www.psychspace.com心理学空间网
ON SEPTEMBER 19, 1955, one of the world's greatest psychologistspassed from the scene. It is quite appropriate to say that LouisLeon Thurstone was to psychology of the first half of the twentiethcentury what Gustav Theodore Fechner was to the last half of thenineteenth. Fechner was the father of quantitative psychology;Thurstone was its chief torchbearer in recent years. In addition tohis many novel contributions during the past thirty years, we haveas memorials to him the Psychometric Society and the journalPsychometrika, both of which were founded through his initiative.Their motto was essentially his own motto: The development ofpsychology as a quantitative, rational science.
Thurstone was born in Chicago, May 19, 1887, to parents of nativeSwedish stock. The family name was Thunstrom but was laterchanged to a form that would better suit the American scene. Hisfather had been an instructor in mathematics in the SwedishArmy, and later became, in turn, a Lutheran minister, a newspapereditor, and a publisher. His mother was interested in music andmusically talented. Leon and his sister, Adele, two years younger,started piano lessons when quite young. Leon showed a transitoryinterest in composing, whereas Adele later completed a collegedegree in music.
Thurstone's elementary education was obtained at several places—Berwyn, Illinois; Centerville, Mississippi; Stockholm, Sweden (bothpublic and private schools); and Jamestown, New York. Highschool was completed in Jamestown. In high school he won acompetition in geometry. With some of the prize money he purchaseda Kodak, which marked the beginning of a lifelong hobbyof photography, in which he later demonstrated considerable artisticskill. As a sophomore, his first notable literary effort was in theform of a letter on "How to Save Niagara," which appeared in theScientific American. This was an attempt to solve the problem ofutilizing large portions of Niagara's waters for power purposeswhile preserving the beauty of the falls, a lively issue at the time.His second publication, also in the Scientific American, demonstratedhis potentialities for creative thinking. As a high schoolsophomore he developed a geometric method for trisecting anangle, a solution that went beyond Euclidian geometry. Later, incollege, he developed an equation for his solution.
Enrolled in electrical engineering at Cornell, Thurstone took aspecial interest in experiments in physics of the singing arc. Theexperiments took the direction of a method by which sounds couldbe recorded on film. He also worked on the designing of a new typeof motion-picture camera and projector that eliminated flicker completelyby means of a continuously running film.
His first notable interests in psychology arose during his studiesof engineering at Cornell. A course on machine design particularlyappealed to him. In this connection he was struck by the factthat among the problems of machine design are some concernedwith the properties of the men who are to operate the machines.Today, of course, there is a growing effort and body of informationalong these lines, sometimes included under the heading of"human engineering." He also became interested in human learningas a scientific problem and attended a few lectures in psychology.Having constructed a working model of his motion-picture projector,on which he secured a patent, he succeeded in having itdemonstrated at the Thomas A. Edison laboratory. Although Edisonwas reported as being impressed, he did not decide to changeover to the production of Thurstone's model. He did, however, offerThurstone a position in his laboratory, which Thurstone acceptedeffective upon his graduation from Cornell. Thurstone's observationsof the way in which Edison went about his work undoubtedlyhad much to do with his interest in the psychology of creativethinking, later in life.
His stay in the Edison laboratory was brief, owing to his desireto return to an academic setting. In the fall of 1912. Thurstonebecame instructor in engineering at the University of Minnesota,where he taught descriptive geometry and drafting. There he tookhis first course in experimental psychology and started his study ofthe learning function (performance as a function of practice time).In the summer of 1914 he started graduate work in psychology atthe University of Chicago. A fellow student, now rather notedin other fields, was Beardsley Ruml.
In the fall of 1915, Thurstone accepted an assistantship in the newand active Department of Psychology at the Carnegie Institute ofTechnology, where the emphasis was on research in applied psychology.Earning his doctorate from Chicago in 1917, he was rapidlypromoted at Carnegie, until he was Professor and head of the departmentin 1920, a position that he held until 1923. His wartimeservice included work in the trade-test division of the Army, wherehis methods of testing and of test appraising were put to use. In1923 and 1924 he devoted a time to research in the Institute of Government,research aimed at the improvement of civil service practices.His efforts in this connection, and his later counsel, have hadlasting effects upon the civil service.
In the summer of 1924 he was married to Thelma Gwinn whomhe had known as a graduate student in psychology. That fall hebecame Associate Professor of Psychology at the University of Chicago.Subsequent academic assignments, briefly listed, includeappointment to full professorship in 1928 and to Charles F. GrayDistinguished Service Professor in 1938. In the meantime, he establishedand directed the Psychometric Laboratory in the SocialSciences Division, in which capacity he continued until his retirementfrom Chicago in 1952. He had much to do with the institutionof the unique examining system at Chicago and with its policiesand practices in the Board of Examinations. During the year 1948-1949 he was visiting professor at the University of Frankfort. Inthe spring semester of 1954 he was visiting professor at the Universityof Stockholm and lectured at other Swedish universities as wellas at the universities of Helsinki and Oslo. Upon his retirementfrom Chicago he became Research Professor and Director of thePsychometric Laboratory at the University of North Carolina,which was his affiliation at the time of his death.
Thurstone's contributions to psychological measurement grew outof his dissatisfactions with psychology as he found it. For example,little or no attention had been given to the description of learningcurves in terms of mathematical equations. In connection with hisdoctoral dissertation Thurstone found that a large number of differentempirical equations might be applied. In 1930 he publishedthe derivation of the first general, rational equation for a learningcurve. He demonstrated that under certain, somewhat standard conditionsthe learning functions is S-shaped. From his general functionhe deduced other relationships, which have been supportedby learning data. For example, he deduced that learning varies asthe 3/2 power of the number of items (beyond the memory span)in a memorized list.
He very early observed that practices of psychological testingwere developing apparently without adequate foundations of theoryto support them. He proceeded to do something about this fromtwo directions, psychological and statistical. A major contributionto psychological theory was in his monograph on intelligence, whichappeared in 1924. He developed and supported the thesis that thedegree of intelligence is related to the degree of incompleteness ofan act at which it becomes focal in consciousness. Becoming highlyaware of an act during its early stages offers much opportunity tobring to bear upon it a wider range of choice or determination.This conception was poorly understood and Thurstone himself didlittle more to investigate intelligence from this particular point ofview, at least explicitly.
From the standpoint of statistically oriented theory, Thurstone'snew conceptions of testing were decidedly more notable. One ofhis chief, early concerns had to do with the development of a rationalmetric for mental ability. He rejected raw-score scales andmental-age scales because of uncertainty concerning equality of unitsand location of a meaningful zero point. He developed a methodcalled "absolute scaling," based upon the concept of item difficulty,on which he published a report in 1925. Using measurements basedupon the absolute-scaling method, he was able to estimate that thezero point of mental ability, as represented in test performance,should be placed at an age several months before birth. Using thesame type of measurements, he concluded that the mental-growthfunction is typically S-shaped, with the inflection point at aboutthe ten-year level.
Of all the statistical bases for tests that Thurstone developed, thatof factor theory and factor analysis is the one for which he will belongest remembered. It is true that factor analysis was not newwith Thurstone; Charles Spearman had initiated factor analysisin psychology as much as a quarter century earlier. Even beforeSpearman, Karl Pearson had proposed the idea of factor analysisas a statistical procedure. Thurstone came to the problem with afresh approach, derived a more generalized theory, and developedprocedures that prevail, in this country at least, today.
Whether it was Thurstone's engineering background or hisnatural habits of thinking, or both, in being faced with a newproblem he usually went to the heart of it, reducing it to its simplestterms. He would, in effect, ask, "What are the variables involved?How are they interrelated?" He also recognized that all too frequentlyin psychology the readily available variables are not thefundamental ones. Thus, latent in any test-score scale, for example,there are probably one or more significant underlying variables ordimensions.
Spearman and his followers had focused most of their attentionupon what they believed to be the single intellective factor g, whichwas believed to be common to all tests involving cognition. Whilerecognizing the existence of group factors (factors of limited generalityamong tests), they played down the importance of suchcomponents. Where only g exists in a group of tests, it alone determinestheir intercorrelations, and the latter exhibit a pattern ofsimple proportionality, except for sampling errors. The group factorswere regarded by Spearman as primarily disturbers of the pictureof simple proportionality and as having little psychological significance.In approaching the factor problem, Thurstone simply asked thequestion, "How many factors are needed to account for the intercorrelationsand how general is each factor?" He regarded a tableof intercorrelations among tests as a matrix and applied to it themathematics of matrix algebra. This called for a number of newconceptions and led to a more general theory in which the Spearman^-factor model is a special case. The number of common factorsis represented by the rank of the matrix. The concept of communalitywas introduced by Thurstone to stand for the sum of theproportions of common-factor variance of a test. This conceptaroused considerable criticism, but it is now generally recognizedthat communality cannot be ignored.
In the practice of factor analysis, Thurstone developed his centroidmethod of extracting factors. A similar summational method hasbeen developed by Cyril Burt in England. But Thurstone differsmost from Burt and his followers in insisting that the centroidmethod gives an arbitrary reference frame that usually makes nopsychological sense. The reference frame must be rotated in order toarrive at positions for the axes that are psychologically meaningful.His chief criterion as to where to rotate lies in his concept of simplestructure. Roughly, simple structure means that with tests representedby vectors extending from the origin of the «-dimensionalreference frame, there are definite regions of higher and lowerdensity of test vectors. The meaningful axes are located at regionsof high density. This procedure results in a factor pattern for agroup of tests, with each test exhibiting relations to a minimal numberof factors and each factor tending to have relations to a minimalnumber of tests. The principle of simple structure thus providescriteria for a unique solution to the rotation problem.Very frequently the optimal achievement of simple-structure solutionscalls for oblique rotations of factor axes. Oblique rotationshave become the rule in the Thurstone procedures of analysis. Interms of theory, this meant to Thurstone that the psychologicalfactors are correlated; they are not statistically independent. Thecorrelation between a pair of factors is estimated from the cosineof the angle of separation between them. The matrix of intercorrelationsof the first-order factors can be factor analyzed, givingrise to second-order factors. These are factors among factors, andThurstone regarded them as having genuine psychological meaning.On this point there is not general agreement among investigators.Noting that Spearman's g factor is not found among the firstorderfactors, Thurstone suggested that it could probably be foundamong the second-order factors. Thurstone's major publications onfactor theory and method were The Vectors of Mind (1935) andMultiple Factor Analysis (1947).
Having developed procedures for factor analysis, Thurstone carriedout a number of factor-analytic studies, often in collaborationwith his wife Thelma. In 1938 he reported his first findings onaptitude factors, which he called "primary mental abilities." Thisstudy involved a battery of 57 tests administered to 240 superioruniversity students. Similar studies were made with children indifferent age groups, even at the kindergarten level, showing essentiallythe same primary mental abilities at all levels. He subsequentlydeveloped and published for general use two different test batteries(for two age groups) for measuring five of the primary mentalabilities. These are commonly known as the Thurstone PMAbatteries. The obvious implication is that a profile of factor scoresshould replace the commonly used single IQ score in describing achild. Such a change in assessing the intelligence of individualshas taken place very slowly.
Thurstone also carried out factor analyses in the areas of temperamentand interest traits, followed by publication of instruments toassess individuals in the factors indicated. With Mrs. Thurstonehe produced for a number of years successive forms of the AmericanCouncil on Education Examination, a very widely used collegeaptitudetest. In his latest years he was very occupied with researchon performance tests of non-aptitude traits of personality.Among Thurstone's most significant contributions were those onpsychophysics and psychological-scaling methods. He was very dissatisfiedwith the classical psychophysics of Fechner, Wundt, andG. E. Miiller because of its restriction to the measurement of limensor thresholds and of points or intervals of subjective equality. Hesaw in the traditional psychophysical methods the possibility fora much broadened use in psychological measurement, includingthe assessment of social, economic, aesthetic, and moral values, asappreciated by human individuals.
Classical psychophysics, being interested ultimately in the functionalrelationships between quantified psychological events andquantified physical events, was confined to those areas of experiencein which measurable physical variables are obvious. Measurementson psychological scales were actually very infrequently made, forlimens and the like were referred back to physical scales. Thurstoneput the emphasis upon psychological scaling. Indeed, where correspondingphysical variables are not readily available this is anatural step.
Starting with human judgments that have the crudest quantitativeproperties, the problem, as Thurstone saw it, was to derivescale values of a relatively high order of measurement by combiningsuch information. As an important step in achieving thisobjective, he developed his well-known law of comparative judgment.The rationale for this law involved the introduction of newconcepts, such as the discriminal process and the discriminal dispersion.A specified stimulus or object at any moment arouses acertain discriminal process in a certain individual. Over a populationof occasions, or over a population of individuals, there is avariability in quantity of process along some specified psychologicalcontinuum. The frequency distribution of those quantities has amean and a standard deviation, two well-known statistical values.Assuming a normal distribution of the discriminal processes, certaindeductions follow concerning the relation between the proportionof the time that one stimulus is judged greater than anotherand their linear separation on a psychological continuum. It is notnecessary to assume independence between discriminal processesfrom pairs of stimuli, over occasions or over individuals. The possiblecorrelation between processes is taken into account in thelaw of comparative judgment.
The law of comparative judgment makes possible a wide rangeof scaling operations, whether the data come from the method ofpair comparisons or from other methods from which comparativejudgments may be inferred such as the method of rank order. Thestimuli being compared may be of almost any kind, such as crimesto be judged for seriousness, nationalities to be judged for desirabilityor handwriting samples to be judged for excellence. Fromthe standpoint of theory, the law of comparative judgment provideda basis of explanation of why it is that Weber's and Fechner'slaws are sometimes not both experimentally verified in the samesituation.
In another area of measurement Thurstone adapted the psychophysicalmethod of equal-appearing intervals to the calibration ofopinions concerning specified issues, institutions, or other socialstimuli. The aim was to arrive at some 20 to 25 statements of opinion,regarding prohibition, for example, that belong at psychologicallyequal steps along a continuum of attitude, from the most extremelyfavorable attitude to the most extremely unfavorable attitude. An"attitude scale" was thus developed. The instrument could then beused to evaluate the characteristic position of a person on this continuumor the average position of a specified population. Thisinvolves the step of asking each person to say which of the opinionshe endorses and noting their scale values. With his studentsThurstone developed attitude scales for such matters as treatmentof criminals, patriotism, war, the Negro, labor unions, communism,birth control, and censorship.
The availability of scaling methods and of attitude scales madepossible some excursions by Thurstone and his students into thestudy of social-psychological problems. For example, there werestudies of the effects of certain motion pictures upon the attitudesof children toward the seriousness of certain crimes, toward theChinese, and toward the Negro. The summational effect of viewingtwo or more films was also detected. Thurstone regarded the fieldof social psychology as being amenable to the isolation of variablesand the study of their quantitative interrelationships, an obviousscientific approach that seemed foreign to too many who had selectedthat field of investigation.
Among Thurstone's last contributions to psychological-measurementtheory was an exploration into the logical problems of predictingfirst choices from knowledge of mean scale positions anddispersions of objects. Some deductions were made that shouldbe of interest in politics and merchandising. He had also developedscaling procedures for judgments that place objects in successivecategories. This method, and others, were used in studies of foodpreferences for the armed services.
Thurstone's breadth of professional interests was shown not onlyby the variety of fields he chose for investigation but also by thevariety of organizations with which he was affiliated and in whichhe showed leadership. He was a Fellow in the American PsychologicalAssociation (President 1932-1933), the American Associationfor the Advancement of Science, the American Statistical Association(Board of Directors), and the American Academy of Arts andSciences. He was Honorary Fellow in the British, Spanish, andSwedish Psychological Associations. He was a member of the MidwesternPsychological Association (President 1930-1931), the Societyfor the Promotion of Engineering Education (Council member),the American Society for Human Genetics (Advisory EditorialCommittee), the American Philosophical Society, the ChicagoPsychology Club (President 1928-1929), and the Chaos Club, Chicago.He was elected to the National Academy of Sciences in 1938.He rendered editorial service to several psychological journals. Hishonorary societies included Phi Delta Kappa, Sigma Xi, and EtaKappa Nu.
His social affiliations included the Acacia fraternity, the QuadrangleClub (Chicago), and the Chicago Literary Club. With hisfamily he frequently spent summers at the family residence atWabigama Club, Elk Lake, Rapid City, Michigan. Special honorsincluded the Award of the American Psychological Association forthe best published paper of 1949 and the Centennial Award, NorthwesternUniversity, 1951.
A very important component of the heritage left by Thurstoneis represented by his former students. He was a stimulating lecturer,always clear and logical. His seminars were regarded as atreat and were attended repeatedly. A number of post-Ph.D. individualswere commonly present. His students generally caught thespirit of the challenge that had dominated Thurstone's career—tomake psychology a quantitative, rational science—and many ofthem have carried on in that same spirit.
The Thurstones had three sons, Robert Leon (born in 1927),Conrad Gwinn (born in 1930), and Frederick Louis (born in 1932),It may be expected that they, too, having chosen the fields ofphysics, medicine, and engineering, will carry on in the scientifictradition set for them by their illustrious father.
n a t i o n a l a c a d e m y o f s c i e n c e s
Any opinions expressed in this memoir are those of the author(s)and do not necessarily reflect the views of theNational Academy of Sciences.
L o u i s l e o n T h u r s t o n e1887—1955
A Biographical Memoir
byJ . P. Guilford
Biographical Memoir
Copyright 1957national academy of sciences washington d.c.
LOUIS LEON THURSTONE1887-1955
BY J. P. GUILFORD
ON SEPTEMBER 19, 1955, one of the world's greatest psychologistspassed from the scene. It is quite appropriate to say that LouisLeon Thurstone was to psychology of the first half of the twentiethcentury what Gustav Theodore Fechner was to the last half of thenineteenth. Fechner was the father of quantitative psychology;Thurstone was its chief torchbearer in recent years. In addition tohis many novel contributions during the past thirty years, we haveas memorials to him the Psychometric Society and the journalPsychometrika, both of which were founded through his initiative.Their motto was essentially his own motto: The development ofpsychology as a quantitative, rational science.
Thurstone was born in Chicago, May 19, 1887, to parents of nativeSwedish stock. The family name was Thunstrom but was laterchanged to a form that would better suit the American scene. Hisfather had been an instructor in mathematics in the SwedishArmy, and later became, in turn, a Lutheran minister, a newspapereditor, and a publisher. His mother was interested in music andmusically talented. Leon and his sister, Adele, two years younger,started piano lessons when quite young. Leon showed a transitoryinterest in composing, whereas Adele later completed a collegedegree in music.
Thurstone's elementary education was obtained at several places—Berwyn, Illinois; Centerville, Mississippi; Stockholm, Sweden (bothpublic and private schools); and Jamestown, New York. Highschool was completed in Jamestown. In high school he won acompetition in geometry. With some of the prize money he purchaseda Kodak, which marked the beginning of a lifelong hobbyof photography, in which he later demonstrated considerable artisticskill. As a sophomore, his first notable literary effort was in theform of a letter on "How to Save Niagara," which appeared in theScientific American. This was an attempt to solve the problem ofutilizing large portions of Niagara's waters for power purposeswhile preserving the beauty of the falls, a lively issue at the time.His second publication, also in the Scientific American, demonstratedhis potentialities for creative thinking. As a high schoolsophomore he developed a geometric method for trisecting anangle, a solution that went beyond Euclidian geometry. Later, incollege, he developed an equation for his solution.
Enrolled in electrical engineering at Cornell, Thurstone took aspecial interest in experiments in physics of the singing arc. Theexperiments took the direction of a method by which sounds couldbe recorded on film. He also worked on the designing of a new typeof motion-picture camera and projector that eliminated flicker completelyby means of a continuously running film.
His first notable interests in psychology arose during his studiesof engineering at Cornell. A course on machine design particularlyappealed to him. In this connection he was struck by the factthat among the problems of machine design are some concernedwith the properties of the men who are to operate the machines.Today, of course, there is a growing effort and body of informationalong these lines, sometimes included under the heading of"human engineering." He also became interested in human learningas a scientific problem and attended a few lectures in psychology.Having constructed a working model of his motion-picture projector,on which he secured a patent, he succeeded in having itdemonstrated at the Thomas A. Edison laboratory. Although Edisonwas reported as being impressed, he did not decide to changeover to the production of Thurstone's model. He did, however, offerThurstone a position in his laboratory, which Thurstone acceptedeffective upon his graduation from Cornell. Thurstone's observationsof the way in which Edison went about his work undoubtedlyhad much to do with his interest in the psychology of creativethinking, later in life.
His stay in the Edison laboratory was brief, owing to his desireto return to an academic setting. In the fall of 1912. Thurstonebecame instructor in engineering at the University of Minnesota,where he taught descriptive geometry and drafting. There he tookhis first course in experimental psychology and started his study ofthe learning function (performance as a function of practice time).In the summer of 1914 he started graduate work in psychology atthe University of Chicago. A fellow student, now rather notedin other fields, was Beardsley Ruml.
In the fall of 1915, Thurstone accepted an assistantship in the newand active Department of Psychology at the Carnegie Institute ofTechnology, where the emphasis was on research in applied psychology.Earning his doctorate from Chicago in 1917, he was rapidlypromoted at Carnegie, until he was Professor and head of the departmentin 1920, a position that he held until 1923. His wartimeservice included work in the trade-test division of the Army, wherehis methods of testing and of test appraising were put to use. In1923 and 1924 he devoted a time to research in the Institute of Government,research aimed at the improvement of civil service practices.His efforts in this connection, and his later counsel, have hadlasting effects upon the civil service.
In the summer of 1924 he was married to Thelma Gwinn whomhe had known as a graduate student in psychology. That fall hebecame Associate Professor of Psychology at the University of Chicago.Subsequent academic assignments, briefly listed, includeappointment to full professorship in 1928 and to Charles F. GrayDistinguished Service Professor in 1938. In the meantime, he establishedand directed the Psychometric Laboratory in the SocialSciences Division, in which capacity he continued until his retirementfrom Chicago in 1952. He had much to do with the institutionof the unique examining system at Chicago and with its policiesand practices in the Board of Examinations. During the year 1948-1949 he was visiting professor at the University of Frankfort. Inthe spring semester of 1954 he was visiting professor at the Universityof Stockholm and lectured at other Swedish universities as wellas at the universities of Helsinki and Oslo. Upon his retirementfrom Chicago he became Research Professor and Director of thePsychometric Laboratory at the University of North Carolina,which was his affiliation at the time of his death.
Thurstone's contributions to psychological measurement grew outof his dissatisfactions with psychology as he found it. For example,little or no attention had been given to the description of learningcurves in terms of mathematical equations. In connection with hisdoctoral dissertation Thurstone found that a large number of differentempirical equations might be applied. In 1930 he publishedthe derivation of the first general, rational equation for a learningcurve. He demonstrated that under certain, somewhat standard conditionsthe learning functions is S-shaped. From his general functionhe deduced other relationships, which have been supportedby learning data. For example, he deduced that learning varies asthe 3/2 power of the number of items (beyond the memory span)in a memorized list.
He very early observed that practices of psychological testingwere developing apparently without adequate foundations of theoryto support them. He proceeded to do something about this fromtwo directions, psychological and statistical. A major contributionto psychological theory was in his monograph on intelligence, whichappeared in 1924. He developed and supported the thesis that thedegree of intelligence is related to the degree of incompleteness ofan act at which it becomes focal in consciousness. Becoming highlyaware of an act during its early stages offers much opportunity tobring to bear upon it a wider range of choice or determination.This conception was poorly understood and Thurstone himself didlittle more to investigate intelligence from this particular point ofview, at least explicitly.
From the standpoint of statistically oriented theory, Thurstone'snew conceptions of testing were decidedly more notable. One ofhis chief, early concerns had to do with the development of a rationalmetric for mental ability. He rejected raw-score scales andmental-age scales because of uncertainty concerning equality of unitsand location of a meaningful zero point. He developed a methodcalled "absolute scaling," based upon the concept of item difficulty,on which he published a report in 1925. Using measurements basedupon the absolute-scaling method, he was able to estimate that thezero point of mental ability, as represented in test performance,should be placed at an age several months before birth. Using thesame type of measurements, he concluded that the mental-growthfunction is typically S-shaped, with the inflection point at aboutthe ten-year level.
Of all the statistical bases for tests that Thurstone developed, thatof factor theory and factor analysis is the one for which he will belongest remembered. It is true that factor analysis was not newwith Thurstone; Charles Spearman had initiated factor analysisin psychology as much as a quarter century earlier. Even beforeSpearman, Karl Pearson had proposed the idea of factor analysisas a statistical procedure. Thurstone came to the problem with afresh approach, derived a more generalized theory, and developedprocedures that prevail, in this country at least, today.
Whether it was Thurstone's engineering background or hisnatural habits of thinking, or both, in being faced with a newproblem he usually went to the heart of it, reducing it to its simplestterms. He would, in effect, ask, "What are the variables involved?How are they interrelated?" He also recognized that all too frequentlyin psychology the readily available variables are not thefundamental ones. Thus, latent in any test-score scale, for example,there are probably one or more significant underlying variables ordimensions.
Spearman and his followers had focused most of their attentionupon what they believed to be the single intellective factor g, whichwas believed to be common to all tests involving cognition. Whilerecognizing the existence of group factors (factors of limited generalityamong tests), they played down the importance of suchcomponents. Where only g exists in a group of tests, it alone determinestheir intercorrelations, and the latter exhibit a pattern ofsimple proportionality, except for sampling errors. The group factorswere regarded by Spearman as primarily disturbers of the pictureof simple proportionality and as having little psychological significance.In approaching the factor problem, Thurstone simply asked thequestion, "How many factors are needed to account for the intercorrelationsand how general is each factor?" He regarded a tableof intercorrelations among tests as a matrix and applied to it themathematics of matrix algebra. This called for a number of newconceptions and led to a more general theory in which the Spearman^-factor model is a special case. The number of common factorsis represented by the rank of the matrix. The concept of communalitywas introduced by Thurstone to stand for the sum of theproportions of common-factor variance of a test. This conceptaroused considerable criticism, but it is now generally recognizedthat communality cannot be ignored.
In the practice of factor analysis, Thurstone developed his centroidmethod of extracting factors. A similar summational method hasbeen developed by Cyril Burt in England. But Thurstone differsmost from Burt and his followers in insisting that the centroidmethod gives an arbitrary reference frame that usually makes nopsychological sense. The reference frame must be rotated in order toarrive at positions for the axes that are psychologically meaningful.His chief criterion as to where to rotate lies in his concept of simplestructure. Roughly, simple structure means that with tests representedby vectors extending from the origin of the «-dimensionalreference frame, there are definite regions of higher and lowerdensity of test vectors. The meaningful axes are located at regionsof high density. This procedure results in a factor pattern for agroup of tests, with each test exhibiting relations to a minimal numberof factors and each factor tending to have relations to a minimalnumber of tests. The principle of simple structure thus providescriteria for a unique solution to the rotation problem.Very frequently the optimal achievement of simple-structure solutionscalls for oblique rotations of factor axes. Oblique rotationshave become the rule in the Thurstone procedures of analysis. Interms of theory, this meant to Thurstone that the psychologicalfactors are correlated; they are not statistically independent. Thecorrelation between a pair of factors is estimated from the cosineof the angle of separation between them. The matrix of intercorrelationsof the first-order factors can be factor analyzed, givingrise to second-order factors. These are factors among factors, andThurstone regarded them as having genuine psychological meaning.On this point there is not general agreement among investigators.Noting that Spearman's g factor is not found among the firstorderfactors, Thurstone suggested that it could probably be foundamong the second-order factors. Thurstone's major publications onfactor theory and method were The Vectors of Mind (1935) andMultiple Factor Analysis (1947).
Having developed procedures for factor analysis, Thurstone carriedout a number of factor-analytic studies, often in collaborationwith his wife Thelma. In 1938 he reported his first findings onaptitude factors, which he called "primary mental abilities." Thisstudy involved a battery of 57 tests administered to 240 superioruniversity students. Similar studies were made with children indifferent age groups, even at the kindergarten level, showing essentiallythe same primary mental abilities at all levels. He subsequentlydeveloped and published for general use two different test batteries(for two age groups) for measuring five of the primary mentalabilities. These are commonly known as the Thurstone PMAbatteries. The obvious implication is that a profile of factor scoresshould replace the commonly used single IQ score in describing achild. Such a change in assessing the intelligence of individualshas taken place very slowly.
Thurstone also carried out factor analyses in the areas of temperamentand interest traits, followed by publication of instruments toassess individuals in the factors indicated. With Mrs. Thurstonehe produced for a number of years successive forms of the AmericanCouncil on Education Examination, a very widely used collegeaptitudetest. In his latest years he was very occupied with researchon performance tests of non-aptitude traits of personality.Among Thurstone's most significant contributions were those onpsychophysics and psychological-scaling methods. He was very dissatisfiedwith the classical psychophysics of Fechner, Wundt, andG. E. Miiller because of its restriction to the measurement of limensor thresholds and of points or intervals of subjective equality. Hesaw in the traditional psychophysical methods the possibility fora much broadened use in psychological measurement, includingthe assessment of social, economic, aesthetic, and moral values, asappreciated by human individuals.
Classical psychophysics, being interested ultimately in the functionalrelationships between quantified psychological events andquantified physical events, was confined to those areas of experiencein which measurable physical variables are obvious. Measurementson psychological scales were actually very infrequently made, forlimens and the like were referred back to physical scales. Thurstoneput the emphasis upon psychological scaling. Indeed, where correspondingphysical variables are not readily available this is anatural step.
Starting with human judgments that have the crudest quantitativeproperties, the problem, as Thurstone saw it, was to derivescale values of a relatively high order of measurement by combiningsuch information. As an important step in achieving thisobjective, he developed his well-known law of comparative judgment.The rationale for this law involved the introduction of newconcepts, such as the discriminal process and the discriminal dispersion.A specified stimulus or object at any moment arouses acertain discriminal process in a certain individual. Over a populationof occasions, or over a population of individuals, there is avariability in quantity of process along some specified psychologicalcontinuum. The frequency distribution of those quantities has amean and a standard deviation, two well-known statistical values.Assuming a normal distribution of the discriminal processes, certaindeductions follow concerning the relation between the proportionof the time that one stimulus is judged greater than anotherand their linear separation on a psychological continuum. It is notnecessary to assume independence between discriminal processesfrom pairs of stimuli, over occasions or over individuals. The possiblecorrelation between processes is taken into account in thelaw of comparative judgment.
The law of comparative judgment makes possible a wide rangeof scaling operations, whether the data come from the method ofpair comparisons or from other methods from which comparativejudgments may be inferred such as the method of rank order. Thestimuli being compared may be of almost any kind, such as crimesto be judged for seriousness, nationalities to be judged for desirabilityor handwriting samples to be judged for excellence. Fromthe standpoint of theory, the law of comparative judgment provideda basis of explanation of why it is that Weber's and Fechner'slaws are sometimes not both experimentally verified in the samesituation.
In another area of measurement Thurstone adapted the psychophysicalmethod of equal-appearing intervals to the calibration ofopinions concerning specified issues, institutions, or other socialstimuli. The aim was to arrive at some 20 to 25 statements of opinion,regarding prohibition, for example, that belong at psychologicallyequal steps along a continuum of attitude, from the most extremelyfavorable attitude to the most extremely unfavorable attitude. An"attitude scale" was thus developed. The instrument could then beused to evaluate the characteristic position of a person on this continuumor the average position of a specified population. Thisinvolves the step of asking each person to say which of the opinionshe endorses and noting their scale values. With his studentsThurstone developed attitude scales for such matters as treatmentof criminals, patriotism, war, the Negro, labor unions, communism,birth control, and censorship.
The availability of scaling methods and of attitude scales madepossible some excursions by Thurstone and his students into thestudy of social-psychological problems. For example, there werestudies of the effects of certain motion pictures upon the attitudesof children toward the seriousness of certain crimes, toward theChinese, and toward the Negro. The summational effect of viewingtwo or more films was also detected. Thurstone regarded the fieldof social psychology as being amenable to the isolation of variablesand the study of their quantitative interrelationships, an obviousscientific approach that seemed foreign to too many who had selectedthat field of investigation.
Among Thurstone's last contributions to psychological-measurementtheory was an exploration into the logical problems of predictingfirst choices from knowledge of mean scale positions anddispersions of objects. Some deductions were made that shouldbe of interest in politics and merchandising. He had also developedscaling procedures for judgments that place objects in successivecategories. This method, and others, were used in studies of foodpreferences for the armed services.
Thurstone's breadth of professional interests was shown not onlyby the variety of fields he chose for investigation but also by thevariety of organizations with which he was affiliated and in whichhe showed leadership. He was a Fellow in the American PsychologicalAssociation (President 1932-1933), the American Associationfor the Advancement of Science, the American Statistical Association(Board of Directors), and the American Academy of Arts andSciences. He was Honorary Fellow in the British, Spanish, andSwedish Psychological Associations. He was a member of the MidwesternPsychological Association (President 1930-1931), the Societyfor the Promotion of Engineering Education (Council member),the American Society for Human Genetics (Advisory EditorialCommittee), the American Philosophical Society, the ChicagoPsychology Club (President 1928-1929), and the Chaos Club, Chicago.He was elected to the National Academy of Sciences in 1938.He rendered editorial service to several psychological journals. Hishonorary societies included Phi Delta Kappa, Sigma Xi, and EtaKappa Nu.
His social affiliations included the Acacia fraternity, the QuadrangleClub (Chicago), and the Chicago Literary Club. With hisfamily he frequently spent summers at the family residence atWabigama Club, Elk Lake, Rapid City, Michigan. Special honorsincluded the Award of the American Psychological Association forthe best published paper of 1949 and the Centennial Award, NorthwesternUniversity, 1951.
A very important component of the heritage left by Thurstoneis represented by his former students. He was a stimulating lecturer,always clear and logical. His seminars were regarded as atreat and were attended repeatedly. A number of post-Ph.D. individualswere commonly present. His students generally caught thespirit of the challenge that had dominated Thurstone's career—tomake psychology a quantitative, rational science—and many ofthem have carried on in that same spirit.
The Thurstones had three sons, Robert Leon (born in 1927),Conrad Gwinn (born in 1930), and Frederick Louis (born in 1932),It may be expected that they, too, having chosen the fields ofphysics, medicine, and engineering, will carry on in the scientifictradition set for them by their illustrious father.
www.psychspace.com心理学空间网
