decline with age. Put differently, it was examined w hether or not frequent "practice" or "use" of the age-vulnerable cognitive abilities am ong experts has positive effects on Gf-components in the dom ain of expertise during the course of cognitive aging. Gf-components (attentiveness, cognitive speed, short-term apprehension and retrieval) were m easured both within and outside the domain of GO expertise. Test measures w ithin the dom ain of GO expertise were developed for this study and m easures from previous research were used to assess Gf abilities outside the dom ain of GO expertise. In order to measure Gf reasoning, Pow er Letter Series test with letters in the Japanese language was developed based on the established version using alphabets in the English language. It seems that the developed test m easures are indeed measuring what they were expected to. Means of test m easures were com pared with ANOVAs and means of factors w ere compared w ith the structural equation m odeling using a Lisrel. It was found that higher GO expertise is associated with higher means of test measures within the dom ain of GO expertise, but not of test measures outside R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. the domain. Similarly, factor means in the dom ain of GO expertise (i.e., GO- ATT, GO-SAR, and GO) were the highest am ong Experts, the second highest among Interm ediates and the lowest among Beginners, although means outside the dom ain of GO expertise were not associated with the levels of GO expertise. This indicated that as GO players have developed GO expertise, they have cultivated GO-related basic abilities of Gf as well. Importantly, no significant association was found between the level of GO expertise and factor means outside the dom ain of GO. This implied that high levels of prior Gf basic abilities (or Gf basic abilities outside the domain of GO expertise) is not a prerequisite either for high levels of Gf basic abilities which can be attained w ithin the dom ain of GO expertise or for the level of GO expertise itself a player can reach after the long, deliberate practice. A factor on processing speed, that is, GO-Gs provided a somewhat different pictures in this trend from other GO-relevant factors. In essence, data suggested that higher GO expertise is not associated w ith a higher factor m ean of GO-Gs. It appears that the acquisition of higher GO expertise does not enhance speed in identifying and com paring information of the kind that is specific to practices of GO. However, this result m ight also be ascribable to a particular condition of this study. For instance, as discussed earlier, research to date has indicated that expertise development is associated w ith increased levels of Gs in a dom ain of expertise at least to a degree of speed that is rew arded in completing real-life expert performance. Tests on Gs in this study might have violated the assumption that the degree of speed that is assessed is to be "realistic" in everyday GO practices. This is probably the case for the task of COMGO. It is possible that the ability to quickly compare GO diagrams which R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 105 are 90 degrees apart is not a fundamental ability of GO to effectively play the game, even though this type of ability is of great use for more successful performance on GO. Further study is needed which identifies practices of GO that need to be executed very quickly. W ith this information, we are able to test whether or not the developm ent of GO expertise is indeed unassociated with the enhancem ent of the speed in com pleting GO-related performance. With regard to the acquisition of LT-WM among advanced GO players, data in this study are in support of the theory of LT-WM by Ericsson and Kintsch (1995). It was suggested that the higher the level of GO expertise of a GO player is, the more he attends, selects, acquires, maintains, evaluates, changes and recalls GO-relevant inform ation in and from the working memory space. For instance, the means of REPGO, RESGO and their common factor, GO-SAR, were positively and significantly associated w ith the levels of GO expertise, while the corresponding means of REPNO, RESNO and their comm on factor, Non-GO-SAR, did not differ significantly across the expertise-specified subgroups. This indicated that high levels of working memory performance of advanced GO players w ithin the dom ain of GO expertise is not the function of prior levels of w orking memory outside the domain. W hile performing RESGO, GO players need to continuously evaluate information, while storing envisioned inform ation and retrieving stored information, when needed, to reevaluate it or compare it with newly acquired information. It seems that successful performance on this highly demanding task is possible for a GO player who has expanded his working memory space to include more information than seven plus or minus two alternatives. It is highly plausible that the acquisition of high expertise in GO is accompanied with the "enlargement" of working memory capacities within R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 106 the dom ain of GO. It seems that this enlarged w orking memory is what Ericsson and Kintsch called LT-WM. Moreover, a player with higher GO expertise is capable of successfully perform ing memory tasks while dealing w ith distractions. The high factor m ean of Experts on GO-ATT, in which the ability to divide attention between counting and recognition memory was assessed, indicated that GO players w ith higher expertise have larger capacity of working memory within which GO-relevant information and GO-irrelevant inform ation are successfully and sim ultaneously processed. This understanding is also in support of the basic notions of the theory of LT-WM by Ericsson and Kintsch. They argued that expanded working memory is equipped w ith an effective retrieval structure w ithin the dom ain of expertise which helps an expert to store information in a retrievable manner. In RECGO, it is likely that GO patterns were stored in LT-WM w ith a retrieval structure while dealing w ith a distraction task of counting. However, in order to further solidify this finding, the relationship betw een GO expertise and the level of performance on the distraction task still needs to be investigated. It seems probable, although not likely, that GO players w ith higher expertise are more concentrated on GO-relevant tasks and that they are less careful and less accurate on the distraction tasks. Thus, data supported hypotheses on the acquisition of LT-WM among GO players with advanced expertise. Yet, w hat this study has shown is mainly the description of the performances of GO players with different levels of GO expertise. These results are not enough to prove that the developm ent of advanced expertise causes one to acquire LT-WM in the dom ain of expertise. The greatest contribution of this study on this issue is, R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 107 however, in a design that enabled the contrasting of memory performance of GO players w ithin the dom ain of GO expertise with performance outside of the domain. Data indicated that the superior memory performance of Experts is not a "reflection" of superior memory abilities in general (i.e., outside the domain of expertise), but instead it represents "established" skills along with the acquisition of GO expertise. This finding is in support of an im portant notion of the LT-WM theory that expertise helps one to "expand" working memory. In order to identify the process of the acquisition of LT-WM and its relation to the developm ent of expertise, future research is needed in which this issue is addressed longitudinally. Concerning the effects of age on interrelations among factors, analyses were done w ith Lisrel structural equation modeling. First, in order to make sure that the same constructs were assessed across the expertise-specified subgroups, invariance of measurement was tested. With this invariant measurement model, both direct and indirect effects of age on factors were considered w ithin and outside the dom ain of GO expertise. It was found that a negative direct effect of age on GO-ATT is the highest among Experts (- .56) and the lowest among Beginners (- .20), and therefore, Experts have the highest negative indirect and total effects of age on GO-Gs, GO-SAR and Gf. These results are quite contrary to the major working hypothesis of this study in which Experts were hypothesized to have the smallest degrees of aging declines in Gf basic abilities within the domain of GO expertise and Gf within the domain. These results imply that even if the basic abilities of Gf reasoning were assessed within a dom ain of expertise, those abilities, and hence Gf, would decline w ith age. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 108 However, at least the following three findings of this study suggest that the expansion of this understanding should be reserved. First, although w hen relations of age with the factors were not considered, the difference in factor means between Interm ediates and Beginners were not significant for GO-ATT and GO-SAR, the difference became significant when effects of age on these factors were taken into account. Since in the latter analysis means of the factors are for the portions that are not affected by aging, w idened mean differences between Beginners and Intermediates for the GO-relevant factors m ight indicate that higher GO expertise helps a GO player to m aintain higher levels of Gf-components w ithin the dom ain of GO expertise through the course of aging during adulthood. Second, it was found that the direct negative effect of age on Gf is the largest among Beginners (-.28 for Beginners, -.07 for Experts and Intermediates). The direct effect of age on Gf was significant for Beginners, while it was nonsignificant for Experts and Intermediates. Because of this largest significant direct effect from age to Gf reasoning, Beginners also experience the largest degrees of total negative effects of age on Gf reasoning across the three expertise-specified subgroups. This indicated that low GO expertise, therefore, small am ount of practice on Gf abilities in the domain of GO expertise was associated with high degrees of aging declines of Gf reasoning. Third, both Experts and Intermediates had significantly higher factor means of Gf after total age effects are controlled for in every factor. That is, means of Gf reasoning, in which aging effects are controlled for, are significantly higher among Experts and Intermediates than Beginners, although their levels of Gf reasoning are not significantly different from each R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 109 other when aging effects are not considered. A possibility was pointed out that this expanded difference in factor means of Gf reasoning is somewhat the function of differences in the prior levels of Gf outside the dom ain of GO expertise, which has yet to be further investigated. However, it is also highly plausible to reason that this w idened difference in the levels of residual Gf reasoning between Beginners and the other more advanced GO players is attributable to the largest degrees of age-related declines of Gf reasoning among Beginners which was also found in this study. There rem ain a lot of questions to be investigated in future studies. For instance, first, data analyzed in this study are exclusively from the Japanese male subjects. It is still unknown if the results obtained in this study are generalizable to the female population or GO players from other ethnicities. Second, because of a small N as well as a restricted age range among professional GO players, they were not included in the analyses with the SEM in this study. Further research is needed in which a large num ber of professional players from a w ide age range are compared w ith other GO players of w ide expertise levels on their process of cognitive aging. Since professional players are the ones w ho have reached at the highest levels of expertise through years of deliberate practices, they are truly representative of so-called "experts." The study which includes those players w ould provide more profound insight into the issue of how expertise, or frequent "practice" of Gf basic abilities is associated w ith the maintenance of Gf abilities, the averages of w hich have been found in previous research to decline with advancing age. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 110 Third, in this study Intermediates were found to have the smallest degrees of aging declines of Gf reasoning abilities across the three expertise subgroups. Also, after the total aging effects are controlled for, they have significantly higher means of Non-GO-Gs and Non-GO-SAR compared to either Experts and Beginners. Further detailed investigation is needed to specify whether these are ascribable to selection effects of this study— i.e., Intermediates are selected to form the most advantaged group on Gf abilities outside the dom ain of GO expertise— or this is a finding that can be generalized to a larger population. Fourth, cautious correlational analyses on test measures need to be conducted. W ith p art/p artial correlation techniques, for instance, the relationship among age, GO expertise, and levels of Gf basic abilities within and outside the dom ain of GO expertise would be further clarified. For instance, if the variance of practice on GO, or GO ratings, is parted out from test measures w ithin the dom ain of GO expertise, part correlations between age and residual test m easures will provide m eaningful information on how "practice" influences relationships between age and test performance. Part/ partial correlation techniques will also allow one to control for selection effects, if there are any, am ong subjects with regard to the levels of prior Gf abilities. Fifth, w ith these (and other types of) thorough correlation analyses, it needs to be examined w hether or not positive correlation between GO expertise and the am ounts of aging declines in Gf basic abilities, which has been found in this study, is a generalizable phenomenon and, if it is, w hether or not the phenom enon is understood with a notion that is analogous to "regression to the mean." R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. I ll To sum up, essential findings of this study are: 1. High levels of expertise help GO players acquire expanded w orking memory capacities within the dom ain of expertise. 2. After aging effects are controlled for in factors, total aging declines of Gf is the largest am ong Beginners. 3. Even though Experts experience the largest am ount of aging decline of Gf basic abilities in the dom ain of GO expertise, the means of GO- relevant factors are still the highest among Experts when the aging effects were taken into account. 4. Attentiveness, in particular capacity for dividing attention, is m ost vulnerable to the factors that produce age-related loss of Gf reasoning. It has to be spelled out here that although finding 2 supports the hypothesis that aging declines of overall Gf are negatively associated w ith the am ount of "practice" on Gf abilities, it did not support the rationale w hich guided the generation of this hypothesis and the major working hypothesis of this study. That is, the largest degree of Gf decline among Beginners are not to be ascribable to their largest Gf declines in the dom ain of GO expertise as was hypothesized. Causes of and mechanisms behind the largest am ount of aging declines of Gf am ong Beginners will m ake an intriguing research topic for a next study. Although the major findings of this study do not fully support the working hypothesis of this study, they do suggest that frequent "practice" of Gf basic abilities have som e positive influence on the process of cognitive aging. In this study, it was found that the GO factor, which represented abilities to successfully solve difficult problems in the dom ain of GO R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 112 expertise, is not significantly influenced by age in any expertise-specified subgroup. This provides a desirable condition for GO players of all age groups and of all expertise subgroups. This finding suggests that GO is the game which older players can enjoy w ithout worrying about losing skills to successfully confront difficult problem s in the game. In this aging society, how to age successfully is a major concern. Learning GO and continuously enjoying GO until the end of life m ight be one of the ways to realize successful aging. The study on the developm ent of expertise is an area which has a rather short history. There are a lot of questions as yet to be investigated in this area of study. However, this study has shown that the theory on the development of expertise provides a very promising framework in which the causes of interindividual differences in cognitive aging are explored. My quest for obtaining clear com prehension on this consequential query has just taken the first step with this study. I will be very happy if I could continue inquiring into this im portant issue, while enjoying m y own experience of aging. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 113 figferengg Anderson, J. R. (1990). Cognitive psychology and its implications (3rd ed.). New York: W. H. Freeman. Anderson, J. R. (1993). Problem solving and learning. American Psychologist. 48(1). 35-44. Anderson, J. R., & Fincham, J. M. (1994). Acquisition of procedural skills from examples, loumal of Experimental Psychology: Learning. Memory, and Cognition. 20 (6), 1322-1349. Anooshina, L. J. (1989). Effects of attentive encoding on analytic and nonanalytic processing in implicit and explicit retrieval tasks. Bulletin of the Psvchonomic Society. 27 (1), 5-8. Baddeley, A. (1992). Working m em ory. Science. 255. 556-559. Baltes, P. B. (1987). Theoretical propositions of life-span developmental psychology: On the dynamics of growth and decline. Developmental Psychology. 23.611-626. Carpenter, P. A., & Just, M. A. (1989). The role of working memory in language comprehension. In D. Clahr and K. Kotovski (Eds.), Complex information processing: the impact of H ervert A. Simon (pp. 31-68). Hillsdale, NJ: Erlbaum. Chamess, N. (1981a). Search in chess: Age and skill differences, journal of Experimental Psychology: Human Perception and Performance. 7 (2), 467- 476. Chamess, N. (1981b). Visual short-term m em ory and aging in chess players, loum al of Gerontology. 36 (5), 615-619. Chamess, N. (1991). Expertise in chess: the balance between knowledge and search. In K. A. Ericsson and J. Smith (Eds.), Toward a general theory of expertise (pp. 39-63). New York: Cambridge University Press. Chamess, N., & Bosman, E. A. (1990). Expertise and aging: Life in the lab. In T. M. Hess (Ed.), Aging and cognition: Knowledge organization and utilization (pp. 343-386). New York: Elsevier Science. Cattell, R. B. (1963). Theory of fluid and crystallized intelligence: A critical experiment, journal of Educational Psychology. 54. 1-22. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 114 Christensen, H., Mackinnon, A., Jorm, A. F., Henderson, A. S., Scott, L. R., & Koren, A. E. (1994). Age differences and interindividual variation in cognition in com m unity-dwelling elderly. Psychology and Aging. 9(3). 381- 390. Ericsson, K. A. (1996). (Ed.) The road to excellence: the acquisition of expert performance in the arts and sciences, sports, and games. M ahwah, NJ: Lawrence Erlbaum Associates. Ericsson, K. A., & Smith, J. (1991). Prospects and limits of the empirical study of expertise: an introduction. In K. A. Ericsson and J. Smith (Eds.), Toward a general theory of expertise (pp. 1-38). New York: Cam bridge University Press. Ericsson, K. A., Krampe, R. T., & Tesch-Romer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review. 100 (3), 363-406. Ericsson, K. A., & Chamess, N. (1994). Expert performance: Its structure and acquisition. American Psychologist. 49 (8), 725-747. Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Review. 102. 211-245. Fogler, J., & Stem, L. (1994). Improving your memory: How to remember w hat you've starting to forget. Baltimore, Maryland: The Johns Hopkins Press. Glaser, R., & Chi, M. (1988). Overview. In M. Chi, R. Glaser and M. J. Farr (Eds.). The nature of expertise (pp. xv-xxvii). Hillsdale, NJ: Hove & London. Hertzog, C. (1991). Aging, information processing speed, and intelligence. In K. W. Schaie and M. P. Lawton (Eds.), Annual review of gerontology and geriatrics (pp. 55-79). N ew York: Springer. Horn, J. L. (1967). Intelligence— Why it grows, why it declines. Trans- Action. 5. 23-31. Horn, J. L. (1968). Organization of abilities and the developm ent of intelligence. Psychological Review. 75. 242-259. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 115 Horn, J. L. (1982). The aging of hum an abilities. In B. B. Wolman (Ed.), Handbook of developmental psychology (pp. 847-870). Englewood Cliffs, NJ: Prentice Hall. Horn, J. L. (1985). Remodeling old models of intelligence: Gf-Gc theory. In B. B. Wolman (Ed.), Handbook of intelligence (pp. 462-503). New York: W iley. Horn, J. L. (1986). Intellectual ability concepts. In R. L. Sternberg (Ed.), Advances in the psychology of hum an intelligence (pp. 267-300). New York: Wiley. Horn, J. L. (1987). A context for understanding information processing studies of hum an abilities. In P. A. Vernon (Ed.), Speed of information- processing and intelligence (pp. 201-238). Norwood, NJ: Ablex. Horn, J. L. (1988). Thinking about hum an abilities. In J. R. Nesselroade (Ed.), Handbook of multivariate psychology (pp. 645-685). New York: Academic Press. Horn, J. L. (1991). Discussion of issues in factorial invariance. In L. M. Collins & J. L. Horn (Eds.), Best methods for the analysis of change (pp. 114- 125). Washington, DC: American Psychological Association. Horn, J. L. (1997 in press). A basis for research on age differences in cognitive capabilities. In J. J. McArdle and W. Woodcock (Eds.), Human cognitive abilities in theory and practice . Chicago, IL: Riverside press. Horn, J. L., & Cattell, R. B. (1967). Age differences in fluid and crystallized intelligence. Acta Psvchologica. 2 6 .107-129. Horn, J. L., Donaldson, G., & Engstrom, R. (1981). Apprehension, memory, and fluid intelligence decline in adulthood. Research on Aging. 3(1), 33-84. Horn, J. L., & Hofer, S. M. (1992). Major abilities and development in the adult period. In R. J. Sternberg and C. A. Berg (Eds.), Intellectual development (pp. 44-99). New York: Cam bridge University Press. Horn, J. L., & McArdle, J. J. (1992). A practical and theoretical guide to m easurem ent invariance in aging research. Experimental Aging Research. 18(3), 117-144. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 116 Horn, J. L., & Noll, J. (1997). Hum an cognitive capabilities: Gf-Gc theory. Chapter 4 in J. L. Flanagan, P. I. Genshaft and Harrison (Eds.), Contemporary intellectual assessment (pp. 53-91). N ew York: Guilford Press. Hultsch, D. F., & Dixon, R. A. (1990). Learning and memory in aging. In J. E. Birren and K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd ed., pp. 258-274). New York: Academic Press. Hum phreys, L. G. (1985). General intelligence: A n integration of factor, test, and simplex theory, In Wolman, B. B. (Ed.), H andbook of Intelligence: Theories, measurement, and applications (pp. 201-224). New York: Wiley. Joreskog, K. G., & Sorbom, D. (1996). Lisrel 8: User's reference guide. Chicago: Scientific Software International. Kail, R., & Pellegrino, J. W. (1985). Hum an intelligence: Perspectives and prospects. N ew York: W. H. Freeman and Company. Kasai, K. (1986). Igo de atam a ga yoku naru hon (Becoming smart with GO). Tokyo, Japan: Shikai. Kausler, D. H., & Kleim, D. M. (1978). Age differences in processing relevant versus irrelevant stim uli in m ultiple-item recognition learning. loum al of Gerontology. 33 ,87-93. Kirasic, K. C., Allen, G. L., & Haggerty, D. (1992). Age-related differences in adults' m acrospatial cognitive processes. Experimental Aging Research. 18(1), 33-39. Logan, G. D. (1988). Toward an instance theory of automatization. Psychological Review. 95 (4), 429-527. Logan, G. D., & Etherton, J. L. (1994). What is learned during automatization? The role of attention in constructing an instance. loumal of Experimental Psychology: Learning. Memory, and Cognition. 22 (5), 1022-1050. Lohman, D. F. (1989). H um an Intelligence: A n introduction to advances in theory and research. Review of Educational Research. 59 (4), 333- 373. McArdle, J. J. (1995). A m odeling examination of Cattell's investment theory of hum an abilities. Proceedings European Congress of Psychology. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 117 McArdle, J. J., & Hamagami, F. (1991). Modeling incomplete longitudinal data using latent growth structural equation models. In L.M. Collins & J. L. Horn (Eds.), Best methods for the analysis of change (pp. 276- 304). W ashington, DC: American Psychological Association. M adden, D. J. (1985). A dult age differences in memory-driven selective attention. Developmental Psychology. 21 (4), 655-665. McDowd, J. M., & Craik, F. I. M. (1988). Effects of aging and task difficulty on divided attention performance, loumal of Experimental Psychology: H um an Perception and Performance. 14 (20), 267-280. McDowd, J. M., & Birren, J. E. (1990). Aging and attentional processes. In J. E. Birren and K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd ed., pp. 222-233). New York: Academic Press. M eredith, W. (1964). Notes on factorial invariance. Psychometrika. 29. 177-185. Meredith, W. (1992). On the misuse of manifest variables in the detection of m easurem ent bias. Psychometrika. 57.289-311. Meredith, W. (1993). M easurem ent invariance, factor analysis and factorial invariance. Psychometrika. 58.525-543. Mituschina, M., & Sats, P. (1991). Analysis of longitudinal covariance structures in assessment of stability of cognitive functions in elderly. Brain Dysfunction. 4 .163-173. Morrow, D., Leirer, V., Altieri, P., & Fitzsimmons, C. (1994). When expertise reduces age differences in performance. Psychology and Aging. 9. 134-148. Nesselroade, J. R. (1988). Some implications of the trait-state distinction for the study of developm ent over the life span: the case of personality. In P. B. Baltes, D. L. Featherman & R. M. Learner (Eds.), Life-span development and behavior (Vol.8). Hillsdale, NJ: Lawrence Erlbaum Associates. Nesselroade, J. R. (1991). Interindividual differences in intraindividual changes. In L. M. Collins & J. L. H orn (Eds.), Best methods for the analysis of change (pp. 92-105). W ashington, DC: American Psychological Association. Nettelbeck, T., & Rabbitt, P. M. A. (1992). Aging, cognitive performance, and mental speed. Intelligence. 16. 189-205. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 118 Noll, J., & Horn, J. L. (1997 in press). Age differences in processes of fluid and crystallized intelligence. Chapter 4 in J. J. McArdle and W. Woodcock (Eds.), Human cognitive abilities in theory and practice . Chicago, IL: Riverside press. OECD (1988). Ageing populations: The social policy implications. Paris: A uthor. Perfect, T. J., & Rabbitt, P. M. A. (1993). Age and the divided attention costs of category exemplar generation. British Toumal of Developmental Psychology, 1 1 ,131-142. Peris, T. T. (1995 Jan.). The oldest old. Scientific American , 70-75. Plude, D. J., Enns, J. T., & Brodeur, D. (1994). The developm ent of selective attention: A life-span overview. Acta Psychologica. 86. 227-272. Plude, D. J., & Hoyer, W. J. (1985). Attention and performance: Identifying and localizing age deficits. In N. Chamess (Ed.), Aging and human performance (pp. 47-99). John Wiley & Sons, Ltd. Rabbitt, P. (1965). An age-decrement in the ability to ignore irrelevant information, loum al of Gerontology. 20. 233-238. Rabbitt, P. (1990). Applied Cognitive Gerontology: Some Problems, methodologies and Data. Applied Cognitive Psychology. 4. 225-246. Rabbitt, P. (1993a). Does it all GO together w hen it goes? The Nineteenth Bartlett memorial Lecture. The Quarterly Toumal of Experimental Psychology. 46 (3), 385-434. Rabbitt, P. (1993b). Crystal quest: A search for the basis of maintenance of practiced skills into old age. In A. Baddeley, & L. W eiskrantz (Eds.), Attention: Selection, awareness, and control (pp. 188-230). Oxford: Clarendon Press. Rabbitt, S., & Abson, V. (1991). Do older people know how good they are? British Toumal of Psychology. 82. 137-151. Reitman, J. (1976). Skilled perception in GO: Deducing memory structures from interresponse times. Cognitive Psychology. 8. 336-356. Richardson, K. (1990). Understanding intelligence. Milton Keynes, Philadelphia: O pen University Press. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 119 Salthouse, T. A. (1985). A theory of cognitive aging. New York: Elsevier Science Publisher. Salthouse, T. A. (1987). The role of experience in cognitive aging. In K. W. Schaie and C. Eisdorfer (Eds.), Annual review of gerontology and geriatrics (Vol., 7, pp. 135-157). N ew York: Springer. Salthouse, T. A. (1988). Initiating the Formalization of theories of cognitive aging. Psychology and Aging. 3(1). 3-16. Salthouse, T. A. (1990). Cognitive competence and expertise in aging. In J. E. Birren and K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd ed., pp. 310-319). Salthouse, T. A. (1991a). Theoretical perspectives on cognitive aging. NJ: Lawrence Erlbaum Associates. Salthouse, T. A. (1991b). Expertise as the circumvention of hum an processing limitations. In K. A. Ericsson and J. Smith (Eds.), Tow ard a general theory of expertise: Prospects and limits (pp. 286-300). Cambridge; Cambridge University Press. Salthouse, T. A. (1992). The information-processing perspective on cognitive aging. In R. L. Sternberg and C. A. Berg (Eds.), Intellectual development (pp. 261-277). New York: Cambridge University Press. Schaie, K. W. (1990). Intellectual development in adulthood. In J. E. Birren and K. W. Schaie (Eds.), Handbook of the psychology of aging (3rd ed., pp. 291-309). New York: Academic Press. Schaie, K. W. (1994). The course of adult intellectual development. American Psychologist. 49 (4), 304-313. Schaie, K. W., & Willis, S. L. (1986). Can decline in adult intellectual functioning be reversed? Developmental Psychology. 22. 223-232. Schneider, W. (1985). Tow ard a model of attention and the development of automatic processing. In M. I. Posner & O. S. Marin (Eds.), Attention and performance XI (pp. 457-492). Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers. Shiffrin, R. M., Dumais, S. T., & Schneider, W. (1981). Characteristics of automatism. In J. Lonf & A. Baddeley (Eds.), Attention and performance IX (p p .). Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 120 Siegel, J. (1993). A generation of change: A profile of America's aging population. New York: Russell Sage Foundation. Snow, R. E. (1981). Toward a theory of aptitude for learning: Fluid and crystallized abilities and their correlates. In M. P. Friedman, J. P. Das, & N. CYConnar (Eds.), Intelligence and learning (pp. 345-362). New York: Plenum. Spearman, C. (1927). The abilities of man: Their nature and measurement. London: Macmillan. Sternberg, R. (1994). (Ed.) The Encyclopedia of Intelligence. New York: Cambridge Press. Strayer, D., & Kramer, A. F. (1990). Attentional requirements of autom atic and controlled processing. Toumal of Experimental Psychology: Learning. Memory, and Cognition. 16 (1), 67-82. Sugar, J. A., & McDowd, J. M. (1992). Memory, learning, and attention. In J. E. Birren, R. B. Sloane, & G. D. Cohen (Eds.), Handbook of mental health and aging (2nd ed., pp. 307-337). New York: Academic Press. Willis, S. L. (1990). Introduction to the special section on cognitive training in later adulthood. Developmental Psychology. 26. 875-878. Willis, S. L., & Schaie, K. W. (1986). Training the elderly on the ability factors of spatial orientation and inductive reasoning. Psychology and Aging. L 239-247. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 121 A ppendix A Examples of the Test M easures - W ithin G O M easures - RECGO Examples of Target GO Diagrams - RECGO Q1 ~ Q7 Q1 0 2 Q3 m o 0 4 £ - • m R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 122 Q 5 Q 6 1 & m Q7 H-HvT11111 1 H- m t l R eproduced with perm ission of the copyright ow ner. Further reproduction prohibited without perm ission. 123 Examples of 6 Choices for a Target GO Diagram - RECGO Q1 ~ Q7 1 : 1 . 3. -O * : f f l T " 1 H I — ■ t>- r - - i§= £ m 4. 5 * i= i 1 m i ML R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 124 1. 2. 3. -O m t 1 s i " T T t- (h- 7 ■ f := * = T f f t l. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 125 1 . <>- 4. o - - o 5. H - - ± 3 m “ ■ * " ™ ■ ---- h - - _ * — * " ----- — ■ := 3 _ _ . •— a ---- m — — ---- 1 -H m w ■ ;:qi t - “ " “ " * m ~ m ------ — :zi m d m R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 126 1 . : = l - -6- I * 4. i I m 5. 3 . f t i « • « « w ' R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. m i £ >=- m f i g i 1 8 5 . I ■ 4KK 4 I « « t ? , e > 7 f e v N T ? < f c $ v \ R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 129 1. 2. 3. J g i R t f S & . S t T ? , ^ < D ^ . - P i i H ) < ? > * ^ ^ • t ? < f £ ^ v \ R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. COMGO Examples - COMGO Q1 - Q10 2. i p i i I i s 1 A Pr' 1 * * IIM! J— L I I m111 i m R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 131 - —f — m f ¥ I 7 . 8. I I 9 . T r io. m o f m mi X \ & K : « > m £ * V ' T * < ) £ S v \ R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 132 IDENTGO Examples - IDENTGO Page 1 ~ Page 2 1 rM - ' 11 r i » ■v- -_t j 4^ _ i ' i i i ! I 1 ! • □ ; i i 1 ^ 1 : “ 1 t i i . < \ L* - 7 + o q * TT- 1 1 1 ' ! 1 1 1 1 i ! . ! i I 1 i i 1 M i i 1 ? * If 1 ! \ > i 1 i y 1 i , : i i i 1 J-l.i J R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 133 Mill I ! J ! 1 1 1 J III!: I I I 1 1 1 1 u 1 J 1 1 1 ~ 1~ n ~ r * 14I f : r — i i 1 J ! 1 1 i! l t i 1 1 r m I n 1 r r i xl M I L TT m r r r r r r LLLI..L! R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 134 REPGO Examples of Target GO Diagrams - REPGO Q1 ~ Q5 Q1 Q3 n ill Q5 0 2 0 4 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 135 An Example of an A nsw er Sheet - REPGO Q1 ~ Q5 s a s a f l g nm ( o ) s r f f l v ' - c , n&ztomx-m'o^xLr ( • ) . K o g ttg fc , c c o ^ ffliK < o ± K :S « L T < fc ^ ^ . 1. 2. 3. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 136 GO Diagram s for RESGO RESGO • • • O ••o o • • • o o o o o • O ' ' • • • O ' . OOOO' • O M T— ■ ' .O O rO fi • • o - o # # • • • • V D — O O o o o o • • • • • • 'O ' ' ' o % m o o m o ~ 8 ' » » « • • • 0 + 0 0 j o o < _ 0 ® f + 0 0 < • + + •0: . ^ + m o o m + u I 1 - f H - H ! 1 |- 1 • • « m ► V M - > § t r ± ! * * J — 1 j ► y h - i i 1 1 1 1 : 1 i : 1 1 ! I — i— l- ! I 1 u 1 \ 1 ! j ! R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 137 i i - 0 0 0 0 - O O 0 i i I I I o - f t ? - S : l i f e . — I . - 4 - R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 138 - M easures from Previous R esearch - RECNO Target Figures ( ) x O O H O % © O CD V o ♦ A 0 0 J R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 139 Choices of Figures for Recognition © c E G ^ o T O % o © @ " C \ f c l c i t , f c $ V \ R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 140 IDENTNO Examples of a Page of Japanese Letters - IDENTNO Pagel - Page2 r ft j £ b 2 * £ to f t f t j & * X . o f t & ft o ft ■ f r f t ft £ j 6 * b T £ tc f t f t t t * * ft ft < ft ft tt w f t £ I ' L o * X f t *5 j & » < 5 to * ft ft b ft 3 5 * ft * b f t ft t t ft ft ft to ft J & > o b b A f t ;i ■ fr f t t t L 1- f t It b ft f t -5 o ft V' ft A t l t t to ft & -5 < t ft £ ft r O t t f t b & a -5 ft b to ft $ * ft IZ A 4 o b r * ft o t t ft t> t l A ft b f t f t f t * o ft V' * > f t to * fc ft * ft ft k £ 3 f c L t > f t fC * » > * & i - * ft ft f t < to f b 5 ft x A t t S > ft ft * ft ft £> 3 f c $ > £ ft & f t L tt ft- ft * A • $ > V « f c * f t ft ft ti ft X 5 t3 fc $ ft * lt ft f t X ft -5 » ? f t ft ft k V' £ 3 & » b * ) X f t T > ft ft iz (2 £ ft < X * t t * > ft ft T fc £ ■ e - L o * f t 1" a f t * ■ f r £ ft * t 5 V Vs f t to £ * A , V ' £ $ t t V' f t * C f t ft ft £ to ft *5 ft it < ti> •fr ft- & £ > ■ f r ft f t to f i ft f t to t) «: c to * ) ft X £ f t -fr ft b ft X t i ft to «t £ ft * to to ft ft •fr o it * K IZ f t to t > vs ft to is X ft b ■ c ft iz ft t ft 0 ft t t f t ft ft ft ft ■ & to 3 d » ft tt ft ft * t t < t i to ft- o ft V' ft h £ (f T ft ti> ft ft ■ £ * £ £ » > & * * ❖ t l ft: £ » > ft ft vs t t «t ft ft ■ fr £ * * ft ft T ft Sr ft ft ia ft ■ & A to -r i t £ ft ■ £ b t t * ■f o X ✓ s . * t ft iZ 3 ft t t ft tt Sr V' X £ ft ft ft * £? * ft J: b * f ft t t * t t tt ft (C # > ti> ft £ t t ft- ft t? b ft * ft ft: 3^ tt A X £ ft . S » t t * * N £ tl X ft » ? ti> to ft ft * 1 2 Q * • 4 > IZ < D ft t t £ tt -e t ti < < > ft * ■ * t i «: ( r - * x > 5 lt is tt £ ft tt t t iz v^ f t ft 3^ £ T X f t f t V . £ X ft to b vs f t to is f t R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 141 r t t j b tt K * £ tt b > 7 r> * * 4 t t V' 1- fc t t t t £ tl ^ t t t o 6 A if - t tt to a < IZ £ £ tt to * V ' ' t* tt X * 6 * 9 * tt X t t tt X b to * o tt t t tt t 4 fc & £ b tt b * 6 <0 t t £ fc T V' £ X o & j G » <0 tt 4> tt tt tr tt 6 £ $ * £ fc $ tt tt fc © tt ft £ * t t tt ft tt T 4 ft & £ t t < 5 9 4 ta t t 4 I' $5 * tt A £ tt £ t t » 7 tt 4 * £ tt 9 t* to X > if fc * ft * t* tt (Z * * 9 i~ -£ 4 * » tt tt IZ ft * i- o X * tt X IZ * tt £ 5 tt T tt tt 9 < tt * tt tt tt < ( 4 . t t * 5 X X L * tt £ * O tp * tt b tt a* L tt fc X tt tt 5 * o Tt tt tt T tt ❖ J : if X ft tt 4 o ft * fC if tt * © 4 t t tt A. tt tt £ a* t t * tt tt £ < ❖ A tt £ £ t t * to X o ft tt tt tt ft < tt to o tt tt V' t t A ft tt tt & * * * i* tt tt tt tt if 4 £ ft tt * tt A £ * > £ t t 5 V' tt tl *. £ * 4 fc £ * Jb if tt tt * » 5 & » * tt fc * < * $ > tt * tt * tt f c o tt tt tt IZ o t t * tt $ ft ft * * t t • 5 tt T tt ;t £ t t * o tt ft • 7 * f c A * £ e if tt £ * * < L •f IZ if 4» t t tt V' * ft ❖ £ tt to < D tt o A (Z £ ft tt tt A tt tt tt t t t o tt tt tt tf tt tt ? 5 » ft K ft b * V' ia £ I' t t 0* A © * tt t t T Jb 6 t t t t 5 tt ft tt * € ■ 4 > 7 £ L tt tt tt A I' • 7 t t tt tt X tt o tt A ft £ t t * tt © * > it *c X. * tt tt t t tt t t £ f c tt tt tt 4 t t & £ tt t t A tt o ft * * fc to tt tt a t t * tt IZ X > L t t A * V' Z. tt tt t t t t £ tt 4 T IZ X. tt i> £ b * fr ft tt £ $ 5 tt to tt V - * # 3 & > tt tt tt 4 t t U tt 4 * b £ A tt fc £ *. tt t t tt tt fc * » 7 tt < T 4 fc tt tt ft tt t t R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. RESNO Examples of M azes - RESNO Q1 - Q5 2 . ± icisia»o-c\ itA ,- e < /c £ v \ T e n . fl6 £ fcv '-t?< * :£ v '. 3. i m * l 8 9 £ < O J » V 'j 6 £ a i § b T < * : £ v \ Z • " * ■ A k fc f e » r < f d iK R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission 143 v % . • * * . / v v \f v w \/ V k v v v i a k ■ A w ® ™ . * * . / \ / \ / \ / * / \ / \ / / \ . " \ / \ / \J \ / \ / \ t t t t t t v t t v A ^ A * r \ A a A A A A A / '• • , a .•••. A A X a * X * * X 3 * a * A a * / • • •• • • ; •• •• • • ; % % / •• • % ; • * \ , * \ .* • . • % . • • / v M • v v Y y # Y M v y w V v t t * t t * t t * k t t y 9 A A A A * * * A • A A • ’ • • • • “ * \ m % , * * , , * * # / * * * , / % * , , * * , / \ x\/xvyvvyy fe r. 9 9. a A A A A • • • • • * • • • • • \ • V V V V V v y v 9 jtt % /*. A /*. A -* *. •• \ / \ / \ w ; ; % .• v / v w V . • - y 0 A A $ * A A A « • . ; * . .* • . / • • / v / v / \ / \ / v w V V tt tt a • * • • « J R J R % : \ : • . .• * . • • * . / / \ / V M V V jtt V v w “ X X ¥. ¥. M P X X X X X t t * * • W » ^ • • a y m ¥ / Z Z frb R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 144 P o w L e t D t t t L 4 : ■ * L 2) it •) 5 5 V' 4i . -------- 3) J> ia i> u\ la ir' 5 fc 5 ;t ia -------------------------------------------- -------- 4) * < $ * £ ■ i ti ■ * L 5) -f * L + t * L * t Z * L t t . -------- tIU 6) U * * fc * . ~ L 7) o t r k t t * 4 * * e k e c c . _ _ 8) ^ It < # it < s » t a * t i ^ i . _ _ 9) *S # < It < I t # U - t L - t - t f c t f c t ) . _____ 1 0) V' it V' # It # L it L t> T *> . _ I D # L t* # L it # U ■* # L 1t . _ n i t i f c i S K t f A i a f i f a o i i ^ ' ' . ML 1 3 ) i t f c - t - t f - t - i s i » L . # w # < I t . ML 1 4 ) # L t" # # L - t U # L - f - t # . __________________________ 1 5 ) « r A , l f ' J 3 l t - « - £ O ' . ML 16) p # V' * i> is A, it . ML 17 ) it *> 'o r x t 1 8 ) H 2 b ^ i i 5 i 3 5 5 Ji 5 b . ML 1 9 ) - S » J » ^ ^ f f t i * * tr * » 2 0 ) * V ' 5 - 3 T £ i t i S j 6 » * > o T # < It. ML 2 1) < It C # < < £L 2 2) 45 & & * * < T o fc It £ # L f . fcL- 23) # < It * It £ # L # t * t t « t 4 ' J . ML 2 4 ) - t - a - - t f - t t - ' t ^ t > o f c t > o r . fck 25) It r L t t 1J . £L, 2 6) #> x # t' is L 5 S* . -------- ML 2 8) Ja * < l± £ L -fc it fc ._________________________________________ 2 9) # L I" S i . ML 3 0 ) 6 J : t i 6 J : i t i i t : * ) 5 ti 6 J: ■ ____ 3 1 ) % 4 t t L t * t L # t f c S . ML 3 2) tS A* # # < tt z tt £ L*-#tf*fc1-t>0 T . _____ 33) * t tf A i tr A & tr A ■ * L R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Topology Examples of Figures - Topology Q1 ~ Q10 1 . 2 . 3. 4. 5. 6. 7. 8 . 9. 10. & A O a A 1 1 s G . £ q e a © I S i a i © c f 3. & ■ P I p ] S L f e l §? R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission 146 Appendix B TABLE 32 Interconelations of GO Tasks within Four Expertise Subgroups RECGO IDENTGO COMGO REPGO IDENTGO Professionals .16 Experts .41** Intermediates .36** Beginners .18 Total .51** COMGO Professionals .15 38 Experts .44** .60** Intermediates .31** .59** Beginners 29* 31* Total .42** .55** REPGO Professionals .47* .08 .23 Experts .59** .43** .42** Intermediates .57** ST * .47** Beginners .50** 32* .38** Total .72** .60** .47** RESGO Professionals .10 .49* .46* .13 Experts .24* .41** .37** 35** Intermediates .37** .33** 23* .44** Beginners .18 .44** .11 32* Total .47** .58** .31** 55** * - Correlation is significant at the .05 level (2-tailed). ** - Correlation is significant at the .01 level (2-tailed). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 147 TABLE 33 Intercorrelations of Non-GO Tasks within Four Expertise Subgroups RECNO IDENTNO COMNO REPNO RESNO IDENTNO Professionals .60** Experts .37** Intermediates .29** Beginners .14 Total .29** COMNO Professionals .44 35 Experts .44** .49** Intermediates .11 .59** Beginners .13 .45** Total .25** 30** REPNO Professionals .16 .10 .03 Experts .40** .28** .40** Intermediates .43** .48** .38** Beginners .40** 33* .37** Total .39** .34** .36** RESNO Professionals -.07 .06 .02 -.06 Experts 20 .16 .19 .30** Intermediates 20 .18 .28** .08 Beginners 33* .33** 27* .50** Total .21** .22** .23** .26** BackSpan Professionals 22 .14 .43 .43 .16 Experts .31** .16 24* .35** .29** Intermediates .11 .05 .19 .11 .28** Beginners -.06 .37** 34** .37** .35** Total .14* .17** .29** .26** .30** PowLet Professionals .30 20 30 .49* 20 Experts .30** .33** .45** .37** .41** Intermediates .36** .37** .33** .47** .39** Beginners 24 .43** .59** .45** .44** Total .30** .37** .44** .43** .41** Topology Professionals 29 .41 -.03 .08 -.02 Experts .18 .14 25* .06 .26* Intermediates .30** .18 .30** .31** .23* Beginners .35** 27* .58** .43** .48** Total .26** 20** .33** .24** .31** (Table Continues) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 148 TABLE 33 (Continued) BackSpan PowLet PowLet Professionals .57** Experts .50** Intermediates .41** Beginners .56** Total .49** Topology Professionals .15 .42 Experts .43** 54** Intermediates 21 .43** Beginners .42** .55** Total .34** 51** * - Correlation is significant at the .05 level (2-tailed). ** - Correlation is significant at the .01 level (2-tailed). TABLE 34 Correlations between G O Tasks and Non-GO Tasks w ithin Four Expertise Subgroups RECGO IDENTGO COMGO REPGO RESGO RECNO Professionals .18 -.13 -.14 .16 -2 1 Experts .28** .34** .31** .19 .14 Intermediates .36** 2 T .28** .43** .19 Beginners .18 .14 .28* 26* -.02 Total .21** .19** .25** .19** .03 IDENTNO Professionals .38 .05 .10 .29 .23 Experts .24* .56** .53** .26* .11 Intermediates .26* .55** .46** .43** .21 Beginners -.01 .47** .49** 28* .05 Total .16* .43** .45** .24** .08 COMNO Professionals .13 .24 .27 .01 -.03 Experts .47** .57** .46** .43** .22* Intermediates .31** .53** .55** 37** 23* Beginners .31* .47** .50** .47** .24 Total .25** .41** .44** .25** .11 (Table Continues) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 149 TABLE 34 (Continued) RECGO IDENTGO COMGO REPGO RESGO REPNO Professionals .33 .05 -.09 .32 .33 Experts .33** .46** .20 .43** .19 Intermediates .29** .40** .23* .49** .26* Beginners .26* .27* .37** .42** .11 Total .23** .31** .21** .30** .14* RESNO Professionals .18 .38 .57** .10 .23 Experts .32** .37** .15 .28** .19 Intermediates .18 .30** .38** .39** .21 Beginners .23 .24 .35** 23 .11 Total .24** .31** .31** .27** .18** BackSpan Professionals .23 .33 .25 .44 .07 Experts .25* .29** .21* .29** .17 Intermediates .24* .13 .12 .32** 28** Beginners .14 .25 .43** .31* .16 Total .27** .27** .27** .33** .25** PowLet Professionals .38 .32 .38 .22 .37 Experts .53** .49** .44** .56** .37** Intermediates .50** .41** .30** .63** .47** Beginners .49** .40** .36** 57** .29* Total .44** .41** .36** .46** .33** Topology Professionals 39 -.02 -.06 .59** .03 Experts .36** .38** .26* .22* .32** Intermediates .25* .34** .23* .42** .36** Beginners .20 .17 .32* 53** .28* Total .29** .33** .25** .33** .33** * - Correlation is significant at the .05 level (2-tailed). ** - Correlation is significant at the .01 level (2-tailed). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 150 TABLE 35 Correlations of GO Tasks with Age within Four Expertise Subgroups RECGO IDENTGO COMGO REPGO RESGO Professionals -.19 -3 1 -.18 -.01 - 2 5 Experts - 56** - 53** - .44** - .62** - 31** Intermediates - 32** - 53** - .41** - .61** - .22* Beginners - .46** - .33* - 53** - .63** - .32* Total - 54** - 52** - .49** - .66** - 32** * - Correlation is significant at the .05 level (2-tailed). ** - Correlation is significant at the .01 level (2-tailed). TABLE 36 Correlations of Non-G O Tasks w ith Age w ithin Four Expertise Subgroups RECNO IDENTNO COMNO REPNO RESNO Professionals .03 .06 - 2 3 -.19 - .56* Experts -.15 - .40** - 39** - .42** - .37** Intermediates - .30* - 52** - 51** - .40** - .33** Beginners - .28* - 38** - 51** - .49** - .47** Total - 21** - 37** - 39** - .36** - .38** BackSpan PowLet Topology Professionals -.44 - .47* -.07 Experts - .26* - .64** - .26* Intermediates - 28** - 50** - 37** Beginners - 52** - .75** - 51** Total - 36** - 57** - .34** * - Correlation is significant at the .05 level (2 tailed). * * - Correlation is significant at the .01 level (2 tailed). R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Factors that influence students' investment of mental effort in academic tasks: A validation and exploratory study

PDF

Application of the graded response model to the revised Tennessee Self-Concept Scale: Unidimensionality, parameter invariance, and differential item functioning

PDF

A structural model of self-efficacy, goal orientation, worry, self-regulated learning, and high-stakes mathematics achievement

PDF

Adult achievement goal orientations and perceived ability: Impact on task choice, persistence and strategy use

PDF

Approaches to inference in ordinal multiple regression

PDF

A comparison of analogical and quantitative methods for teaching mechanical functions in science education

PDF

Inaccuracies in expert self -report: Errors in the description of strategies for designing psychology experiments

PDF

Individual motivation loss in group settings: An exploratory study of the social -loafing phenomenon

PDF

Cognitive and non-cognitive factors as predictors of retention among academically at-risk college students: A structural equation modelling approach

PDF

Collective efficacy, anxiety, creativity/innovation and work performance at the team level

PDF

Cognitive predictors of intention -behavior consistency in safer sex practices

PDF

Commitment to deliberate practice in the development of writing expertise: A survey of the motivational patterns of academic writers in Tier One research universities

PDF

Change beliefs and academic outcomes: A construct validation

PDF

Biosocial antecedents of schizophrenia-spectrum personality disorders: A longitudinal study

PDF

Examining the relationship between writing self -efficacy, writing performance and general achievement for third graders

PDF

Characteristics and cortical localization of subliminal visual priming

PDF

Beyond skill development: The effects of training and development on the attitudes and retention of employees

PDF

Development and validation of the Cooper Quality of Imagery Scale: A measure of vividness of sporting mental imagery

PDF

A comparison of values among counseling psychologists characterized as scientist-practitioners, practitioners, or scientists

PDF

Associations and mechanisms among attention deficit hyperactivity disorder symptoms, cognitive functioning, and drinking habits

Asset Metadata

Creator Takagi, Hiromi (author)

Core Title Cognitive aging: Expertise and fluid intelligence

Degree Doctor of Philosophy

Degree Program Education-Educational Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag education, educational psychology,OAI-PMH Harvest,psychology, cognitive,psychology, developmental

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Clark, Richard (committee chair), Guiton, Gretchen (committee member), Hocevar, Dennis (committee member), Horn, John (committee member), Peterson, David (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-279438

Unique identifier UC11352814

Identifier 9733148.pdf (filename),usctheses-c17-279438 (legacy record id)

Legacy Identifier 9733148.pdf

Dmrecord 279438

Document Type Dissertation

Rights Takagi, Hiromi

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA