2 '%^^ ^^^ xO^^. .x^' <^. c> ^^" *= /T^v^ ^\ ^o^ " r xOc> A^ lEjmxmtxmnl Paiirtiologti jTOfluograyl^g No. 18 THE STANFORD REVISION AND EXTENSION OF THE BINET-SIMON SCALE FOR MEASURING INTELLIGENCE Bt Lewis M. Terman, Grace Lyman, George Ordahl, Louise Ellison Ordahl, Neva Galbreath AND WiLFORD TaLBERT Absistbd bt Herbert E. Knollin, J. H. Williams, H. G. Childs, Helen Trost, Richard Zeidler, Charles Waddle AND Irene Cuneo BALTIMORE WARWICK & YORK, Inc. 1917 LZB//3/ ■Ts- Copyright. 1917, By Warwick & York, Inc. M -4 1918 ©CI,A481957 .1 /VvO Dedicated by Lewis M. Terman to Those whose loyal cooperation made the study possible EDITOR'S PREFACE The labors of Professor Terman and his co-workers at Stanford University in the critical examination and improvement of the Binet-Simon Scale for Measuring Intelligence are so well and so favorably known by psychologists and by the many users of the method that no words of editorial introduction are needed to call attention to the importance of the present mono- graph. The results of these labors are embodied in the Stan- ford Revision of the Binet-Simon Scale. A general guide for the application of this Revision has been' published elsewhere. In the present monograph, how- ever, the reader is taken '' behind the scenes/' is shown the precise methods by which the Revision was made, the actual data on which it was based. There is intro- duced also an instructive discussion of a number of very salient questions: What is the nature of intelli- gence? How is intelligence distributed? What sex differences exist in intelligence? What is the relation between intelligence and social status? Between intelligence and school success? Is the intelligence quotient a vaUd measure? How shall the validity of any single test in an intelligence scale be determined? What principles should govern the assembling of tests into a system, or scale? These questions have more than a merely technical interest: they bear in many ways upon practical problems of school instruc- tion and administration. The monograph should do much to stimulate and to clarify thinking, both in psychological and in pedagogical circles. G. M. W. PREFACE The present monograph summarizes the data on which the Stanford revision and extension of the Binet scale rests and gives an analysis of the results secured by the application of the revised scale with nearly 1000 unselected school children. The complete guide for giving and scoring the tests and for the interpretation of results is pubUshed separately: The Measurement of Intelligence (Houghton Mifflin Co., 1916). This and the present monograph are in a sense companion volumes, and it is especially hoped that all who use the guide will also make them- selves familiar with source material herein offered. The responsibility of each of the various collabora- tors is related in Chapter I. Terman is responsible for the assembling of the source material, the arrange- ment of the trial series, the scoring of all the records, the elaboration of the revision from the results, the formulation of the procedure, the analysis of the data, and the preparation of this monograph for the press. In all these matters, however, invaluable help was rendered by all who collaborated in the work. What- ever merit the present revision possesses must be credited in no small degree to the loyal and painstaking work of those who assisted in the tests. Hearty thanks are also due the public-school officers, teachers, prin- cipals and superintendents for their always willing cooperation in furnishing pupils for the tests and in supplying the supplementary information called for. Lewis M. Tekman. Stanford University, June 12, 1916. 3 TABLE OF CONTENTS Page Chapter I. — Brief Account of the Revision and Its History 7 Chapter II. — The Distribution of Intelligence 26 Chapter III. — The Rate of Growth and the VaHdity of the I Q. . . 51 Chapter IV. — Sex Differences 62 Chapter V. — The Relation of IntelKgence to Social Status 84 Chapter VI. — The Relation of InteUigence to School Success 104 Chapter VII.— The Validity of the Individual Tests 129 Chapter VIII. — Some Considerations Relating to the Formation of an InteUigence Scale 146 Appendix (1). — Statistics on the individual tests 163 Appendix (2). — Form used for supplementary information 179 CHAPTER I BRIEF ACCOUNT OF THE STANFORD REVI- SION AND ITS HISTORY Terman and Childs' tests of 396 children in 1910- 1911 afforded data for a tentative revision and exten- sion of the Binet 1908 scale. The most important changes introduced into the test series by that revision involved a shifting downward of most of the tests in the lower end of the scale, a shifting of several of the upper tests in the opposite direction, and the addition of the following new tests: the ^^ball and field" test, a graded completion test, and a graded test of vocabu- lary and fable interpretation. ^ In 1911-1912 the tentative revision was applied to 310 public school children in Palo Alto, San Jose and Los Angeles. Of these, 127 were tested by Miss Helen Trost, a senior student at Stanford University, 52 by Dr. Charles Waddle, of the State Normal School, Los Angeles, and the remainder by Terman. The children were selected from each grade by arbitrary rule, accord- ing to seating when this had not been determined by scholarship, otherwise alphabetically. The schools selected were attended chiefly by children of the middle classes. The number at each age is given on p. 9. For several reasons the results of this study did not afford satisfactory data for a further revision of the scale. The method of selecting subjects failed to give representative children at all the various ages, and too ^Lems M. Terman and H. G. Childs: A Tentative Revision and Extension of the Binet-Simon Measuring Scale of Intelligence. /. oj Educ. Psych., Vol. 3, Feb., Mar., Apr. and May, 1912. 7 8 STANFORD REVISION OF BINET-SIMON SCALE little attention had been given to securing uniformity of procedure. Moreover, some of the features of the Terman and Childs revision proved impracticable in actual use and showed the necessity of a more thorough- going revision based on more extensive data. Ac- cordingly, a new investigation was undertaken, much more extensive than the earlier ones and more care- fully planned. The work was begun in the autumn of 1913 by Ter- man, Lyman and Galbreath. Miss Lyman and Miss Galbreath were at the time graduate students in 1];. Department of Education at Stanford Universii\\ Later, Professor and Mrs. Ordahl, of the University of Nevada, and Mr. Wilford Talbert, a former gradu- ate student of Stanford University, kindly offered to cooperate in the work. During the school year '>! 1913-1914 approximately 1000 public-school child: e. were tested by Miss Lyman, Miss Galbreath, I^tr. Talbert, Professor Ordahl and Mrs. Ordahl. The following year another Stanford graduate. Miss Irene Cuneo, secured further data on the lower tests by applj ing the revised scale to the 54 kindergarten child}> '^ attending the training department of the California State Normal School at San Jose. The accompanying table shows the number of children of each age tested by each examiner. The data secured from these 982 children made possible the revision of the scale up to the 14-year level, but left the extreme upper part of the scale still insecure. Revision of this part was fortunately made possible by the following data: (1) Tests of 40 high-school students in San Jose and Campbell, Cal. Thirty-two of these, tested by Tea- man, were members of classes in ^Uife-career stud- " I THE STANFORD REVISION AND ITS HISTORY 9 TABLE I Number of Children at Each Age Tested by the Several Examiners Examiner Place 4 5 6 7 8 9 10 11 12 13 14 15 16-17 Total Lyman Prof, and Mrs. Ordahl Galbreath Talbert Cuneo Santa Barbara Los Angeles Los Gatos, Cal. Reno, Nevada San Jose and Mt. View, Cal. Oakland Kindergarten St. Nor. School San Jose 3 14 17 10 11 15 18 54 18 27 18 32 22 117 23 31 23 15 92 34 31 23 12 100 35 38 28 12 113 27 29 20 11 33 22 10 14 26 22 25 10 42 33 11 12 29 33 10 10 15 19 5 7 4 8 1 1 299 304 176 151 54 Total at each age 87 79 83 98 82 46 14 982 and were of junior or senior grade. Their ages ranged from 17 to 20, with a median of 18+. The others, tested by Mr. Zeidler, were first and second-year students from 14 to 16 years of age. The ^^high school group," referred to in the statistics, included only the 32 tested by Terman. (2) Tests of 30 business men in Palo Alto and vicin- ity? by Knolhn and Zeidler. The men selected had had Uttle or no formal education beyond the common school but had shown themselves ordinarily successful in the various lines of business represented in a smaU city. (3) Tests of 150 "migrating unemployed" men by Mr. Knollin. These were temporary residents at a "hobo hotel" conducted at Palo Alto for transient pedestrians who were willing to work a few hours for a night's lodging and a couple of meals. The 10 STANFORD REVISION OF BINET-SIMON SCALE ages ranged from 18 to 65, but were chiefly between 25 and 40. Tests of somewhat more than 100 unem- ployed men were also made with our trial series by Mr. Glenn Johnson, of Reed College, Portland, Ore- gon, who kindly loaned us his data for comparative purposes. (4) Tests of 150 juvenile delinquents in the Whittier (Cal.) State School. These tests were made by Dr. J. H. Williams, at that time Fellow on the Buckel Foundation, Stanford University. The ages of the delinquents ranged from 10 to 21, but most were between 14 and 19.^ Returning now to the tests of 1000 unselected chil- dren, this part of the investigation may be described as follows: 1. We first assembled as nearly as possible all the results which had been secured for each test of the Binet scale by all the workers of all countries, includ- ing per cents passing the test at various ages, conditions under which the results were secured, method of pro- cedure, etc. After a comparative study of these data, and in the light of results we had ourselves secured, a provisional arrangement of the tests was prepared for trial. 2 As the foregoing studies of delinquents, unemployed, and business men are to be published separately by their several authors, it is not necessary to enter into them here in detail. Mr. Williams has since increased his tests of delinquent boys to nearly 500, and Miss Cuneo her tests of kindergarten children to approximately 100. More re- cently another Stanford University student has used the revision with 150 employees, mostly unskilled or semi-skilled. The mental ages found are given in Chapter II. About a dozen additional studies, involving tests of nearly 1000 school children and adults, were carried out at Stanford University during the school year of 1916-1917. These studies, which will be reported in a forthcoming monograph, have sought especially to deter- mine the vahdity of the Stanford Revision as a means of diagnosing a child's educability. THE STANFORD REVISION AND ITS HISTORY 11 2. A plan was then devised for securing subjects who should be as nearly as possible representative of the several ages. The method was to select a school in a community of average social status, a school at- tended by all or practically all the children in the dis- trict where it was located. In order to get clear pic- tures of age differences, the tests were confined to children who were within two months of a birthday. ^ To avoid accidental selection, all the children within two months of a birthday were tested, in whatever grade enrolled (below the high school). Tests of foreign-born children, however, were eliminated in the treatment of results. 3. The children's responses were for the most part recorded verbatim. This made it possible to re-score the records according to any desired standard and thus to fit a test more perfectly to the age level assigned it. 4. The tests were made at an average rate of about fifty minutes per test. The time was rarely below 40 minutes, except with the children of four and five years. The older children and adults more often required from fifty minutes to an hour. In spite of the rather long time required for the test we are con- vinced that fatigue has been a negligible factor in our results. The tasks required of the child are so novel that the reserve energies are brought into play and attention is kept at high efficiency much longer than would be the case with ordinary school work. 5. As may be inferred from the time required, the testing was reasonably thorough. It is possible, how- ever, that occasionally a success has been missed by not carrying the test high enough, or a failure missed ' The only exception to this was in the case of 14 five year olds, tested by Miss Cuneo. 12 STANFORD REVISION OF BINET-SIMON SCALE by not going back far enough. Errors of this sort doubtless about balance in the long run, and so do not affect appreciably the distribution of mental ages. They do affect, however, the statistical treatment of the results for individual tests, and as a rule we have given the per cents passing a test only for those ages at which all the children were given the test. 6. Much attention was given to securing uniformity of procedure. A half-year was devoted to training the examiners and another half-year to the supervi- sion of the testing.^ In the further interests of uni- formity all the records were scored by one person (Terman) . In working out a revision of the scale the guiding principle was to secure an arrangement of the tests and a standard of scoring which would cause the median mental age of the children of each age-group to coincide with the median chronological age. If the median mental age at any point in the scale was too high or too low, it was only necessary to change the location of certain of the tests, or to change the standard of scoring, until an order of arrangement and a standard of passing were found which would throw the median mental age where it belonged. We had already be- come convinced that no satisfactory revision of the Binet scale was possible on any theoretical consider- ations as to the percent of passes which an individual test ought to show in a given year in order to be con- sidered standard for that year, although such a plan might be feasible with a scale differently founded. ^ This statement does not apply, however, to Professor and Mrs. Ordahl, who had to rely on a 20-page guide, supplemented by a few demonstration tests and such further direction as could be given by correspondence. Mr. Talbert and Miss Cuneo had also somewhat less specific training for the tests than had the others, though both had taken a half-year course in clinical psychology. THE STANFORD REVISION AND ITS HISTORY 13 As was to be expected, the first draft of the revision did not prove satisfactory. The scale was still too hard at some points and too easy at others. Three successive revisions were necessary, involving three separate scorings of the data and as many tabulations of the mental ages, before the desired degree of accur- acy was secured. As finally left, the scale gives a median intelligence quotient closely approximating 100 for our non-se- lected children of each age. The revision contains six regular tests and from one to three alternative tests in each year from 3 to 10, eight tests at year 12, six at 14, and six in each of two higher groups which are named, in order, ^^ average adult'' and '^superior adult." The tests in the two highest groups were standardized chiefly on the basis of results from 400 adults. The extension of the scale in the upper range is such that ordinarily intelUgent adults, little educated, test near to what is called the '^average adult" level. Adults whose intelligence is known from other sources to be superior are found to test well up to the '^superior adult" level, whether they are well educated or prac- tically unschooled. Of 30 uneducated business men, 15 tested at '^average adult" (15-17), 8 at ^^ superior adult" (17-19), 6 at ^'inferior adult" (14-15) and 1 at 13. Of 32 high-school students who were 16 years of age or older, 22 tested at ^'average adult," 5 at "superior adult," 5 at "inferior adult." The trial arrangement of tests included, in addition to those of the Binet 1908 and 1911 series, 31 addi- tional tests, as follows: Kuhlmann's test of discrimin- ation of forms, two new tests of comprehension (Terman), four tests of repeating digits in reversed order (suggested by Bobertag), repeating 8 digits, test of ability to tie a bow-knot (Terman), two tests of finding similarities 14 STANFORD KEVISION OF BINET-SIMON SCALE (Terman), six vocabulary tests (Terman and Childs), two form-board tests (Healy and Fernald), the Healy- Fernald code test, two tests of fable interpretation (Ter- man and Childs), two ^^ ball and field " tests (Terman and Childs), an ^' induction'^ test (Terman), a test of arith- metical reasoning (selected from Bonser's series), an ''ingenuity'' test (Terman), a test of ''comprehension of physical relations" (suggested in part by Meumann), a test of observation (drawing an apple with pencil through it, suggested by an experiment of Professor Earl Barnes), and the "problem of enclosed boxes'' (Terman). The test of observation proved too un- satisfactory to be included in the revision, as was true also of one of the Healy-Fernald form boards and Binet's "suggestion" and "reversed triangle" tests. Counting both regular and alternative tests, the re- vision contains 90 tests, as contrasted with 54 in the Binet 1911 series. As far as possible, the original Binet tests have been retained in the form in which they were used by their author, although in a number of cases it has seemed advisable to introduce alterations either in procedure or scoring. In preparing the directions, special at- tention has been devoted to the difficulties encountered by inexperienced examiners in giving and scoring the tests. 5 While it is not claimed that the revision here offered is satisfactory in every respect, the authors believe that it possesses a number of distinct advantages over other versions of the Binet scale. Among these ad- vantages are the following: ^ An extended guide for the giving and scoring of the individual tests and for the interpretation of test results has been pubhshed in a separate volume, The Measurement of Intelligence (Houghton Mifflin Co., 1916). With the latter is furnished all the necessary material for the use of the Stanford Revision. THE STANFORD REVISION AND ITS HISTORY 15 1. Correction of the too-great ease of the original scale at its lower end and its too-great difficulty at the upper end. This correction should have the im- portant result of tending to prevent the overlooking of borderline cases of deficiency among young children and the overestimation of deficiency among adults of somewhat inferior or borderline intelligence.® 2. The revision not only contains a much larger number of tests than any other series, but also brings into operation a much greater variety of mental func- tions. This is especially true for the upper part of the scale. 3. It is believed that the detailed directions set forth in the companion volume for giving and scoring the tests should tend materially to promote uniformity of procedure. The following copy of the forms used in applying the tests will serve to indicate their nature. YEAR III. (6 tests, 2 months each.) 1. Points to parts of body. (3 of 4.) Nose Eyes Mouth Hair 2. Names familiar objects. (3 of 5.) Key Penny Closed knife Watch Pencil 3. Pictures, enumeration or better. (At least 3 objects in one picture. ''TeU me everything you can see in this picture.") a. Dutch Home b. Canoe c. Post Office 4. Gives sex. (Note form of question.) 5. Gives last name 6. Repeats 6-7 syUables. (1 of 3.) a. "I have a httle dog." h. "The dog runs after the cat." c. "In summer the sun is hot." Al. Repeats 3 digits. (1 of 3. Order correct. Read 1 per second.) 6-4-1 3-5-2 8-3-7 YEAR IV. (6 tests, 2 months each.) 1. Compares hues. (3 of 3, or 5 of 6.) 1 2 3 2. Discrimination of forms. (Kuhlmann. 7 of 10.) Circle Square Triangle Other errors ^ For a fuller discussion, see L. M. Terman, Some problems related to the detection of border-line cases of mental deficiency, J. of Psycho- Asthenics, Sept. and Dec, 1915. 16 STANFORD REVISION OF BINET-SIMON SCALE 3. Counts 4 pennies. (No error.) 4. Copies square. (Pencil. 1 of 3.) 1 2 3 5. Comprehension, 1st degree. (2 of 3.) "What must you do: a. ** When you are sleepy? b. "When j^ou are cold? c. " WTien you are hungry? " 6. Repeats 4 digits. (1 of 3. Order correct. Read 1 per second.) 4-7-3-9 2-8-5-4 7-2-6-1 Al. Repeats 12-13 syllables. (1 of 3 absolutely correct, or 2 with 1 error each.) a. "The boj^'s name is John. He is a very good boy." h. "When the train passes you will hear the whistle blow." c. "We are going to have a good time in the country." YEAR V. (6 tests, 2 months each.) 1. Comparison of weights. (2 of 3.) 3-15 15-3 3-15 2. Colors. (No error.) Red Yellow Blue Green 3. Aesthetic comparison. (No error.) Upper pair Middle Lower 4. Definitions, use or better. (4 of 6.) Chair DoU Horse Pencil Fork Table 5. Patience, or divided rectangle. (2 of 3 trials. 1 minute each.) 1 Time 2 Time 3 Time 6. Three commissions. (No error. Order correct.) Puts key on chair Brings box Shuts door Al. Age YEAR VL (6 tests, 2 months each.) 1. Right and left. (3 of 3, or 5 of 6.) R. hand L. ear R. eye 2. Mutilated pictures. (3 of 4.) Eye Mouth Nose Arms 3. Coimts 13 pennies. (1 of 2 trials, without error.) 4. Comprehension, 2d degree. (2 of 3.) "WTiat's the thing to do: a. "If it is raining when you start to school? b. "If you find that your house is on fire? c. "If you are going some place and miss your car?" 5. Coins. (3 of 4. Present in order given below.) Nickel Penny Quarter Dime 6. Repeats 16-18 syllables. (1 of 3 absolutely correct, or 2 with 1 error each.) a. "We are having a fine time. We foimd a little mouse in the trap." 6. "Walter had a fine time on his vacation. He went fishing every day." c. "We will go out for a long walk. Please give me my pretty straw hat." Al. Morning or afternoon. (Note form of question.) THE STANFORD REVISION AND ITS HISTORY 17 YEAR VII. (6 tests, 2 months each.) 1. Fingers. (No error.) R L Both . . 2 P ictures, description or better. (Over half of performance descrip- tion. "Tell me what this picture is about?" "What is this a picture of?") a. Dutch Home b. Canoe c. Post Office w ' \ jT * * 3 Repeats 5 digits. (1 of 3. Order correct. Read 1 per second.) 3il_7_5_9 4-2-8-3-5 9-8-1-7-6 4. Ties bow knot. (Model shown. 1 mmute. "Single" bow half credit.) Time Method 5. Gives differences. (2 of 3.) a. Fly and butterfly h. Stone and egg c. Wood and glass 6. Copies diamond. (Pen. 2 of 3.) a b .c Al. 1. Names days of week. (Order correct. 2 of 3 checks correct.) Mon., Tues., Wed., Thurs., Fri., Sat., Sun. Al. 2. Repeats 3 digits backwards. (1 of 3. Read 1 per second.) 2-8-3. 4-2-7 9-5-8 YEAR VIII. (6 tests, 2 months each.) 1. Ball and field. (Inferior plan or better.) 2 Counts 20-0. (40 seconds. 1 error allowed.) Time Errors 3. Comprehension, 3rd degree. (2 of 3.) "What's the thing for you to do: , . , , , a. ''When you have broken something which belongs to someone else? b. "When you are on your way to school and notice that you are in danger of being tardy? • • • • • c. " If a pla5niiate hits you without meaning to do it? 4. Gives similarities, two things. (2 of 4. "In what way are wood and coal alike?" etc. Any real likeness is plus.) o. Wood and coal b. Apple and peach c. Iron and silver d. Ship and automobile 5. Definitions superior to use. (2 of 4. "Thing" as genus counts plus.) a. Balloon b. Tiger c. Football d. Soldier ; * ' ', 6. Vocabulary, 20 words. Score Total Vocab Al. 1. Six coins. (No error. Give in order indicated.) .05 01 25 10 1.00.... .50 Al. 2. Dictation. ("See the little boy." Easily legible. Pen, 1 minute.) Time Score by Ayres scale • 18 STANFORD REVISION OF BINET-SIMON SCALE YEAR IX. (6 tests, 2 months each.) 1. Date. (Allow error of 3 days in c, no error in a, h, or d.) a. Day of week. ... 6. month. . . . c. day of m d. year 2. Weights. (3, 6, 9, 12, 15. Procedure not iUustrated. 2 of 3 correct.) a Method b Method c Method 3. Makes change. (2 of 3. No coins, paper, or pencil.) 10-4 15-12 25-4 4. Repeats 4 digits backwards. (1 of 3. Read 1 per second.) 6-5-2-8 4-9-3-7 8-6-2-9 5. Three words. (2 of 3. Oral. 1 sentence or not over 2 coordinate clauses.) a. Boy, river, ball b. Work, money, men c. Desert, rivers, lakes 6. Rhymes. (3 rhymes for each word. 1 minute for each part. Illustrate with hat, rat, cat.) a. Day Time b. Mill Time c. Spring Time Al. 1. Months. (15 seconds and 1 error in naming. 2 checks of 3 correct.) Jan., Feb., Mch., Apr., May, June, July, Aug., Sept., Oct., Nov., Dec. Al. 2. Stamps, gives total value. (2d trial if individual values are known.) YEAR X. (6 tests, 2 months each.) 1. Vocabulary, 30 words. Score Total Vocab 2. Absurdities. (4 of 5. Warn. Spontaneous correction allowed.) a. "A man said: 'I know a road from my house to the city which is down hill all the way to the city and down hill aU the way back'^home. ' " b. "An engineer said that the more cars he had on his train the faster he could go." c. "Yesterday the poUce found the body of a girl cut into 18 pieces. They beheve that she killed herself." d. "There was a railroad accident yesterday, but it was not very serious. Only 48 people were killed." e. "A bicycle rider, being thrown from his bicycle in an accident, struck his head against a stone and was instantly killed. They picked him up and carried him to the hospital, and they do not think he wiU get well again." 3. Designs. (1 correct, 1 half correct. Expose 10 seconds.) a. . .b. . . 4. Reading and report, (8 memories. 35 seconds and 2 mistakes in reading.) Memories Time for reading Mistakes New York. ] September 5th. | — A fire | last night | burned ] three houses I near the center | of the city. | It took some time | to put it out. | THE STANFORD REVISION AND ITS HISTORY 19 The loss I was fifty thousand dollars, | and seventeen famihes | lost their homes. | In saving | a girl \ who was asleep | in bed, | a fireman | was burned | on the hands. 5. Comprehension, 4th degree. (2 of 3. Question may be repeated.) a. "What ought you to say when someone asks your opinion about a person you don't know very well? " h. "What ought you to do before undertaking (beginning) some- thing very important? " c. "Why should we judge a person more by his actions than by his words? " 6. 60 words. (Score half-minutes separately. Illustrate with clouds, dog, chair, happy.) 1 2 3 4 5 6 Method Al. 1. Repeats 6 digits. (I of 2. Order correct. Read 1 per sec- ond.) 3-7-4-8-5-9 5-2-1-7-4-6 Al. 2. Repeats 20-22 syllables. (1 of 3 correct, or 2 with 1 error each.) a. "The apple tree makes a cool pleasant shade on the ground where the children are playing." h. "It is nearly half-past one o'clock; the house is very quiet and the cat has gone to sleep," c. "In summer the days are very warm and fine; in winter it snows and I am cold." Al. 3. Form board. (Healy-Fernald Puzzle A. 3 times in 5 min- utes.) Time: a b c Method YEAR XII. (8 tests, 3 months each.) 1. Vocabulary, 40 words. Score Total Vocab 2. Abstract words. (3 of 5.) a. Pity b. Revenge c. Charity d. Envy e. Justice 3. Ball and field. (Superior plan.) 4. Dissected sentences. (2 of 3. 1 minute each.) a Time b Time c Time 6. Fables. (Score 4, i. e., two correct or the equivalent in half credits.) a. Hercules and wagoner b. Maid and eggs c. Fox and crow d. Farmer and stork e. MiUer, son and donkey 6. Repeats 5 digits backwards. (1 of 3. Read 1 per second.) 3-1-8-7-9 6-9-4-8-2 5-2-9-6-1 7. Pictures, iQterpretation. (3 of 4. "Explain this picture.") a. Dutch Home b. Canoe 20 STANFORD REVISION OF BINET-SIMON SCALE c. Post Office d. Colonial Home 8. Gives similarities, three things. (3 of 5. ''In what way are — , — , — , alike?" Grade fairly closely.) a. Snake, cow, sparrow b. Book, teacher, newspaper c. Wool, cotton, leather d. Knife-blade, penny, piece of wire e. Rose, potato, tree YEAH XIV. (6 tests, 4 months each.) 1. Vocabulary, 50 words. Score Total Vocab 2. Induction test. (Gets rule by 6th folding. Unfold after each cutting.) 1 2 3 4 5 6 3. President and king. (Power. . . . accession. . . , tenure. ... 2 of 3.) a b c 4. Problems of fact. . (2 of 3. Query on a and b.) a. "A man who was walking in the woods near a city stopped suddenly, very much frightened, and then ran to the nearest poUceman, saying that he had just seen hanging from the limb of a tree a a what? " b. "My neighbor has been having queer visitors. First a doctor came to his house, then a lawyer, then a minister (preacher or priest). What do you think happened there?" c. "An Indian who had come to town for the first time in his life saw a white man riding along the street. As the white man rode by the Indian said — 'The white man is lazy; he walks sitting down.' What was the white man riding on that caused the Indian to say 'he walks sitting down'?" 5. Arithmetical reasoning. (1 minute each. 2 of 3.) a. If a man's salary is $20 a week and he spends $14 a week, how long wiU it take him to save $300? 6. If 2 pencils cost 5 cents, how many pencils can you buy for 50 cents? c. At 15 cents a yard, how much will 7 feet of cloth cost? 6. Clock. (2 of 3. Error must not exceed 3 or 4 minutes.) 6 :22 Time required 8 :10 Time required 2 :46 Time required Al. Repeats 7 digits. (1 of 2. Order correct. Read 1 per second.) 2-1-8-3-4-3-9 9-7-2-8-4-7-5 THE STANFORD REVISION AND ITS HISTORY 21 YEAR XVI, AVERAGE ADULT. (6 tests, 5 montlis each.) 1. Vocabulary, 65 words. Score Total Vocab 2. Interpretation of fables. (Score 8.) (First explain what a fable is, and after reading each say, "What lesson does that teach us?") a. Hercules and wagoner h. Maid and eggs c. Fox and crow d. Farmer and stork e. Miller, son and donkey 3. Difference between abstract words. (3 real contrasts out of 4.) a. Laziness and idleness 6. Evolution and revolution c. Poverty and misery d. Character and reputation 4. Problem of the enclosed boxes. (3 of 4.) One large box con- taining: a. 2 smaller, 1 inside of each h. 2 smaller, 2 inside of each c, 3 smaller, 3 inside of each d. 4 smaller, 4 inside of each 5. Repeats 6 digits backwards. (1 of 3.) 4-7-1-9-5-2 5-8-3-2-9-4 7-5-2-6-3-8 6. Code, writes "Come quickly." (2 errors. Omission of dot counts half error. Illustrate with "war," "trench," and "spy.") Errors C-O-M-E Q-U-I-C-K-L-Y Time Method Al. 1. Repeats 28 syllables. (1 of 2 absolutely correct.) a. Walter likes very much to go on visits to his grandmother, because she always teUs him many funny stories. 6. Yesterday I saw a pretty little dog in the street. It had curly brown hair, short legs, and a long tail. Al. 2. Comprehension of physical relations. (2 of 3.) o. Path of cannon baU h. Weight of fish in water. c. Hitting distant mark. . . XVIII, SUPERIOR ADULT. (6 tests, 6 months each.) 1. Vocabulary, 75 words. Score Total Vocab 2. Binet's paper cutting test. Draws folds and locates holes. (If given, must come before XlVa.) 3. Repeats 8 digits. (1 of 3. Order correct. Read 1 per second.) 7-2-5-3-4-8-9-6. . . . 4-9-8-5-3-7-6-2. . . . 8-3-7-9-5-4-8-2. . . . 4. Repeats thought of passage heard. (1 of 2. E. reads each in about 3^ min.) "I am going to read you a Httle selection. Listen carefully, and when I am through I will ask you to tell as much of it as you can remember. Ready — " 22 STANFORD REVISION OF BINET-SIMON SCALE a. "Tests such as we are now making are of value both for the advancement of science and for the information of the person who is tested. It is important for science to learn how people differ and on what factors these differences depend. If we can separate the influence of heredity from the influence of environ- ment we may be able to apply our knowledge so as to guide human development. We may thus in some cases correct defects and develop abihties which we might otherwise neglect." "Many opinions have been given on the value of life. Some call it good, other call it bad. It would be nearer correct to say that it is mediocre, for on the one hand our happiness is never as great as we should like, and on the other hand our misfortunes are never as great as our enemies would wish for us. It is this mediocrity of life which prevents it from being radically unjust." 5. Repeats 7 digits backwards. (1 of 3.) 4-1-6-2-5-9-3 3-8-2-6-4-7-5 9-4-5-2-8-3-7 6. Ingenuity test. (2 of 3. 5 minutes each. If S fails on 1st, E explains that one.) a. "A mother sent her boy to the river to get seven pints of water. She gave him a 3-pint vessel and a 5-pint vessel. Show me how the boy can measure out exactly 7 pints without guessing at the amount. Begin by filling the 5-pint vessel." h. Same, except 5 and 7 given to get 8. ("Begin with 5.") c. Same, except 4 and 9 given to get 7. ("Begin with 4.") THE STANFORD REVISION AND ITS HISTORY 23 Time required. 1. orange. . . . 2. bonfire 3. roar 4. gown 5. tap 6. scorch 7. puddle 8. envelope. . 9. straw 10. rule 11. haste 12. afloat 13. eye-lash. . . 14. copper. . . . 15. health 16. curse 17. guitar 18. mellow 19. pork 20. impolite. . . 21. plumbing. 22. outward. . , 23. lecture. . . . 24. dungeon. . 25. southern. . 26. noticeable. 27. muzzle 28. quake 29. civil 30. treasury. . , 31 . reception. . 32. ramble 33. skni 34. misuse. . . . 35. insure 36. stave 37. regard. . . . 38. nerve 39. crunch 40. juggler 41. majesty. . . 42. brunette. . 43. snip 44. apish 45. sportive. . . 46. hysterics. . 47. Mars 48. repose 49. shrewd THE VOCABULARY TEST Score 51. pecuHarity 52. coinage. 53. mosaic 54. bewail 55. disproportionate. , 56. dilapidated , 57. charter , 58. conscientious ... , 59. avarice 60. artless 61. priceless 62. swaddle 63. tolerate 64. gelatinous 65. depredation 66. promontory 67. frustrate 68. milksop 69. philanthropy 70. irony 71. lotus 72. drabble 73. harpy 74. embody 75. infuse 76. flaunt 77. declivity 78. fen 79. ochre 80. exaltation 81. incrustation. . . . 82. laity 83. selectman 84. sapient 85. retroactive 86. achromatic 87. ambergris 88. casuistry 89. paleology 90. perfunctory 91. precipitancy. . . . 92. theosophy 93. piscatorial 94. sudorific 95. parterre 96. homunculus .... 97. cameo 98. shagreen 99. limpet. 50. forfeit 100. complot Note : To get the entire vocabulary, multiply the number of correct definitions by 180. 24 STANFORD REVISION OF BINET-SIMON SCALE The Fable Test "Fables, you know, are little stories which teach us a lesson. Now I am going to read a fable to you. Listen carefully and when I am through I will ask you to tell what lesson the fable teaches us." After reading each fable say, "What lesson does that teach us?" Ask also if fable has been heard before. A. Hercules and the Wagoner A man was driving along a country road, when the wheels suddenly sank in a deep rut. The man did nothing but look at the wagon and call loudly to Hercules to come and help him. Hercules came up, looked at the man, and said: "Put your shoulder to the wheel, my man, and whip up your oxen." Then he went away and left the driver. Lesson B. The Milkaiaid and Her Plans A milkmaid was carrying her paU of milk on her head, and was thinking to herself thus: "The money for this milk will buy 4 hens; the hens will lay at least 100 eggs; the eggs will produce at least 75 chicks; and with the money which the chicks will bring I can buy a new dress to wear instead of the ragged one I have on." At this mom- ent she looked down at herself, trying to think how she would look in her new dress ; but as she did so the pail of milk slipped from her head and dashed upon the ground. Thus all her imaginary schemes perished in a moment. Lesson C. The Fox and the Crow A crow, having stolen a bit of meat, perched in a tree and held it in her beak. A fox, seeing her, wished to secure the meat, and spoke to the crow thus: "How handsome you are! and I have heard that the beauty of your voice is equal to that of your form and feathers. Will you not sing for me, so that I may judge whether this is true?" The crow was so pleased that she opened her mouth to sing and dropped the meat, which the fox immediately ate. Lesson D. The Farmer and the Stork A farmer set some traps to catch cranes which had been eating his seed. With them he caught a stork. The stork, which had not really been steahng, begged the farmer to spare his life, saying that he was a bird of excellent character, that he was not at all like the cranes, and that the farmer should have pity on him. But the farmer said: "I have caught you with these robbers, the cranes, and you have got to die with them." Lesson THE STANFORD REVISION AND ITS HISTORY 25 E. The Miller, His Son, and the Donkey A miller and his son were driving their donkey to a neighboring town to sell him. They had not gone far when a child saw them and cried out: "What fools those fellows are to be trudging along on foot when one of them might be riding." The old man, hearing this, made his son get on the donkey, while he himself walked. Soon they came upon some men. "Look," said one of them, "see that lazy boy riding while his old father has to walk." On hearing this the miller made his son get off, and he chmbed upon the donkey himself. Farther on they met a company of women, who shouted out: "Why, you lazy old fellow, to ride along so comfortably while your poor boy there can hardly keep pace by the side of you!" And so the good-natured miller took his boy up behind him and both of them rode. As they came to the town a citizen said to them, "Why, you cruel fellows! you two are better able to carry the poor donkey than he is to carry you." "Very well," said the miller, "we will try." So both of them jumped to the ground, got some ropes, tied the donkey's legs to a pole and tried to carry him. But as they crossed the bridge the doiJiey became frightened, kicked loose and fell into the stream. Lesson CHAPTER II THE DISTRIBUTION OF INTELLIGENCE The question as to the manner in which intelligen^^e is distributed relates itself at once to fundamental issues in biological theory and suggests social and educational problems of great importance. Perhaps the most vital question which can be asked by any nation of any age is the following: ^^How high is the average level of mental endowment among our people, and how frequent are the various grades of ability above and below the average?'' With the development of standardized intelligence tests we are approaching, for the first time, a possible answer to this question. The future of such tests is guaranteed by the importance of the problems which they undertake to answer. This would still be true even if it could be shown that all the mental tests which have yet been devised or suggested are of little worth. Difficulties in Finding the True Distribution of Intelli- gence In view of the large number of investigations mad with the Binet-Simon tests in many countries, the light which these have thrown upon the distribution of intelligence is less than might have been expected. The reasons for this are various. In the first place, the number of children tested by any one investigator, and particularly the number at any one age, haB usually fallen short of that required for far-reaching; conclusions. If we could mass the results of different investigators the problem would be made much easier; 26 THE DISTRIBUTION OF INTELLIGENCE 27 but owing to the lack of uniformity in the methods by which the data have been secured, this is usually a dangerous procedure. Because of the small numbers we can seldom be sure that the children tested were representative. Educational advantages, social status, racial differences, and other possible selective influences must be taken into account. To get representative children of a given age is especially difficult, and it is by no means easy even when questions of education, social status and race have been eliminated. Studies of the progress of school children through the grades have shown that children of any given age are scattered over an astonishing range of grades. It has been a common mistake to select certain school grades for the testing and to suppose that the results secured could be used as norms for the ages found in those grades. One factor or another has entered to impair the value of almost every experiment with the scale. Kuhlmann, for example, in his tests of 1000 children, avoided selection by examining all the children enrolled in the public schools of a small middle-class city; but Ms examiners were untrained. Fewer tests by trained examiners would have made his experiment of greater value in several respects. Binet's 1908 scale was based on tests of only 200 children, 15 to 25 of each age, and these were situated in one of the poorest quarters of Paris. What further selection of subjects took place in this experiment we are not informed. Bobertag's subjects were in the main pupils attending the Volks- sdiule, and these are known to have a lower average L 5vel of mental endowment than pupils attending the higher schools. Goddard^s numbers were fairly large, but the tests were made by persons of limited training and at a rate (as high as thirty per day for one tester) 28 STANFORD REVISION OF BINET-SIMON SCALE which rendered thoroughness impossible. Some of the tests of Terman and Childs were made by only partly-trained examiners, and here, again, too little attention was paid to social class and age selection. In hke manner it would be easy to point out serious shortcomings in every study which has been made with the Binet scale, including those of Jeronutti at Rome, Treves and Saffiotti at Milan, Dr. Anna Schu- bert at St. Petersburg, Mrs. Wolkowitsch at Moscow, Bloch and Preiss at Kattowitz, Miss Johnston at Shef- field, Winch in London, Rogers and Mclntjrre at Aberdeen, Decroly at Brussels, Levistre and Morle in Paris, Miss Dougherty in Kansas City, Dr. Morse and Miss Strong in South Carolina, Dr. Rowe in Michigan, the Weintrobs in New York City, Dr. Schmidt in Chicago, etc. But even had the procedure and the method of selecting cases not been at fault in these studies, the results would still have been misleading as regards the distribution of intelligence, owing to the acknowl- edged imperfection of the scale used. The arrange- ment of tests in the earlier years magnified decidedly the amount and range of superior ability, and to a corresponding degree covered up the presence of re- tardation. At the upper end of the scale the tests were so few and so unsatisfactory that individual differences above the level of eleven years were hardly brought out at all. Only at the middle point was the scale reasonably accurate. As regards the distribution of intelligence we believe that the Stanford 1914-1915 data have more than ordinary significance, and for reasons which it may be well to enumerate: (1) The children were as nearly representative of the different ages as it is possible to get. The method THE DISTRIBUTION OF INTELLIGENCE 29 was to select a school attended by all the children of school age in the community, and to test the children of the various ages in whatever grades they might happen to be. This obviates accidental age selection at least for the years 7 to 13, inclusive. It is possible, however, that six-year-old and fourteen-year-old school children are not quite representative of children of those ages, since mentally retarded six-year-olds are likely to enter school a year late and since a consider- able number of fourteen-year-olds have been promoted to the high school. (2) The children tested were all within two months of a birthday. Our curve of distribution for nine- year intelhgence, for example, really represents the distribution of intelligence for nine-year olds, and not that of children ranging all the way from eight and a half to nine and a half years. (3) The schools selected for the tests were such as almost any one would classify as middle-class. Few children attending them were either from very wealthy or very poor homes. The only exception to this rule was in the case of Reno, where all the children within two months of a birthday were tested throughout the city. The large majority of these, however, were from homes of average wealth and culture. Only 2 per cent., in fact, were classified by the teachers as of ^'very inferior" social status, and only 1.6 per cent, as of ^'very superior" social status. (4) Care was taken to avoid racial differences and the difficulties due to lack of familiarity with the language. None of the children was foreign-born and only a few were of other than western European descent. The names were chiefly English, Irish, Scotch and German, with a few Swedish, French, Spanish, ItaUan, and Portuguese. Tests of Spanish, 30 STANFORD EEVISION OF BINET-STMON SCALE Italian and Portuguese children were eliminated from our study of distribution, for the reason that in west- em cities children of these nationalities are likely to belong to unfavorably selected classes. We are justi- fied in beheving, therefore, that the distribution of intelligence among our subjects is less influenced by extraneous factors than has been the case in other studies of this kind. (5) The numbers tested were relatively large, namely 80 to 120 at each age from 6 to 14 years, with somewhat fewer at 5 and 15. Moreover, the use of only such children as were within two months of a birthday greatly enhances the value of the numbers employed. In each school this near-birthday group composed approximately one-third of all the children in attendance, and by the laws of chance we can be sure that the results are approximately as they would have been had the entire school enrollment been tested, that is, 3000 instead of 1000. This plan has the further advantage that three times as many schools were sampled as would otherwise have been the case; and if it happened, in spite of the care exercised, that some of the schools were below average in social status, this would probably be counterbalanced by other schools somewhat above average. (6) The scale by which the mental ages were com- puted is certainly much more accurate than Binet's or the earlier revisions. As already stated, the tests were worked over until the average mental age se- cured at each level approximately coincided with the physical age. In this way the scale was made of nearly equal accuracy at every point. (7) Correcting the rather large error of the scale at the upper and lower ranges gives another advantage of extreme importance, for it enables us to combine THE DISTEIBUTION OF INTELLIGENCE 31 the intelligence quotients of the children of all ages into a single surface of distribution. As long as in- telligence was reckoned in terms of years and months of retardation or acceleration it was of course not permissible to combine the distributions for different ages. The range of mental ages is approximately twice as great at 10 years as at 5 years, so that one year of retardation or acceleration at 5 is equivalent to two years of retardation or acceleration at 10. Ac- cordingly, to combine the results for children of several different ages so as to show what percent of the entire number are retarded or accelerated one year, two years, three years, etc., is an absurd procedure. The range of intelligence quotients, however, as measured by the revised scale, is not far from constant from five to fourteen years, and these may therefore be combined into a single surface of distribution. The curve thus obtained differs from those of the indi- vidual years only in being somewhat more regular. (8) Finally, though by no means least in importance, the tests from which the following curves of distribu- tion were derived were made with more than ordinary care. The examiners were trained, the procedure was kept as uniform as possible and the scoring was all done by one person. Wherever evidence was found of mistaken procedure, the examiner was questioned, and if the records of that examiner for that test were not comparable with the others, they were thrown outv After making the necessary eliminations because of incomplete testing, foreign parentage, etc., there remained 981 children, distributed as shown herewith, TABLE 2 Age 4 5 6 7 8 9 10 11 12 13 14 15 16 Cases.... 16 54 117 92 100 113 87 97 83 98 82 46 14 32 STANFORD REVISION OF BINET-SIMON SCALE The Distribution of Intelligence for the Ages Separately The intelligence quotients were calculated for all the children and those for a given year were then thrown into groups so that each group represented the cases include in a range of ten points intelligence quotient. The middle group includes the intelligence quotients from 96 to 105; the ascending groups include in order those from 106-115, 116-125, 126-135, and 136-145. The corresponding descending groups are 86-95, 76-85, 66-75, and 56-65. Only one case tested below 56, a girl of 8 years who had an intelligence quotient of 51. None tested above 145. Graphs 1 to 13 show the results of this grouping for each age sepa- rately from 4 to 16. It is evident at a glance at these graphs that the distribution of intelligence quotients is fairly sym- metrical at each age from 5 to 14. At 14 the number of exceptionally superior children decreases, as we should expect, since a few of the brightest have by that age finished the eighth grade. At 15 the intelli- gence quotients range on either side of 90 as a median, and at 16 years on either side of 80 as a median. This, again, is what we should expect, because we know from other facts that a majority of school retardates are below average in intelligence. The 17 4-year-olds averaged high, with a median intelligence quotient of 103. It is reasonably certain, however, that children of this age attending school are usually somewhat beyond the average in intelligence. It will be noted that at every year from 5 to 14 the middle group is the largest and that the groups on either side decrease in size somewhat regularly with increasing distance from the median. In the lower years, however, the 106-115 and 116-125 groups are larger than the 86-95 and 76-85 groups respectively THE DISTKIBUTION OF INTELLIGENCE 33 5&-65 66-7S 7G-8S 86-SS 9G-I05 lOG-IIS ilB-lZS l2Q>-ldS l5Q-/fS 9'Jo lyi. 327. zn, 18I0 f'h Graph 1. Distribution of intelligence quotients of 16 kindergarten children, age 4 years. (These pupils are a selected group and test high.) J-//S //6 -/^S /2G-/JS /J6-/¥-J^ Z'l, (o'l. Zb'l. it'/. 26'/. ^°'. Graph 12. Distribution of intelligence quotients of 47 children, age 15 years (over-age for grade). (Median 90.) Most of the average and superior 15-year-olds had been promoted to the high school and do not appear in the above graph. 5Q>-^5 G€>-7S /f/. 2Z'/. Graph 13. 7G-8S 86 -3S $6 -tOS lOb -US /Ib-IZS IZC ■IJS/3^-/4-S 35'/, zru 7% Distribution of intelUgence quotients of 14 children, age 16 years (over-age for grade). (Median 80.) AH of the average and superior 16-year-olds had been promoted to the high school and do not appear in the above graph. THE DISTRIBUTION OF INTELLIGENCE 39 on the other side. The most noticeable lack of sym- metry occurs at ages 5, 6, and 14. This is due in part to a certain amount of unavoidable selection. The five-year-olds were enrolled in kindergartens, and since school attendance at this age is not compulsory, we can not be sure that kindergarten children represent the median intelligence for five-year-olds. The same is true of six-year-olds, though to a less extent. In both cases the distribution of intelligence quotients suggests that at these ages inferior children are some- what less likely to be found in school than those of superior endowment. The reverse is the case at 14, since a few of the brighter children of that age have completed the eighth grade and have either dropped out or passed on to the high school. It is possible that a few children under 14 have managed to evade the compulsory attendance laws and are not in school, but it is certain that in the cities and towns where our testing was done the amount of such evasion was practically negligible. On the whole, it is evident that the distribution of intelhgence quotients is fairly regular and uniform in the various years. This is further shown by the fact that the range including the middle 50 percent of the intelligence quotients does not vary greatly between 5 and 14 years. Combining each two successive years in order to overcome the chance effects of the limited number of children of any one age, we have the ac- companying table in which, it is of interest to note, the distribution of intelligence quotients for the differ- ent age-levels furnishes no support to the very generally accepted belief that variability materially increases at adolescence. As far as 14, at least, there is no evidence that this occurs. 40 STANFORD REVISION OF BINET-SIMON SCALE TABLE 3 Limits of quotient Range of quotients Ages including middle 50% including middle 50% 5 and 6 combined 97 to 111 15 7 and 8 combined 95+ to 111 16+ 9 and 10 combined 94 to 108 15 11 and 12 combined 92 to 108 17 13 and 14 combined 90 to 105 16 All ages combined 92+ to 108 16+ The Distribution of Intelligence for the Ages Combined As already explained, the corrected scale has the advantage of enabling us to combine the intelligence quotients of the children of different ages into a single surface of distribution, something that we could not do when the scale was too easy at some levels and too hard at others. By combining ages 5 to 14, we have the distribution shown in Graph 14. Ages 4, 15 and 16 have been omitted from this combined dis- tribution because of the selection which has taken place at these years. ^ S6'GSGG-rS 7G-8S 8e-9S SG-/05 /0&-//S //G-/£^ /2e-/JSf36-/^5' .337. 2.37. 8.GX 20./7. JJ.S'A 23J7. 901 zyL ^Sl Graph 14. Distribution of intelligence quotients of 905 unselected children, ages 5-14 years. (Median at 99.) ^ At 15 and 16 there is the additional reason, that growth at this time is probably slowing down suflSciently to impair the vaUdity of the intelligence quotient. THE DISTRIBUTION OF INTELLIGENCE 41 Exception may be taken to this combined distribu- tion on the ground that it fails to take account of possible selection which may have taken place at ages 5, 6, 13, and 14. As shown in Graph 15, however, the distribution for ages 7 to 12 combined is little different from that for 5 to 14 combined. f£-G5 G6'7S 7e-B5 8&-35 9G-I0S J0&-I15 l/G-125 {ZS-/35 /3Q-li-5 XL 1:87. 8% 13'/. Jjr. EiJ'L 3.81 Z'L 7'A Graph 15. Distribution of intelligence quotients of 554 unselected children 7-12 years of age. (Sexes combiaed.) TheHntelligence quotients may, of course, be grouped in ranges of any desired extent. If the numbers dealt with had been larger, it would have been interesting to group them in ranges of 5 instead of 10. With the numbers available, however, the curves resulting [from this method of grouping would be much less regular than when larger groups are used. On the other hand, if ranges above 10 are used for the grouping, the distri- bution becomes still more symmetrical. This may be seen from Graph 16, which shows the relative sizes of groups contained in ranges of 20, for ages 5 to 14 combined. 42 STANFORD REVISION OF BINET-SIMON SCALE Graph 16. Distribution of intelligence quotients of 905 unselected children, age 5-14 years. Intelligence quotients grouped in ranges of 20. Followdng is a comparison of the observed distri- bution (Ages 5-14) with that called for by the theoretical normal" curve of distribution (the Gaussian curve): TABLE 4 Intelligence Quotient Ranges (( Obtained Theoretical 56-65 .33 .4 66-75 2.3 3.1 76-85 8.6 11. 86-95 20.1 22. 96-105 33.9 27. 106-115 23.1 22. 116-125 9. 11. 126-135 2.3 3.1 136-14? .55 A The normal nature of the distribution of mental ability is further borne out by the teachers' rankings according to intelligence and school work, which were made according to the form given in the appendix. These rankings were made in a httle more than half THE DISTRIBUTION OF INTELLIGENCE 43 the schools and included 489 children, about equally divided between the sexes. The results are shown on page 48. The distribution shown in the graphs of this chapter are for the boys and girls combined. There are cer- tain sex differences, however, which will be set forth in Chapter III. We may ignore these for the present, since they are not great, and proceed to an examination of the frequency with which various departures from the normal are encountered. An examination of the distribution for all the ages from 5 to 14, taken together, and for the sexes com- bined, gave the following significant facts: TABLE 5 rhe lowest 1 percent go to 70 or below; the highest 1 percent reach 130 or abov 2 3 5 10 15 20 25 33.3 u u 73 . a 2 u 128 " " u u 75 c u 3 u 125 " " u u 7g c a 5 u 122 " " u u g5 i " 10 " u 115 u u a u gg ^ u 15 u u 113 u u a u 91 . " 20 " u 110 u u " " 92+ ' " 25 " U IQg u u " " 95 ' " 33.3 " " 106+ " " A perfectly normal distribution would cause 93 per- cent of the cases to fall between 76-125, instead of the observed 94.7 percent; and 99.2 percent between 66-135, instead of the observed 99.3 percent. Or, to put some of these facts of distribution in another form, we may say, speaking approximately: The child reaching 110 is equaled or excelled by 20 out of 100 u a u 115 u u u u u iQ u u jqq " " " 125 " " " " " 3 " " 100 II II (I 1^0 " '' " '' " 1 " " 100 Again, for those whose intelligence quotient is below 100: The child testing at 90 is equaled or excelled by 80 out of 100 u u u u g5 U U U U U 9Q u u iQo it cc (c li 75 a a a (i u 07 ce ci lAT) a (( ii (I 7Q II u ii a li QQ a n iqq 44 STANFORD REVISION OF BINET-SIMON SCALE When we examine the above data, it is difficult to avoid the conclusion that superior intelligence of any given degree occurs with approximately the same frequency as intelligence which is inferior to a corres- ponding degree. The usual assumption, however, is that extreme degrees of mental deficiency are much more numerous than extreme degrees of mental superiority. As far as intelligence quotients between 60 and 140 are concerned, our figures do not support this assumption.2 As regards the relative frequency of intelligence quotients below 60 and above 140, the assumption is in all probability valid; for while all idiot and most imbecile children have an intelligence quotient as low as 50, there are extremely few cases of budding genius which reach as high as 150. Indeed, notwith- standing diligent search, the writer has found only a few cases testing above 150, and only two testing as high as 170.3 The significance of various intelligence quotients will be dealt with more fully in another chapter. It is 2 This statement requires some modification in view of the fact that om* data were collected entirely from children who were attending pubHc schools. There are, of com-se, in any community a few children with inteUigence quotients between 60 and 80 who are either kept at home or placed in institutions. No investigation seems to have been made which would show what proportion of such children are not in school, but our experience suggests that it is very small. At any rate, relatively few children testing as high as 65 or 70 are sent to an institu- tion until they have first been tried for several years in the schools, usually until well toward adolescence. It is well known that only a small minority with an inteUigence quotient of 75 to 80 ever get into an institution for the feeble-minded. 3 This is not to deny that cases of considerably higher intelUgence quotient are to be found. A five-year-old child reported by Miss Langenbeck seems to have tested not far from ten years. If the test can be accepted at its face value, the child, therefore, had an intelhgence quotient of about 200. See A study of a five-year-old child, Pedag. Seminary, March, I9I5, 65-88. More recently Dr. Leta S. Holhng- worth has favored us with a copy of a test of an eight-year boy whose inteUigence quotient is 190. THE DISTEIBUTION OF INTELLIGENCE 45 already evident, however, that the term '^ feeble- mindedness '^ is a matter of arbitrary definition. In one sense it could be said that a child with an intelli- gence quotient of 85 or 90 is as truly feeble-minded as one testing at 50, only he is mentally feeble to a much less degree. It becomes merely a question of the amount of intelligence necessary to enable one to get along tolerably with his fellows and to keep somewhere in sight of them in the thousand and one kinds of competition in which success depends upon mental ability. The definition of feeble-mindedness, too, is a constantly shifting one. Until recent years the standard was one which would have classed a majority of children having two-thirds intelligence (intelligence quotient 67) in the normal group. Even yet the usual medical standard is no higher than this. The child of moron grade is rarely classified by the physician as ^' feeble-minded." Social workers, psychologists, and criminologists, however, are constantly meeting facts which would seem to justify the application of the term feeble-minded to many children with three- fourths intelligence (intelligence quotient 75). It is possible that the development of civilization, with its inevitable increase in the complexity of social and industrial life, will raise the standard of mental nor- mality higher still. But whatever the standard, the number of borderline and debatable cases will probably be greater than the number of those whom all would agree to call feeble- minded. The attempt to classify all children as either definitely feeble-minded or definitely normal involves exactly the same difficulties as we should encounter in trying to classify all adult men as either ^'normally tall'' or ^ ^abnormally short," and we may add that 46 STANFORD REVISION OF BINET-SIMON SCALE the one attempt is just about as much worth while as the other. To regard feeble-inteUigence as always a disease, which, like small-pox, one either has or does not have, is a view which is contradicted by all we know about the distribution of mental traits. Physicians find special difficulty in freeing themselves from this fallacy, since for them diagnosis consists essentially in deter- mining the presence or absence of definitely patho- logical conditions. There is other evidence than that of mental tests to support the hypothesis that intelligence is dis- tributed in the manner indicated by oiu" distribution of intelligence quotients. It has often been shown that a similar distribution results when teachers are asked to classify children into three groups (or five groups) according either to school success or intelli- gence. Thus, Bobertag had teachers classify 2772 pupils into three groups according as their school work was '^unsatisfactory," '^ satisfactory," or '^ better than satisfactory. ' ' The resulting classification showed 50.8 percent in the middle group, 25.7 percent in the superior group, and 23.5 percent in the inferior group. The numerous other studies which have been made of teachers' marks give similar results, though in some cases the curve shows a more noticeable tendency to be skewed in the direction of superiority. About half our teachers supplied for each of their children tested the supplementary data called for on the blank shown in the appendix. It will be noted that Question 4 calls for a classification of the children into five groups according to the quality of school work, and Question 5 a similar classification on the basis of the teacher's judgment as to a child's intelli- THE DISTRIBUTION OF INTELLIGENCE 47 gence. The classes were designated in the blank as "very inferior/' "inferior," "average/' "superior" and "very superior." The resulting distributions are shown herewith. TABLE 6 Percents Rated Very Very Inferior Inferior Average Superior Superior Schoolwork (503 Chil- dren) 5.2 17.9 51.0 22.1 3.8 Teacher's judgment of intelligence (489 Children) 3.4 14.4 55.8 23.1 3.3 If for quality of school work, we combine the two superior groups and combine similarly the two inferior groups, the distribution coincides remarkably with that found by Bobertag. TABLE 7 Inferior Bobertag 23 . 5 Ours 23.1 It is interesting to compare the teachers' groupings for intelligence with those for school success, as shown in Graphs 17 and 18. The "piling up" of the intelligence distribution in the direction toward the "superior" end indicates that teachers are able to judge the degree of a child's intelligence less objectively and therefore less accu- rately than they judge the quahty of his school work. Personal factors are more likely to enter into judgments about intelligence, and in case of uncertainty as to the proper classification there is a natural tendency to give the child the benefit of the doubt, even at the risk of grading him too high.^ 4 The relation between the inteUigence tests and the teachers' judg- ments of inteUigence is treated in full in Chapter VI. Average Superior 50.8 25.7 51.0 25.8 48 STANFORD REVISION OF BINET-SIMON SCALE \/ery Inferior Jnjtnor Average Superior Very superior 3.1-7. /f/f-y- 55 8'/. 23l'f. 33'/r Graph 17. Distribution of teachers' rankings of 489 children ac- cording to intelligence. Very Inferior Inftnor Average Superior Very superior 5.Z% 17.97. 51'/. 221'/. 3.81 Graph 18. Distribution of teachers' rankings of 503 children" ac- cording to quaUty of school work. (Mostly the same children as appear in Graph 17.) THE DISTRIBUTION OF INTELLIGENCE 49 The ''Mental Ages'' of 62 Normal Adults The use of the Stanford revision with 30 business men of moderate success and of very limited educa- tional advantages gave a distribution of mental ages differing Uttle from that found with 32 high school pupils pupils from 16 to 20 years of age. The results for both groups are shown in tabular form. TABLE 8 Mental Age 13 to 14 to 15 to 16 to 17 to 18 to 13-11 14-11 15-11 16-11 17-11 18-11 Business Men 1 6 7 8 6 2 H. S. Pupils 5 12 10 4 1 If we combine the business men with the high school pupils who are over 16 years of age chronologically, we have the distribution of mental ages shown in Graph 19. It will be noted that the middle part of the graph represents the "mental ages'' falUng within the range of two years, namely 15 to 17. This range we may designate as the "average adult" level. Summary 1. The revised scale gives a median intelHgence quotient of approximately 100 when used with unse- lected children of any age from 5 to 14. 2. The distribution of intelligence quotients for unselected children of each age conforms fairly closely to the Gaussian curve. This holds particularly for our subjects of ages 7 to 13. 3. Since the revised scale yields the same form of distribution of intelhgence quotients at each age, it is permissible to combine the intelhgence quotients for the different ages from 5 to 13 or 14 into a single surface of distribution. 4. The mental ages found by testing 30 uneducated business men and 32 high school pupils over 16 years of age range from the "inferior adult" level to the "superior adult" level, with the greatest number at. "average adult." 50 STANFORD REVISION OF BINET-SIMON SCALE l3to 13-11 H-U 1^-11 l5UiS-ll I7UI7-II iShlB'll LQr. nil, S9.77, IGZ'A 4-.SZ Gkaph 19. Mental ages of 62 adults, including 30 business men of little education and 32 high school students over 16 years of age. CHAPTER III THE RATE OF GROWTH AND THE VALIDITY OF THE INTELLIGENCE QUOTIENT The previous chapter showed that for unselected children the distribution of inteUigence as measured by the revised scale maintains a certain constancy from 5 to 13 or 14 years, when the degree of intelligence is expressed in terms of the intelligence quotient. Any given deviation from the median occurs with much the same frequency at all the ages. The intelligence of children has usually been esti- mated, however, in terms of years and months of retardation or acceleration. Binet, while using this method, realized that a year of retardation is less serious with older children than with younger, and accordingly he suggested the rule that while a retarda- tion of 2 years usually means feeble-mindedness in children under ten years of age, older children should not be regarded as feeble-minded unless retarded as much as 3 years. This is obviously crude, but Binet did not suggest any more definite adjustment to allow for the decreasing significance of a given amount of retardation in the upper years. Even this slight adjustment is often ignored by those who use the scale. One person, after testing a large number of juvenile delinquents ranging from 10 to 18 years in age, lumps all the ages together and counts up the number who were retarded 1 year, 2 years, 3 years, 4 years, etc., concluding finally that about 75 percent of the total number were feeble-minded, since that many were retarded 3 years or more. This error appears again and again in the Hterature of Binet testing. Others, starting from the same erroneous 51 52 STANFORD REVISION OF BINET-SIMON SCALE assumption, have defined ^^at-age'' intelligence as that which is within one year of the child's physical age and have expected to find the number of children testing ^^at-age" to be the same at all chronological ages. It is obvious, however, from the distribution of intelhgence quotients as shown in Chapter II, that a given number of years of retardation can have no definite meaning except in relation to the age of the subject. Whatever the age of a group of non-selected children, approximately the same percent will always be included in any range of the intelligence quotients. As already shown, the middle fifty percent are at all ages included in the range of about 92 or 93 intelligence quotient to 108 or 109 intelligence quotient. Trans- muting these values into months, we have for 6-year- olds, 50 percent included in a range of a little less than one year of mental age; while for 12-year-olds, the middle 50 percent range over about twice this distance, or nearly two years. Retardation of two years is about as common at 12 years as retardation of one year at 6; and either is about as common as retardation of a year and a half at 9. That is, the curve of dis- tribution of mental ages becomes progressively flat- tened, the older the children with which we deal. This is shown in Graphs 20, 21, and 22, which give the distribution of mental ages for children of 6, 9, and 12 years, respectively. The range including 50 percent of the mental ages increases in a fairly constant ratio from Age 6 to Age 14, as shown in Graph 23. The use of the intelhgence quotient as a means of expressing a child's intelligence status is based, of course, on the assumption that the intelligence quo- VALIDITY OF THE INTELLIGENCE QUOTIENT 53 -3 -2 -/ 5-7 +1 +2 +3 o7. 07. 7.37. 797. iZ'U 2.7% 7, Graph 20. Distribution of mental ages of the 117 iinselected 6-year- olds. -■f -3 -2 -1 8-10 +/ O'/o 07o 4-'/z '1, 16 V, eo'/o /5% + 2 7. + 3 oy. Graph 21. Distribution of mental ages of the 113 unselected 9-year- olds. -•f -3 -2 -/ 11-13 +/ ^2 +3 +^ ^7, 3.5% /2 7o !?'/> ^3 7o 1^3% 8.5% I'l. 07» Graph 22. Distribution of mental ages of the 83 imselected 12-year- olds. 54 STANFORD REVISION OF BINET-SIMON SCALE Graph 23. Showing range of months including the middle 50% of mental ages at various years. tient remains practically constant during the years of mental growth; that, for example, the child of five years who tests at 4 (intelligence quotient 80) will at later ages have the mental ages shown in the follow- ing figures: 10 monThi 13.4- monThs) J6 mortThi zo ■monThs ZG moTiThi TAB LE 9 8 9 10 11 12 13 14 6.4 7.2 8 8.8 9.6 10.4 11.2 Physical Age 5 6 7 Mental Age.... 4 4.8 5.6 6.4 7.2 The facts which have already been presented argue in favor of the validity of the intelligence quotient at least for Ages 5 to 13 or 14. It has been shown that the distribution of intelligence quotients for the different years remains essentially the same, and that the distribution of mental ages (in terms of years and months) flattens out in the upper years in approxi- mately the expected ratio; 79 percent test within one year of physical age at 6, 60 percent at 9, and 43 percent at 12. The percents called for at 9 and 12 by a theoretically valid intelligence quotient would be 59.25 and 39.5, granting 79 percent to be correct for VALIDITY OF THE INTELLIGENCE QUOTIENT 55 Age 6. The range of mental ages including the middle 50 percent of cases is 10 months at Age 6, and increases in the succeeding years as shown in the accompanying table. TABLE 10 Observed Percentage Percentage of In- Range Including of Increase Over crease Called for by- Age Middle 50 Percent That of Year 6 a TheoreticaUy VaUd Intelligence Quotient 8 13.4 months 34 33.3 10 16 months 60 67.6 12 20 months 100 100 The crucial test of the validity of the intelligence quotient would be to measure the intelligence of the same children several times during their period of mental growth. No experiment of this kind appears to have been made on any considerable scale, barring a few repetitions of tests after an interval of only one year. The results of such tests, however, support in a general way the hypothesis that the intelligence quotient of a given child tends to remain constant.^ The matter has been complicated, however, by the uneven inaccuracy of the Binet scale at different levels. Repeated tests are being made of a considerable number of children with the Stanford revision, and although the investigation is not complete at this writing, the results of 140 such tests show that as far as the age of 13 or 14, even when the tests are separated by as much as five years, changes of 10 points in 12 are relatively rare. In general, it can be said that the superior children of the first test are found superior in the second, the average remain average, the inferior ^ See W. Stern : Der Intelligenzquotient als Mass der kindlichen Intelligenz, inbesondere der unternormalen. Zeitsch. f. Angewandte Psychologie, 1916, Bd. II, Heft, 1-19. (Argues for the constancy of the inteUigence quotient from 7-12.) 56 STANFORD REVISION OF BINET-SIMON SCALE remain inferior, the feeble-minded remain feeble- minded, and nearly always in approximately the same degree. The most marked exceptions to this rule are found with the feeble-minded, whose intelli- gence quotient shows a tendency to decrease consid- erably. If future investigations should confirm the validity of the intelligence quotient and its necessary corol- laries, the practical consequences would be of the greatest importance. It would mean that, after a mental test consuming no more time than an ordinary medical examination, the psychologist would be able to predict, with some degree of accuracy, the future of the child's mental development. There is nothing else which the average parent would more like to know about the child, and nothing else which would prove of greater value in directing its education. Whether the intelligence quotient holds even ap- proximately with very young children, or with children much beyond the age of 14, is a question on which the data available afford little light. We are war- ranted in believing, however, that general intelligence practically ceases to develop by the age of 18 or 20 years. The mental age of high-grade morons appears to change little after the age of 14 or 15 years. With normal children development continues a little longer, though at a decreasing rate. It is practically certain, however, that growth of intelligence comes to a stand- still somewhere in the later years of adolescence, and that the cessation is gradual rather than sudden. It is evident, also, that beyond the time when the cessa- tion begins, the intelhgence quotient rapidly loses its meaning. The 12-year-old moron with a mental age of 8 years has an intelligence quotient of 67, which VALIDITY OF THE INTELLIGENCE QUOTIENT 57 at this period of life probably indicates his mental status fairly well. When 6 years old, the same child probably had a mental age of about 4 years, and when 9 years old, a mental age not far from 6 years. But inasmuch as mental growth slows down rapidly some- time after the age of 14 or 15, the mental age of this subject is unlikely ever to go beyond 10 years. Sup- posing it to stop at 10, the intelHgence quotient, if we continued to use it, would be reduced to 50 at the age of 20 years, to 25 at the age of 40 years, etc. Such a use of the term, of course, would be absurd, since the subject's intelligence is really a constant quantity throughout adult life. If it could be shown that mental growth continues its earlier rate up to a certain age, say 16, and then stopped quite suddenly, we could continue to use the intelligence quotient with adults of any age by merely ignoring the years beyond 16. That is, all adults, for purpose of reckoning the intelligence quotient, would be regarded as exactly 16 years of age.^ There are two difficulties, however, with a plan of this kind: (1) mental growth probably does not come to a stand- still suddenly; and (2) the time of its cessation is not accurately known. A practical way to get at the matter is to adopt some hypothesis of this general nature, a quite tenta- tive one, of course and by subjecting it to the pragmatic test of experiment, to see whether it is in harmony with ascertainable facts. If the hypothesis first adopted is unable to satisfy the requirements, it may be altered or replaced by a better. In this way, by 2 Such a scheme would demand, however, that the upper end of the scale be so framed that the inteUigence of superior adults as well as that of superior immature subjects could be expressed in an intelligence quotient above 100. 58 STANFORD REVISION OF BINET-SIMON SCALE checking up every step and profiting from our mis- takes, we should be able finally to arrive at a solution of the problem which would be correct enough for all practical piu-poses. Some of the data which have been presented would seem to justify the assumption, as a tentative working basis, that mental age maintains approximately a fixed ratio to chronological age until the latter has reached about 14 or 15 years, that during the next year or two the ratio diminishes, and that after the chronological age of 17 or 18 years, mental age re- mains constant. According to this hypothesis, the intelligence quotient would be a proper expression of the intellectual status with subjects as old as 14 or 15 years. This is the hypothesis which has guided us in the extension of the scale at the upper end. We have at least succeeded in shaping it in such manner that a child, for example, whose mental age at 7 was 8 years (intelligence quotient about 115) will have at 14 a mental age in the neighborhood of 16 (intelUgence quotient about 115), with the possibility of further increasing his mental age considerably before growth ceases. Our high school students usually test between 15 and 17 years, as do also Knollin's and Zeidler's business men. College students average slightl> higher, as we should expect from the fact that they belong to a selected group. It will be understood naturally that the numbers expressing such mental ages as 17 years, 18 years, 19 years, etc., have only a conventional value and are not to be interpreted literally. Their use offers a feasible, if arbitrary, method of enabling the superior adolescent or adult to earn a quantitative expression VALIDITY OF THE INTELLIGENCE QUOTIENT 59 of his superiority in the tests. Tentatively, we may use the intelligence quotient with normal adults by merely ignoring years of age beyond 16. That is, the adult^s chronological age is always, for this pur- pose, reckoned as 16. Further trial of our revision by repeating the tests between early and late adolescence, supplemented by tests of different groups of adults, will determine the adequacy of our arrangement. It is not offered as a finished product, but as material for further elaboration and refinement. Although the intelligence quotient maintains a fairly constant value from rather early in childhood until late in the growing period in the case of children of all grades of intelligence above mental deficiency, it is possible that this constancy may not be main- tained with defectives, particularly those of low grade. The child of 4 years who has a mental age of 1 year is an idiot and may never develop higher than a mental level of 2 years, perhaps not so high. His intelligence quotient of 25 at 4 years will gradually diminish, say to 15, at the age when mental maturity is attained in the normal child. We must look to the institutions where low-grade defectives are abundant to supply the facts regarding mental growth in these subjects. In closing this discussion it may be interesting to point out that the facts presented in this and the preceding chapter are not entirely in harmony with certain wide-spread opinions about the rate of mental growth. The view has often been expressed that intelligence normally develops by alternate leaps and rests. Starting from observations on certain instincts, the doctrine of "nascent stages" has come to be applied to the phenomena of mental development generally 60 STANFORD REVISION OF BINET-SIMON SCALE and is now almost a dogma in the literature of child psychology. Researches are rapidly showing, how- ever, that instincts themselves have less a Minerva- birth than we had supposed, a fact which Freudian psychology has demonstrated in the case of the sex instincts. As far as general intelligence is con- cerned, there is little evidence of periodicity or irregu- larity. If such periodicity or irregularity occurred, the intelligence quotient of the developing child would be now high, now low, instead of maintaining a fairly constant value. ^ The facts we have presented offer little hope to the parent who would like to believe that his backward boy who is 6 or 8 years old will '^ catch up" in the supposed spurt of adolescence. Indeed, the much- talked-of adolescent spurt begins to look like a myth. These same facts, however, furnish some consolation to the parents of a young genius. It has been the custom for teachers and even for some psychologists to inspire in them a dismal and uneasy fear that such a child is in danger of retrograde development. We hear of great men who in childhood were famous blockheads, and of genius children who became numb- skulls! Perhaps it would not be safe to assert that such cases do not occur, for logic teaches that uni- versal negatives are hard to establish. If they do occur, we may suppose that a concrete example will sometime come to light, vouched for by the necessary scientific proof. 3 Baldwin has recently shown that physical growth also proceeds at a nearly uniform rate from 7 to 14 years of age, and that a child's physical status, as expressed in height and weight, maintains a fairly constant position, with reference to norms, from age 6 or 7 through adolescence. Bird T. Baldwin: Physical Growth and School Progress. Bull U. S. Bur. of Ed., 1914, No. 581. VALIDITY OF THE INTELLIGENCE QUOTIENT 61 Summary 1. Retardation or acceleration of any given number of years has no meaning apart from the age of the child, and the method of expressing intelligence status in absolute years only should be abandoned. 2. The distribution of intelligence quotients for the separate years argues strongly in favor of the intelligence quotient as a valid method of expressing a child's development status, at least for the years 5 to 14. 3. Data are presented which indicate that the intelligence quotient of a child of any grade of intelli- gence remains fairly constant until well into the period of adolescence. Doubt is thrown upon the existence of the supposed ''nascent stages," the "adolescent spurt," and other popularly assumed irregularities of mental growth. CHAPTER IV Sex Differences Our revised scale has been constructed by massing the results from boys and girls, and our discussion of the distribution of intelligence in Chapter II took no account of sex differences. However, when we treat the intelligence quotients of the boys and girls separately, we find a somewhat constant, though rather slight superiority of the girls from Ages 5 to 13, with the exception of Age 10. At 14 years the boys appear to be about as superior to the girls as the girls were superior to the boys at 5. This is shown in Graph 24, which gives the median intelligence quotient for the boys and girls separately at each age from 5 to 14. Graphs 25-34 show the distribution of intelligence quotients for the sexes separately, when grouped in ranges of ten, 56-65, 66-75, 76-85, 86-95, etc. Be- cause of the small number of boys or girls in any one year, successive ages have been combined for the graphs, as 5-6, 7-8, 9-10, etc. If, now, we combine the intelligence quotients for all ages from 5 to 14, inclusive, we have Graphs 35 and 36 for boys and for girls, respectively. Comparison of the sexes with regard to the range of intelligence quotient that includes the middle 50 percent shows that half the boys He between 91-107 and half the girls between 93 and 109. Table 11 shows the frequency of some of the more extreme degrees of variation according to sex. 62 SEX DIFFERENCES 63 1.00 .90 .60 ,4-0 .ZO r 6 7 8 3 10 II 12 13 14- Soys 1.00 .9S I.OI 1.00 .98 m SG .97 M IDO Girls 1 04- 1.05 I.OZ I.OZ 1.01 t.03 LOl ,99 S7 .96 Graph 24, Showing median intelligence quotients for boys and girls separately at each age, number of boys, 457; girls, 448. 64 STANFORD REVISION OF BINET-SIMON SCALE ^e-GS 6G-75 76-85 86-95 9Q-I05 /oe-llS IIG-IZS I2G-I55 136-1^5 ri, in, ZZ'L 3G7, I83'L 9'J. ZS'U Graph 25. Distribution of intelligence quotients of 87 boys, ages 5 and 6 combined. 5G'G^ e&'7j5 76-85 86-S5 96-/05 i06-ii5 Ii6-i25 i26'i35 l3G'i4'5 I'l, 5% 14-V. 3G'/, Z3'/. I^y. 67. l7. Graph 26. Distribution of intelligence quotients of 87 girls, ages 5 and 6 combined . SEX DIFFERENCES 65 c 56-65 66-75 7S'85 SG-95 9e-/05 106-//5 JJG'/25 /26-J35/5G-l4'5 1% /7o ^.57, 165% "hl'L Z37. lOV. Z'U ''U Graph 27. Distribution of intelligence quotients of 100 boys, ages 7 and 8 combined. 56-85 66-75 76-35 86-95 96-105 106-115 II6-/Z5 /26-135138'l'^S ru 7'U 19% 3J'L Z^'i. iZ'h "h'A Graph 28. Distribution of intelligence quotients of 91 girls, ages 7 and 8 combined. 66 STANFORD REVISION OF BINET-SIMON SCALE 56-65 66-75 76-85 86-95 96-/05 i06-li5 ii6-iZ5 i2G-i35 i36-l^3 65% 207, ^I'Jo Z^°/o S7o /'/. Graph 29. Distribution of intelligence quotients of 92 boys, ages 9 and 10 combined. 56-65 66-75 76-85 86-S5 38-/05 /06-//5 /I6-/25 /26-/35 /56-/4-5 ^% 857. /9.5% 3/5;. ZG^ $7. / 7, 2 7, Graph 30. Distribution of intelligence quotients of 108 girls, ages 9 and 10 combined. SEX DIFFERENCES 67 56-65 £6-73 76-85 8S-9S 96-/03 106-/15 llG-125 l2Q-i35 I3Q-I^5 1.5°/. 3% /3'/. 2f/. 2^1. Z0% 81 37, Graph 31. Distribution of intelligence quotients of 74 boys, ages 11 and 12 combined. 5G-G5 Q6-75 TGS^ 88-95 96-/05 106-1/5 116-/25 126-12)5 /3Q-/^5 Zy. 9V, 171 32% 2^7. /2 5% Z'A /% Graph 32. Distribution of intelligence quotients of 88 girls, ages 11 and 12 combiaed. 68 STANFORD REVISION OF BINET-SIMON SCALE Se -G5 66' 7S 76 SS 36 -95 96- 105 106 -115 1/6 -125 J 26 ■ 135 /36 -/4-5 2y, 6"/. /fV. ZO'I, 32'/. ZZ'/, 5'/. Graph 33. Distribution of intelligence quotients of 106 boys, ages 13 and 14 combined. X 56-G5 &e-75 76-86 86-9S 96-/05 /06-//S //6-I25 /Z6-/35I5G-I^5 5.5'/. 87. 271 35'/. 17.51. S.51. 1.5'/. Graph 34. Distribution of intelligence quotients of 74 girls, ages 13 and 14 combined. SEX DIFFERENCES 69 .S6 -es 66 - 75 76 -8S 86 -95' 36 -/OS 106 -IIS 116-125 l26'/3S/36'f4'S .i3% Z.t7t I0.£7'/. 2/J57, J3P57. ZZPl S.OGl 2J7t Gbaph 35. Distribution of intelligence'^quotients of 457 boys, 5-14 years of age. ^6-65 66-75 76-85 86-95 96-105 /06-1 15 II6-/Z5 IZ6-J35 J36-/^S .S2l 2077. 6f7'l. (8B71 JJ.7Z 2f.33l lOMl ZfS'L /J/7 Graph 36. Distribution of intelligence quotients of 448 girls, 6-14 years of age. 70 STANFOKD KEVISION OF BINET-SIMON SCALE TABLE 11 Distribution of Certain Intelligence Quotients in 457 Boys and 448 Girls Total 60 or lower Total 70 or lower Total 75 or lower Total 125 or higher Total 130 or higher Total 140 or higher Boys... Girls. . . No. 1 1 % .21% .22% No. 4 5 % .87% 1.11% No. 13 14 % 2.83% 3.12% No. 11 17 % 2.39% 3.47% No. 5 8 % 1.08% 1.74% No. 3 % .66% The facts we have presented indicate that, apart from a sUght superiority of the girls from 5 to 12 years, the distribution of intelligence is much the same for the sexes. There is no evidence of any wider range of intelligence among boys, such as has commonly been supposed to exist. The difference, if any exists, seems to be in the other direction. A slightly larger percent of girls than of boys falls to 75 or below, which is the point frequently taken as indicating feeble- mindedness, and a decidedly larger percent of girls reaches as high as 125. The range that includes the middle 50 percent is almost exactly the same in extent for the two sexes. This is all quite contrary to the traditional belief that both feeble-mindedness and exceptionally superior ability are more frequent with boys than with girls. ^ Although the superiority of the girls is not great in amount, it appears over a long enough period to suggest that it may represent a genuine difference and not some accidental condition of the experiment. ^ In support of our results we are glad to cite the study of Mrs. Leta Hollingworth : ''The frequency of amentia as related to sex," The Medical Record, Oct. 25, 1913, which is an analysis of 1000 cases ex- amined in the New York Clearing House for Mental Defectives. See also, by the same author: ''Variabihty as related to sex differences in achievement, Am. J. Sociology, Jan., 1914, pp. 510-530; and the comparative variability of the sexes at birth, same Journal, Nov., 1914, pp. 335-370. SEX DIFFERENCES 71 It was first thought that part of the difference might be accounted for by the fact that two-thirds of the tests were made by women and only one-third by men, but when the inteUigence quotients were classi- fied according to the sex of the examiners no such influence was discovera,ble. There remains the possi- bility that the superiority of the girls in the tests may be the result of a somewhat more ready facility of the girls in the use of language, or of their greater willingness to respond. Fortunately the supplementary information fur- nished by the teachers affords us valuable data as to the genuineness of the sex difference in intelligence quotients. On page 48 was given the grouping of 476 children, boys and girls together, according to tke teachers' estimates of intelligence. When these estimates are summarized for the boys and girls sep- arately (Table 12), the superiority of girls appears only in one respect: viz., 12, or 4.8 percent of the girls are classified in the ^Very superior'' group, as con- trasted with 3, or 1.3 percent of the boys. TABLE 12 Teachees' Estimates of Intelligence for 229 Boys and 247 Girls, BY Sex Percent Very Very- Ranked Inferior Inferior Average Superior Superior 229 Boys 3.9 14.4 56.7 23.5 1.3 247Girls 3.2 13.7 57.0 21.0 4.8 Table 12 is for the ages 5 to 14 combined. The teachers' estimates for the ages 13 and 14 were treated separately from the other ages in order to find out whether the tendency of the girls to lose their advan- tage in intelUgence quotient at this point is confirmed or contradicted by the judgment of the teachers. The results, shown in Table 13, bear out the tests, 72 STANFORD REVISION OF BINET-SIMON SCALE for while the teachers have judged the inteUigence of the girls at earlier ages as fully equal, if not superior to that of the boys, they give the advantage at age 13-14 to the boys. The table shows 19.5 percent of the boys of 13 and 14 years of age classed as below average, as contrasted with 24.5 percent of the girls; on the other hand, 26 percent of the boys are classed as above average, as contrasted with 13.5 percent of the girls. TABLE 13 Teachers' Estimates of Intelligence of Boys and Girls for the Ages 5-12 and 13-14 Percent Very Very- Ranked Inferior Inferior Average Superior Superior Ages 5-12 Boys 3.2 14.7 57.3 22.9 1.6 Combined Girls 3.8 11.9 56.1 22.3 5.7 Ages 13- Boys 6.5 13.0 54.3 26.0 0.0 14 Combined.. Girls 0.0 24.5 62.0 13.5 0.0 In like manner we have compared the boys and girls with reference to the quality of their school work, as judged by the teachers. When we classify the judgments for the ages 5-12 separately from those for 13-14, we have Table 14. TABLE 14 Teachers' Estimates of the Quality of the School Work of Boys and Girls for the Ages 5-12 and 13-14 Percent Very Very Kanked Inferior Inferior Average Superior Superior Ages Boys 3.8 11.1 60.0 23.8 1.1 5-12 Girls 3.3 12.6 57.7 20.8 5.3 Ages Boys 4.1 26.5 45.0 22.5 2.0 13-14 Girls 2.4 19.5 53.6 22.0 2.4 Table 14 agrees with the tests in showing a larger number of cases of greatly superior ability in school work among girls than among boys from 5 to 12 years. The data for Ages 13-14 agree with the tests less closely, for, while the boys are less inferior to the SEX DIFFERENCES 73 girls than in the ages 5-12, they are still somewhat inferior. There remains still another means of checking up the evidence of the tests as to the relative intelligence of boys and girls; we can compare their age-grade distribution in school. Fortunately, our data enable us to do this for all the children tested. Table 15 gives the age-grade distribution for Ages 7 to 14. Ages 5 and 6 are left out of account because children of this age have not had time to become retarded or accelerated in school, and those above 14 are eliminated because they represent a selected group — the '^ left- overs. '^ TABLE 15 Age-Grade Distribution of Boys and of Girls for the Ages 7 to 14. (In Percents) Age Grade I II III IV V VI VII VIII 7 Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls Boys Girls 77.7 76.2 22.3 21.5 2.5 24.0 18.2 2.0 2.5 15.5 14.0 1.9 3.5 25.0 16.4 2.8 2.0 20.0 35.2 35.2 2.7 9.1 27.7 40.5 13.3 39.5 8 22.0 27.0 3.9 5.2 52.0 52.5 9 32.8 26.1 2.8 8.1 5.3 46.2 50.8 10 25.0 22.5 8.4 9.3 13.3 3.4 44.4 51.0 11 31.6 16.4 10.8 9.1 10.3 2.6 2.2 3.0 34.3 37.2 2.3 12 13 14 32.4 32.0 15.5 5.4 2.2 40.6 47.9 38.0 35.0 24.5 24.0 2.5 5.2 16.2 57.6 33.Z 74 STANFORD REVISION OF BINET-SIMON SCALE Inspection of Table 15 will show that the results lack uniformity. At Age 7 , the age-grade status of the girls is slightly better; at 8, the boys have a shade the advantage; at 9, the girls; at 10, the boys; at 11, 12 and 13, the girls are much in advance; while at 14, the boys are the first time decidedly ahead. On the whole, we may say that there is httle difference for Ages 7, 8, 9, and 10; that for the next three years the girls are much more advanced than the boys; and that at 14 the boys have much the advantage. Ignoring Year 7, since at this age pupils have had little time to become retarded or accelerated, we are justified in grouping the ages together as follows: Years 8, 9, and 10; years 11, 12, and 13; and finally, Year 14 by itself. Table 16 shows the percent of boys and girls retarded or accelerated 1, 2, 3, or 4 years in each of these age-groups. TABLE 16 Percent of Boys and Girls Retarded or Accelerated in School BY 1, 2, 3, OR 4 Grades in Various Age-groups Age Sex — i —3 —2 —1 Normal + 1 + 2 +3 + 4 Ages .8, 9, 10 combined Boys. . . Girls. . . 1.4 4.6 27.0 25.3 48.1 51.3 21.1 16.0 2.1 2.6 Ages 11, 12, 13 combined Boys. . . Girls. . . 1.5 10.0 .8 12.3 8.0 34.6 27.5 33.0 42.0 8.5 20.1 1.6 Age 14 Boys. . . Girls. . . 2.2 3.2, 2.2 24.5 24.2 13.3 39.3 57.8 33.3 SEX DIFFEEENCES 75 TABLE 17 Percent of Boys and Girls Retarded or Accelerated in School BY 1, 2, 3, OR 4 Grades, for the Ages 8 to 14 Combined —4 —3 —2 —1 Normal + 1 + 2 + 3 + 4 Boys .95 4.47 9.58 28.11 43.13 12.77 .95 Girls .32 .32 8.11 27.59 45.45 15.90 1.94 .32 Table 17 shows the percent of boys and girls retarded or accelerated 1, 2, 3, or 4 years for all the ages 8-14 combined. It should be emphasized, however, that the facts we want to know are best disclosed in Table 16, and that the evidence goes to show that the grade progress of our hoys and girls differs little up to, and including Age 10, that for the next three years the girls are clearly in- advance, and that the reverse is the case at 1/f.. In the main, therefore, the school progress of our subjects agrees with the intelligence tests, with the teachers' estimates of intelligence, and with the teachers' judgments of the quality of the school work, in showing a sex differ- ence which is in favor of the girls before 14, ctnd in favor of the boys thereafter. Before accepting this conclusion there is one other factor to be taken into account which might help to explain the apparent superiority of the boys at Age 14. A certain amount of selection has taken place in this age-group. A considerable number of the 14-year-olds have been promoted to the high school, and these are not included in our group of subjects of this age. This has doubtless occurred more often with the girls than the boys, for we have already showed that marked school acceleration occurs much oftener with girls than with boys at the ages 12 and 13 years. At Age 76 STANFORD EEVISION OF BINET-SIMON SCALE 12 there are 9.1 percent of the gMs in Grade VII and 2.5 percent in Grade VIII, as compared with only 2.7 percent of the boys in Grade VII and none at all in Grade VIII. Similarly, at Age 13 there are 16.2 percent of the girls in Grade VIII as compared with only 5.2 percent of the boys. If all our 13-year-olds in Grade VIII should be promoted to the high school at the end of the year, the number of 13-year-old girls receiving such promotion would accordingly be more than three times as great as the number of boys. If this situation holds true generally, it can not be safely ignored in making a comparative study of the intelli- gence of boys and girls at the age of 14. Even at 13, unequal selection appears to have taken place to no small degree, as would appear from the following age-grade distribution (in percent s) of 13-year-old boys and girls: TABLE 18 III IV V VI VII VIII 13 Boys.... . 3.4 10.3 15.5 38 27.7 5.2 year olds Girls 2.6 5.4 35 40.5 16.2 The foregoing distribution would suggest that our 13-year-old boys are almost free from any selection as far as pro motion to the high school is concerned, but that probably 5 percent or more of the girls who ought to be present at this age are not in our group. Comparison of the relative number of boys and girls tested at different ages fuUy confirms this suspicion. Table 19 shows that, while the number of girls found in the grades equals or exceeds the number of boys at every age but one below 13, yet at 13 and 14 the girls make up only about 41 percent of the entire number. This is significant when it is remembered that, in all the schools where testing was done, all SEX DIFFERENCES 77 the boys and girls below the high school were tested who were within two months of a birthday, whatever grade they may have been in. By the laws of chance the number of boys and girls found at each age ought to have been nearly equal, barring selective influences. TABLE 19 Percent the Girls Form op the Entire Number of Pupils in Grades Below the High School at Each Age Age 5 6 7 8 9 10 11 12 13 14 15 16 Percent... 50 50 49.5 46 51.3 57.5 56 53 41 41.5 45 28.6 The only possible conclusion seems to he that the apparent superiority of hoys at the age of 14-, as well as also their diminished inferiority at 13, is due solely to the unequal selection which has taken place at these ages. The results of the tests themselves, the teachers' estimates of intelligence, the teachers' judgments about the quality of school work, and the age-grade distribution, offer four separate and distinct lines of evidence pointing in this direction. The same four lines of inquiry are also in general agreement in show- ing a distinct, though slight superiority of the girls in the ages below 13. Unfortunately, most of the studies made with the Binet tests throw little light on sex differences. In only two studies besides our own have the subjects tested been nearly enough non-selected to make such a comparison worth while, namely, in Goddard's tests of 1500 Vineland school children and in Kuhl- mann's tests of 1000 Faribault school children. Kuhl- mann, however, has not analyzed his data for the sexes separately, and Goddard has done so only to the extent of giving the percent of boys and girls testing 1, 2, 3, 4, or more years above or below age. 78 STANFORD REVISION OF BINET-SIMON SCALE Even this is given by Goddard only for all the ages taken together, a procedure which ignores the unequal significance of a given number of years of retardation with children of different ages. However, his data for the ages combined agree with our own in indicating a slight superiority of the girls. TABLE 20 Sex Differences as Shown by the Binet Tests (Goddard) — 2 years — 1 year At age +1 5 ear +2 years Percents or more or more or more Boys 18.5 23 34.5 20 4 Girls 18.5 17 36.5 23 5 Bloch and Preiss made comparisons for sex differ- ences in their tests of 79 boys and 71 girls aged 7 to 11 years. However, their results, which indicated a marked superiority of the boys, throw no light on the question of real sex differences, for the reason that their subjects had been selected from a large number as being supposedly "average'^ in intelUgence, and we have no means of checking up the effect of this selec- tion. The only other Binet results thus far published which may be used for comparative purposes are those of Yerkes and Bridges, who present the sex differences found in the use of their Point Scale with 760 grammar school children from the kindergarten to the eighth grade. The school chosen was located in a ^ ^medium to poor'^ district, but otherwise there seems to have been no selection of subjects which could have influ- enced sex differences except the fact that in this study, as in our own, the 14 and 15-year-olds tested were composed wholly of those who had not progressed in school beyond the eighth grade. To make comparison easier, Yerkes' and Bridges' curves showing the average scores earned by boys SEX DIFFEEENCES 79 and girls of different ages are reproduced here. We have chosen the curves obtained by excluding the children of non-English-speaking parents. When the latter were included, the curves crossed and re-crossed so often as to have no clear significance. It will be noted that the results of Yerkes and Bridges are not altogether in harmony with our own. The two scales agree in that both show a superiority of the girls in the earlier ages and a superiority of the boys at Ages 14-15. The latter, in all probability, has the same cause in both cases, namely, the more frequent elimination of 13, 14, and 15-year-old girls from the grades by promotion. Yerkes and Bridges, however, seem not to have considered this possibility. In the middle ages the results of the two studies are quite in contrast. In evaluating these somewhat contradictory find- ings it should be remembered, (1) that the Stanford data are based on more than twice as many children as those entering into the Yerkes-Bridges curve; (2) that all of our children were within two months of a birthday, thus obviating large possible errors likely to result from dividing a small number of children with age differences ranging up to one year into sex groups; and (3) that the number of tests in the Stan- ford revision is much larger than that included in the Yerkes-Bridges scale, thus reducing the part played by chance. In conclusion, we may say that the evidence seems to us to point to the existence of a small sex difference in intelligence, which, but for the influence of selection, would probably be in favor of the girls at all ages from 5 to 13 or 14 at least. It should be emphasized, how- ever, that the difference is small, amounting to no 80 STANFORD REVISION OF BINET-SIMON SCALE JO y , / / 80 ; K / / / 1 / < r" IL 1 / 1 / 70 {/ ^A {l { -^ 1 / ^ ^ ► bO / \ r / / / / / / ( c fi SO II 1 1 1 1 1 1 y w fO / / I ^ 1 /- i / / / / / / / y > / X 30 , , / / i y / / / / /' / i > / / / / / zo < ' . / in 4- S 6 7 8 9 JO // /Z 13 14- /S- Average Scores Earned by 468 Boys and Girls of Different Ages with the Yerkes-Bridges Point Scale SEX DIFFERENCES 81 more than 6 percent in terms of intelligence quotient at any age, and at most ages from only 2 percent to 4 percent. In view of the wide distribution of intelli- gence in each sex (from 50 to 150 intelligence quotient), a difference of 2-4 percent in median intelligence would be practically negligible, even if it were dem- onstrably genuine. The difference actually found is so small that it might conceivably result from a sex difference in temperamental traits having nothing to do with intelligence, such, for example, as a differ- ence in willingness to attempt a novel task, difference in timidity, or what not. We prefer not to indulge in speculation. At any rate we find no reason to share the opinion voiced by Yerkes and Bridges ^^that at certain ages serious injustice will be done to individ- uals by evaluating their scores in the light of norms which do not take account of sex differences." Finally, the individual tests were examined sep- arately for sex differences. Since the number of our pupils of one sex was ordinarily not larger than 40 to 50 at one age, it was found that the increase in the percent passing in successive years was so irregular as to be very confusing. One way out of this difficulty is to mass together the percents of boys (or girls) passing a test at three separate age levels: the age at which the test appears in the scale, and the adjoining ages above and below. We have not deemed a sex difference worth noting unless it appeared in all of these three successive age levels and to such an extent that the superiority averaged 10 percent for the three ages taken together. This is, of course, an arbitrary basis, but some such plan is necessary to escape the confusion and contradiction engendered by chance variations due to small numbers. As will be seen 82 STANFORD REVISION OF BINET-SIMON SCALE from Table 21, the number of tests in which significant sex differences appear, is not large — orly 19 out of 58.» TABLE 21 Sex Differences in Individual Tests SuPEiiioEiTY OF Boys Superiority of Girls Tests ia which the super- iority of Boys in three successive years averaged 10 percent or more Amoiint of such superiority in percent Tests in which the super- iority of GmLS in three successive years averaged 10 percent or more Amount of such superiority ia percent Arithmetical reasoning, XIV President and king, XIV Form-board, IX Fables, XII Making change, IX Hands of clock, XIV BaU and field, XII Similarities, XII [nduction, XIV 33 25 20 15 15 12 11 11 10 Designs from memory, X Aesthetic compr., V Ball and field, VIII Giving differences, VII Comprehension, VIII 4 digits backwards, IX 6 digits, X 7 digits, XIV Bow-knot, VII Rhymes, IX 19 17 16 13 12 11 10 10 10 10 Smaller differences were found in favor of the boys in Copying a Diamond (VII), Giving Easy Similarities (VIII), Naming the Months (VIII), and Defining Abstract Words (XII) ; in favor of the girls in Repeat- ing 5 Digits (VII), Naming the Days of the Week (VII) and Hard Comprehension (XII). We refrain from extended comment on the list of tests in which sex differences have been found. It will probably agree badly enough with anyone's a priori expectations. There are apparent contra- dictions, also, as well as surprises. We have no theory 2 The tests of Years 3 and 4, and those of the "average adult" and "superior adult" were left out of this comparison because of insufficient data. SEX DIFFERENCES 83 to explain why the girls are superior on the ball and field test of Year VIII (score 2) and the boys on the same test at Year XII (score 3); or why the boys are better in the tests of giving similarities, and the girls in the test of giving differences. Perhaps we should have expected the superiority of the boys in arithmetical reasoning, the form-board, and making change; likewise the superiority of the girls in aesthetic comparison, tying a bow-knot, and repeating digits. Summary 1. The tests indicate a slight superiority of girls over boys at each age from 5 to 13. The apparent superiority of the boys at 14, however, is probably accounted for by the unequal selection which has taken place in the promotion of pupils to the high school. 2. The small superiority of the girls in the tests probably rests upon a real superiority in intelligence, age for age. At least, this conclusion is supported by the age-grade distribution of the sexes, and by the teachers^ rankings according to intelligence and quality of school work. 3. Apart from the small superiority of the girls, the distribution of intelligence shows no significant difference in the sexes. The data offer no support to the wide-spread belief that girls group themselves more closely about the median or that extremes of intelligence are more common among boys. 4. Not many of the individual tests show large sex differences in the percent passing in three con- secutive years. In certain of the tests, however, the differences were marked and unexpected. CHAPTER V THE RELATION OF INTELLIGENCE TO SOCIAL STATUS In the use of the Binet scale with different social classes it has generally been found that children who come from superior environment test higher than those who come from homes where the degree of culture is inferior. As already noted, the arrangement of tests in the 1908 Binet scale was based on an exami- nation of about 200 children from one of the poorest quarters of Paris. WTien the scale thus derived was used by Decroly and Degand in the examination of children from wealthy and cultured homes, in a Brus- sels private school, it was found that many of the tests were passed two years below the location assigned them by Binet. Jeronutti's tests of 144 better-class children of Rome agree closely with those of Decroly and Degand. The same is true of Madame Wolk- owitsch's tests of private-school children at Petrograd. On the other hand. Dr. Anna Schubert's data gathered from 229 lower-class children in a Moscow orphanage gave a distribution of mental ages skewed in the opposite direction. In fact, none of Dr. Schubert's children were advanced more than one year, only 27 percent tested '^at age," while 75 percent were retarded one j^ear or more and 39 percent two years or more. Binet, himself, took up the question by having tests made of 54 children who had been classified into four groups by the teachers according to social status. Unfortunately, the school chosen was one which was 84 RELATION OF INTELLIGENCE TO SOCIAL STATUS 85 not attended by children of the highest social classes, and the number tested was very small. The results failed to show any correlation of mental age with social status. A later comparison by Binet of 30 children attending a poorly situated school with 30 others attending a school in a well-to-do neighborhood of Paris showed a marked difference in favor of the better situated children. The Breslau experiment, of which a partial account has been given by Stern, ^ indicated that pupils of the Volksschule are at the age of 10 years somewhere near the level of mental development which is at- tained by pupils of the Vorschule at 9 years. The Volksschulen are attended mainly by children of the laboring and lower business classes and the Vorschulen by children of the better classes. Likewise, in com- paring children of the upper and lower social classes in English infant schools, Winch found the children of the higher class superior to the other group in a majority of the tests. Study of the data which we have collated for the individual tests of the scale shows that large differ- ences found by investigators in the percentage of children who pass certain tests may often be accounted for by a difference in the social class of the subjects. Yerkes and Bridges compared 54 pupils of a better- class school with an equal number attending a poor school. The sexes were represented equally, and the pupils were selected in such a way that a boy or girl of the favored group was matched by a boy or girl of approximately the same age from the unfavored group. The comparison showed that by the Yerkes- ^ The Psychological Methods of Testing Intelligence, These Mono- graphs, No. 13, 1914, pp. 54 fiF. 86 STANFORD REVISION OF BINET-SIMON SCALE Bridges scale the favored boys averaged 7.7 points higher and the girls 8.4 points higher than did the members of the unfavored group of the same ages. If we compare this difference in points with the age- norms given by Yerkes and Bridges, we find that it represents about a year of difference in mental ad- vancement with children of this age. Binet at one time estimated that social status might make as much as a year and a-half difference in mental age, though in making this statement he seems to have overlooked the fact that a retardation of a year and a-half is not of equal significance in the lower and upper ages. The tacit assumption which most writers seem to have made in their discussions of such facts as those we have just set forth is that the difference found is due wholly, or at least mainly, to the influence of environment. Meumann believes that the most seri- ous fault of the Binet scale is its failure to take account of the influence of social environment on the ability to pass certain tests. Yerkes and Bridges assert that ''it is obviously unfair to judge by the same norm of intelligence two children, the one of whom comes from an excellent home and neighborhood, the other from a medium-to-poor home and neighborhood.'' As will be shown presently, we believe that the facts may be more reasonably explained on an entirely different h^^^othesis. First, however, we will set forth somewhat in detail the Stanford data that bear upon this question. As stated elsewhere, we were able to secure a classi- fication of 492 of our children according to social class into five groups: 'Very inferior," "inferior," "average," "superior" and very superior." Although the schools chosen for the tests were on the whole RELATION OF INTELLIGENCE TO SOCIAL STATUS 87 as nearly average as could be found, it will be readily understood by anyone who is acquainted with the democracy of the American educational system that in almost any small city an ' 'average" school contains some children of every social class. As was expected, therefore, all the social classes were found to be rep- resented in every school. Graph 38 shows the dis- tribution by social class of the 492 children regarding whom the supplementary information was obtained. l/fir/ Inferior /riferior Ai/trac/e Superior \/ery Superior 27. IBBI 5G.51 20BV. I^y. Graph 38. The Distribution by Social Class of 492 Children OF All Ages. We have next classified these same pupils according to the intelligence quotients resulting from the tests, and for this purpose also we have made use of five groups, as follows: I Q 120 or 1 Q below 80 I Q 80-89 I Q 90-109 I Q 110-119 above "Very infe- "Very rior" "Inferior" "Average" "Superior" superior" 88 STANFORD REVISION OF BINET-SIMON SCALE This grouping of the inteUigence quotients is, of course, arbitrary, but some sort of grouping is neces- sary, and the one we have employed has the advantage of giving five groups of inteUigence quotients which agree roughly in size with the groups according to social class. This correspondence is as follows: TABLE 22 Percent op Pupils in Each Group I Q I Q below 80 1 Q 80-89 I Q 90-109 I Q 1 10-119 I Q 120 or above 6.9 14.5 57.7 15 5.6 Social Very Very Status inferior Inferior Average Superior superior 2 16.6 56.5 20.6 1.6 Table 23 shows where the children in each social group fall with reference to intelligence quotient, and also where those of any intelligence quotient group fall with reference to social class. TABLE 23 The Relation of Intelligence to Social Status I Q Social 120 or Status Below 80 81-89 90-109 110-119 above Total Very Inferior 4 4 3 11 Inferior 18 15 43 4 102 Average 9 43 181 48 11 292 Superior 1 7 59 21 14 80 Very Superior 3 2 2 7 Total 32 69 289 75 27 492 RELATION OF INTELLIGENCE TO SOCIAL STATUS 89 That there is a certain degree of correlation between intelligence quotient and social status is quite evident from a glance at the table. Of the 27 children with an intelligence quotient of 120 or above, not a single one comes from a social class below the average; of the 75 with an intelligence quotient between 110 and 119, only 4 belong to the "inferior'' social group, and none to the "very inferior group." Conversely, of the 32 children with an intelligence quotient below 80, only one is classified below the "average" social group. Only the middle intelligence quotient group, 90-109, is represented in all the social classes; and only the "average" and "superior" social group is represented in all the intelligence quotient groups. Application of the Pearson formula to the data in this table gives a correlation of .40 between social status and intelli- gence quotient. Another way to express the relationship between intelligence and social status is to compare the median intelhgence quotient for the children of each social group, as follows: Social Group Very Inferior Inferior Average Superior Very Superior Median I Q 85 93 99.5 107 106 Since only 8 pupils are included in the "very in- ferior" and only 10 in the "very superior" social groups, the medians for these extreme groups have limited significance. The case is different, however, in the "inferior," "average" and "superior" groups, which include 80, 292, and 102 cases, respectively. Graphs 39, 40, and 41 show the distribution of intelligence quotients grouped by lO's for the three social classes. In this case the 7 "very inferior" pupils are thrown with the "inferior" class, and the 11 "very superior" pupils with the "superior" class. 90 STANFORD REVISION OF BINET-SIMON SCALE SG - 65 66 - 7S 7G-7S 76 'S5 SG-SS 36-/05 /0G-//5 I (6-IZ5 IZ6-l35/36'l4\ r/o M. Zt6lo'37% Z0% 8.1 fc .fl &J. \tO,Ui^Tri'buT'iOti of I Q's of ZSS" chi/dren of " a-verajc'' socio/ clo&'i S6' 65 66-75 76-85 36-35" $6-/05 /06-//S //G 'IZ5/Z6-I35 136-/4 f'/p l^% 18.6% Z2J% 30,3% 81% Z3% ^^j, Oi^'Tri buTi'on of I Qs oj 86 ch ( Idran of'iTifcr ('or a-nd ''\/trj^ inf trior* social classes. RELATION OF INTELLIGENCE TO SOCIAL STATUS 91 The median intelligence quotient of the ^'average'' social group is 99.5, and so practically coincides with the median for all classes taken together. It is sig- nificant that the median intelligence quotient of the ^^inferior'' group is 14 points below that of the '^superior" group, and that the median intelligence quotient for the ^^average^' group lies approximately mid-way between the two. The difference between the median intelli- gence quotients of the ''inferior'^ and the '^superior" groups means a difference in mental age at the various actual ages as follows: TABLE 24 Median Diffekence in Mental Age Age Age 4 6.7 mo. 10 16.8 mo 5 8.4 " 11 18.5 " 6 10.1 " 12 20.2 " 7 11.7 " 13 21.8 " 8 13.4 " 14 23.5 " 9 15.1 " 15 25.2 " Our results, accordingly, agree closely with those of other workers. Our next task is to find the most rational hypothesis which will explain the correlation between social status and intelligence quotient. The usual assumption has been that the correlation is the artificial product of environmental influences; that the child from a cultured home does better in the tests by reason of his superior home training and because he has had more opportunity to pick up the informa- tion which success in the various tests calls for. This explanation has seemed to us from the beginning a most improbable one. Several investigations of the influence of environment on mental traits suggest the conclusion that this influence is much less important than is original endowment in determining the nature 92 STANFORD REVISION OF BINET-SIMON SCALE of the traits in question. From an a-priori stand- point, the endowment hypothesis explains the correla- tion between intelligence quotient and social status just as adequately as does the environment hypothesis. To conclude, as Meumann and Yerkes have done, that the demonstration of the existence of such a correlation invahdates the Binet scale as a method of measuring intelhgence is to make a gratuitous assumption — an assumption, indeed, which is con- tradicted by much evidence from investigations bearing on the mental endowment of different social groups. We have thought it worth while, therefore, to sift our data somewhat carefully for evidence on this point. First, we have compared the social status of the children with the teachers' estimates of intelligence. The tests themselves are brief. Success in some of them, it must be admitted, hinges upon information, the possession of which might conceivably be largely conditioned by home environment. One thinks in this connection of such tests as naming coins, making change, repeating the days of the week and the month of the year, giving definitions, giving the moral of fables, etc. Success in certain others would appear to depend rather too much on facility in the use of language. But the teacher's judgment as to a child's intelligence is based upon months of acquaintance, in this case from half to an entire school year. The teacher has had abundant opportunity to distinguish between real mental ability on the one hand and the accidents of knowledge, or facility in the use of langu- age, on the other. Accordingly, Table 25, which gives the teacher's judgment as to the intelligence of each child in the various social groups, should be of interest. RELATION OF INTELLIGENCE TO SOCIAL STATUS 93 TABLE 25 The Intelligence op Children op Various Social Classes as Estimated by the Teachers Teachers' Estimate of Intelligence Social Very- Very- Status Inferior Inferior Average Superior Superior Total Very Inferior 6 2 3 11 Inferior 4 31 36 5 1 77 Average 7 34 93 50 4 188 Superior 1 40 56 4 101 Very- Superior 2 6 8 Total 18 67 172 113 15 385 Casual inspection of this table shows that the judg- ment of the teachers accords with the evidence from the tests in crediting greater mental ability to the children of superior social status. Not one of the children of the 'Very superior^' social group is ranked below '^superior" intelligence, and of the 101 included in the "superior'' social group only one falls below "average" intelligence. Conversely, not one of the 11 children of "very inferior" social status ranks above "average" in intelligence, while 6 of them are classified as intellectually "very inferior." By the Pearson formula the correlation between social status and the teachers' estimates of intelligence is .55. This is considerably higher than the .40 correlation found between social status and intelligence quotient. But children from superior homes are likely to be better dressed, cleaner and more attractive in appear- ance than children from the poorer homes. Perhaps, 94 STANFORD REVISION OF BINET-SIMON SCALE too, they are better behaved, more responsive, and socially more adaptable on account of superior training in the home. It is conceivable that external appear- ances of this kind, which, all would agree, are in part an expression of home conditions, have deceived the teachers and influenced their ranking of the children according to intelligence. If this were true, the actual quality of the school work done by the children of various social groups might be expected to afford a corrective for this possible error. School work, because it is more definite and objective, is easier to judge than the complex of mental traits called intelligence. Table 26 shows the quality of the school work, as judged by the teachers, for all the children of the several social groups. TABLE 26 The Quality of School Work Done by Children of the Various Social Groups Quality of School Work Social Status Very Inferior Inferior Average Superior Very- Superior Total Very Inferior 5 4 3 12 Inferior 10 29 35 6 2 82 Average 9 52 160 60 4 285 Superior 4 51 46 4 105 Very- Superior 2 6 8 Total 24 89 249 114 16 492 By the Pearson formula the correlation expressed in this table amounts to .47, which is only a little lower than that found for the teachers' estimates and EELATION OF INTELLIGENCE TO SOCIAL STATUS 95 social status, and somewhat higher than that between social status and intelligence quotient. We would attach especial importance to this correla- tion, for the reason that the wide-spread use of peda- gogical tests in recent years has demonstrated that the individual differences in subject-proficiency which such tests bring to light among school children repre- sent, in large part, individual differences in native endowment and not the effects of unequal home or school training. Even spelling abihty, contrary to common opinion, is largely a function of general intelli- gence. As Mr. Houser has shown, the correlation between the two ranges from 35 to 71 percent. 2 In- dividual instruction, ^^ special class' ^ methods, indeed, the concentrated efforts of an entire school system, are unable to wipe out the major differences of this kind in a dozen years. As a rule, the longer the child is in school, the more evident the inferiority of the inferior child becomes. It would hardly be reason- able, therefore, to expect that a little incidental experi- ence and instruction in the home, amounting perhaps in most cases to not more than a few minutes per day, would weigh very heavily against these native differ- ences. Even in good homes, children are likely to learn less from their parents than from their play fellows and nurses. For further evidence on the relation of school success to social status, we will examine the age-grade dis- tribution of the children in the five social groups. This is shown in Table 27 for the ages 8-16. Children below 8 were omitted from this comparison because 2 See J. D. Houser: ''The relation of ability to general intelligence and to meaning vocabulary. The Elementary School Journal, Dec, 1915. 96 STANFORD REVISION OF BINET-SIMON SCALE TABLE 27 The Relation of Age-grade Distribution to Social Status Location in the Grades Social Status Retarded In Advanced Grade 4 yr. 3yr. 2 yr. 1 3-r. for Age 1 yr. 2 yr. 3 yr. Total Very Inferior 1 1 2 2 2 8 Inferior 3 4 10 27 11 4 59 Average 1 4 21 66 109 21 2 224 Superior 4 16 30 14 64 Very Superior 3 2 1 6 Total 5 9 2,1 114 154 40 2 321 they have not had time to become much retarded or accelerated in school. Normal progress is here defined as Grade II at Age 8, Grade III at Age 9, etc. The correlation here is positive, but somewhat smaller than in the case of the intelligence quotient, the teachers' estimates of intelligence, and the quality of the school work. This is what we should expect, knowing that the tendency of schools is to promote children by age rather than by quality of school work or native abihty. Again, if home environment really has any con- siderable effect upon the intelligence quotient we should expect this effect to become more marked, the longer the influence has continued. That is, the correlation of intelligence quotient with social status should increase with age. We have accordingly RELATION OF INTELLIGENCE TO SOCIAL STATUS 97 worked out the correlation between intelligence quo- tient and social status for three separate age-levels: for Years 5, 6, 7 and 8 taken together; for 9, 10, and 11 combined; and for 12, 13, 14, and 15 likewise combined. The coefficients of correlation for these three age- levels were, respectively, .43, .41, and .29. In other words, the longer the supposed powerful influence of home environment is continued, the more independent of it the intelligence quotient becomes. The con- clusion indicated is that the home environment, as environment^ has in all ages of childhood relatively little weight in determining the intelligence quotient. One other line of argimaent remains. Anyone who has done much testing knows that if sufficiently large numbers are taken, every degree of intelligence from profound Idiocy to very superior ability is represented in every social class. This did not happen to be true in the case of our 500 non-selected children from whom supplementary data were available, but we may be certain that it would have been true if a hundred or a thousand times as many children had been tested. In miscellaneous testing, the data from which are not included in the present study of non-selected children, we have found two children of extraordinarily poor home environment who had an intelligence quotient of 150. The highest we have ever found among children of any class is 170. It is a coramon- place that dull and feeble-minded children of all grades of deficiency may be found in any social class. We have tested two feeble-minded children whose fathers were men of substantial reputation for scientific achievement. It goes without saying that in each case the home environment was everything that could have been desired. In each family there are other 98 STANFORD REVISION OF BINET-SIMON SCALE children whose intelHgence quotients range from 115 to 125. Three children were tested in another family, in which the home conditions were about as wretched as could be imagined. Two of these children had an intelligence quotient between 75 and 85, the third an intelligence quotient of 120. The two former have since shown their inability to do fifth-grade school work by the age of 15 years, while the latter entered the high school at the age of 12. Since the unfavorable home environment did not prevent the superior endow- ment of the one child from evidencing itself in the tests, we must conclude that the inferior showing of the other two could have not been caused by this same environment. These are individual cases, and we would not stress them unduly. They do illustrate, however, a most important fact, — that exceptionally superior endow- ment is discovered by the tests, however unfavorable the home from which it comes, and that inferior men- tality can not be overcome by all the advantages of the most cultured home. Of course, we would not deny all possibility of environmental conditions affecting the result of an intelligence test. On the contrary, we have no doubt that the influence, is always present in some degree. What it accounts for in terms of the intelligence quotient, we do not know. That it accounts for the larger and more significant differences seems to us wholly improbable. We have little reason to believe that ordinary differences in social environment (apart from heredity) — differences such as those obtaining between the higher and lower classes of children attending approximately the same general type of school in a civilized community — impair the validity of the scale to any great extent. RELATION OF INTELLIGENCE TO SOCIAL STATUS 99 A crucial experiment would be to take a large number of young children of the lower classes and, after placing them in the most favorable environment obtainable, to compare their later mental development with that of children born into the best homes. No such study, properly safe-guarded, seems to have been made. The study would be quite feasible if carried out with the cooperation of a well-conducted orphanage. Some of the tests which have been made in such institutions indicate that mental subnormality of both high and moderate grades is extremely frequent among children who are placed in these homes. Most, though ad- mittedly not all of them, are children of inferior social classes. Of 20 orphanage children tested by the writer only 3 were fully normal. The other 17 ranged in intelligence quotient from 75 to 95. Nearly all of these children had been in the orphanage for from two to several years. The orphanage in question is a reasonably good one and affords an environment which is about as stimulating to normal mental development as average home life among the middle classes. The children live in the orphanage and attend an excellent public school in a California village. ^ After all, does not common observation teach us that, in the main, native qualities of intellect and character, rather than chance, determine the social class to which a family belongs? From what is already known about heredity should we not naturally expect to find the children of well-to-do, cultured, and success- ful parents better endowed than the children who have been reared in slums and poverty? An affirmative 3 Additional data will be published shortly on the influence of orphan- age life on the intelligence quotient of children who have come from 1 ow-grade homes. 100 STANFOED REVISION OF BINET-SIMON SCALE answer to the above question is suggested by nearly- all the available scientific evidence. The suggestion urged by Meumann and also by Yerkes, that it is unfair to evaluate the intelligence of any child except in terms of the average intelligence of his own social class, is not warranted. It would be just as logical to insist that it is unfair to the dull or feeble-minded child to judge his intelligence with reference to standard intelUgence for the mentally normal. Finally, it should be pointed out that Meumann's strictmres on the Binet scale in this connection had their origin in certain discrepancies observed in the results of various investigators, which seemed to him attributable entirely to differences in the social status of the subjects tested. It can be shown, however, that the observed discrepancies may be largely ac- counted for in other ways. They may have resulted in part from failure to follow the same procedure in giving or scoring the tests. Experience in training a fairly large number of individuals in the correct use of the Binet method leads one to stress this factor as a possible source of large discrepancies in results. Moreover, closer inspection of the discrepancies shows that they are much smaller than Meumann seems to have estimated them. As a criterion of agreement between two workers he uses the relative percentages of children found to test '^at age," +1, +2, — 1, — 2, etc. In certain cases, this criterion has been mis- leading. It is well known that the original Binet scale was much too easy at the lower end and much too difficult at the upper. Accordingly, whether the mental ages found by a given worker turn out to be predominantly plus or predominantly minus depends largely on the age of the subjects. If these are yoimg, RELATION OF INTELLIGENCE TO SOCIAL STATUS 101 the mental ages, by the original Binet scale, will tend to run too high. If they have reached an age which demands the use of the upper tests, the resulting mental ages will be too low. Meumann instances the high mental ages found by Jeronutti in his tests of better-class children in Rome as evidence purely of the influence of milieu. However, an examination of Jeronutti' s table of results by ages reveals the fact that the excess of plus mental ages is present only in the years below 10. Above 10, the mental ages fall predominantly on the minus side. Meumann seems to have been similarly misled with reference to the high mental ages found by Madame Wolko- witsch at Petrograd. Since her children were of kindergarten age, we may be sure that their high mental ages resulted in part from the incorrectness of the scale at this point and only in part from the high social status of the children. On the other hand, Meumann instances the excess of minus mental ages found in Miss Johnston's tests as an example of the unfavorable influence of low social status on test performance. The fact is that Miss Johnston's excess of minus mental ages occurs most noticeably in the years above 9, where the scale is demonstrably too hard. Of her 7-year-olds, only 5 tested minus and 24 at or above age. We would agree with Stern that the remarkable fact is not that minor discrepancies have appeared in the statistics of different workers, but that their results, gathered in France, England, Germany, Russia, Italy, Belgium and diverse parts of the United States, from different classes of children and by methods which have undoubtedly fallen short of the desired uniformity, should agree as closely as they do. 102 STANFOED REVISION OF BINET-SIMON SCALE It is quite possible that some of the individual tests of the Binet scale are affected by accidents of environ- ment and training to an extent which largely invali- dates them as measures of intelligence. We do not know which tests are included in this class, but re- search will ultimately disclose their identity. Kuhl- mann has shown how unsafe it is to condemn a test off-hand as subject to this or that disturbing influence.* The classification of the tests in Meumann's three-fold test series (tests of maturity, tests of endowment, and tests of milieu) is based upon inspection, and has little value beyond the program of research which it sug- gests.^ To ascertain the extent to which a test is influenced, by environment, apart from endowment, is not easy. An attempt was made to analyze the Stanford data for evidence of this influence on the individual tests, but it was abandoned. Of the 80 to 120 children at each age from 10 to 15 were ordinarily found in each of the two social classes ^'inferior" and ^ ^superior.'' Such numbers are at best too small to have statistical value, and in this case the matter was further com- plicated by the presence of large differences due presumably to native endowment. To all appear- ances, the average child of any social class behaved in the tests like a child of any other class who had the same intelligence level. * F. Kuhlmann: The Binet-Simon tests in grading feeble-minded children, J, of Psycho- Asthenics, XVI, 1912, pp. 173-193. ^ See Terman's review of Meumann on the Psychology of Endow- ment, J. of Psych.-Asthenics, 14: 1914-1915: pp. 75-94: 123-134: and 187-199. RELATION OF INTELLIGENCE TO SOCIAL STATUS 103 Summary 1. The median intelligence quotient for children of the superior social class is about 7 points above, and that of the inferior social class about 7 points below the median intelligence quotient of the average social group. This means that by the age of 14, inferior-class children are about one year below, and superior-class children about one year above, the median mental age for all classes taken together. 2. That the children of the superior social classes do better in the tests is almost certainly due primarily to superior original endowment. This conclusion is supported by five supplementary lines of evidence: (a) the teachers' rankings of the children according to intelligence, (6) the age-grade progress of the chil- dren, (c) the quality of the school work, (c?) the com- parison of older and younger children as regards the influence of social environment, (e) the study of individual cases of bright and dull children in the same family. 3. In order to facilitate comparison, it is advisable to express the intelligence of children of all social classes in terms of the same objective scale of intelli- gence. This scale should be based on the median for all classes taken together. 4. Meumann's criticism of the scale with reference to the influence of social environment on the test results was based on insufficient examination of the data, particularly on the failure to take account of the prevailing ages of the children tested in different investigations. 5. In their responses to individual tests, our children of a given social class were not distinguishable from children of the same intelligence level in any other social class. CHAPTER VI The Relation of Intelligence to School Success The degree of school success offers a partial check on the accuracy of the intelligence scale. While we should not expect complete agreement between scholar- ship and the results of even a perfect test of intelligence, nevertheless a very marked disagreement between the two would suggest some fault either in the tests or in the method of estimating school success. There are three main indices of the degree of school success which a given child has attained: (1) his advancement in the school grades, (2) the quality of school work he is doing in the grade where he is enrolled, and (3) the extent to which he is regarded by the teacher as intelligent. At first thought the last point may seem irrelevant. We are taking the term ^'school success," however, in an inclusive sense. In estimating an individual's success in life we ordinarily take into account not only the objective record of his achievements, but also the impression he has made on his associates and superiors. The latter is a real part of his success. For our present purpose it is important to know whether a child whose intelligence is judged by the teacher to be ^ ^inferior'' is really capable of doing ^ ^superior" school work; or conversely, whether a child judged by the teacher as of '^superior" intelHgence is likely to do "inferior' ' school work. 104 RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 105 Comparison of Intelligence Quotient with the Quality of the School Work as Judged by the Teachers Table 28 shows the teachers' judgments as to the quahty of school work done by children of the intelli- gence quotient groups below 80, 80-89, 90-109, 110- 119, and 120 or above. TABLE 28 The Relation Between Intelligence Quotient and Quality of School Work (as Judged by the Teachers) Quality of Intelligence quotients School Work Below 80 89-89 90-109 110-119 120-above Total Very Inferior 12 5 8 25 Inferior 9 28 49 6 92 Average 8 34 173 32 10 257 Superior 1 5 60 32 15 113 Very Superior 12 3 2 17 Total 30 72 302 73 27 504 It is evident from the table that a fairly high correla- tion is present. No child doing ^ Very superior' ' school work has an intelligence quotient below the middle group 90-109, and no child doing ^Very inferior" work ranks in intelligence quotient above the middle group. The agreement, however, is far from perfect. The group doing '^average" school work contains intelligence quotients all the way from ^^below 80" to ^^120 or above." One child with an intelligence quotient below 106 STANFORD REVISION OF BINET-SIMON SCALE 80 is ranked as doing '^superior" school work. The correlation by the Pearson formula is .45. There are 51 cases out of a total of 504 in which the quality of the school work is two steps removed from the location required for perfect correlation. This is nearly 10 percent of all. An occasional dis- agreement of one step would naturally be expected, since we know that the quality of school work depends partly on factors other than intelligence, such as health, industry, conscientiousness, quality of teach- ing, etc. But a disagreement of two steps is serious. We have examined our data to see if any evidence could be found of the presence of constant factors tend- ing to explain the disagreement between intelligence quotient and quality of school work. In the table are 26 children whose school work is at least two grades better than the intelligence quotient would itself warrant. On looking up the facts about these children, we find that 19 of them are over-age for their grade. Of these, 10 are from two to four years over-age. Of course nothing else is needed to explain the disagree- ment in these cases. We know it to be true that a 10-year-old child with a mental age of 8 years (intelli- gence quotient 80) is usually just about able to do •school work of average quality in the second grade, or that a 13-year-old with a mental age of 10 years (intelligence quotient 77) can manage to get along fairly well in the third or fourth grade. Of the other cases of disagreement in this direction, 4 were kindergarten children. Perhaps these should be thrown out altogether, on the ground that the work of the kindergarten fails to bring out clearly differences of intelligence. Of the remaining cases, one girl of 8 years was described by the teacher as ^Very timid KELATION OF INTELLIGENCE TO SCHOOL SUCCESS 107 and sensitive/^ and it is possible that this caused an unfavorable showing in the tests. Another, also a girl, was described as a child of ^Vonderfully sweet disposition," that is, the kind of child we are always ready to give the benefit of a doubt. In two other cases there was an additional explanation, namely, an intelHgence quotient which was just over the border between two groups. If one of these had tested at 90 instead of 89, and the other at 110 instead of 109, one step of the disagreement would have been elimi- nated. Accordingly, we may say that of the 26 cases of serious disagreement, 19 are largely accounted for by over-ageness, 4 by the fact that the school attended was a kindergarten which has no formal work, and that in only 2 of the 26 cases was no explanation suggested by the data at hand. We will now consider the 24 children the quality of whose school work ranked two grades below what the intelligence quotient would lead us to expect. Since over-ageness accounted for nearly two-thirds of the disagreements in the other direction, we will nat- urally expect to find under-ageness a frequent cause of displacement downward. This is true, but to a less extent than one might have expected. Of the 24 children, 10, or nearly 40 percent, are under-age for the grade they are in. Seven of these, however, are only one year advanced beyond age, a degree of under- ageness which would hardly account for more than half of the observed disagreement in these cases. Four others are kindergarten children and so may be left out of account. In one other case the teacher had contradicted herself. The child in question had an intelhgence quotient of 125 but had been ranked only ^ ^average" in school work. The teacher's supple- 108 STANFORD REVISION OF BINET-SIMON SCALE mentar^^ statement, however, showed that the marks ranged from B to A in every subject except arith- metic, and other statements indicated that the child had unusual talent in composition and in the apprecia- tion of literature. Two cases were sufficiently ex- plained by illness and long absence from school. An- other child was described as lazy and incorrigible — traits which would affect the school work unfavorably. In two other cases about half of the disagreement was accounted for by an intelhgence quotient just over the dividing line. In the other 4 cases there was nothing in our data which suggested a reason for the inferiority of the school work below apparent intel- lectual ability, though it is possible that a fuller knowl- edge of the facts would have cleared up these cases also. An analysis of the one-step disagreements disclosed the same factors, though here there remained a some- what larger number for which the data failed to offer a clear explanation. This amount of disagreement, however, is not particularly significant, since it may come about in a great variety of ways having nothing to do with the validity of the tests. In conclusion we may say that even a two-step disagreement between intelligent quotient and the quality of a child's school work does not in itself argue against the validity of the intelligence test. At least 90 percent of these disagreements are found to be explainable wholly or partly by other facts. It is especially to be emphasized that rarely, if ever, is a child able to do school work, in the grade where he belongs by age, more than one degree superior to that which the mental age would lead us to expect. On the other hand, it more often occurs that the quality of RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 109 TABLE 29 Grade Distribution of 676 Children by Mental Age Mental Grade Attended Age I II III IV V VI VII VIII HS I HS II HS III Total 8 25 25.5% 4 4% 55 56.6% 18 18.4% 19 19.4% 1 1% 15 14.2% 1 1% 6 5.7% 17 20% 1 1% 1 .9% 3 3.5% 16 16.6% 1 1.2% 12 12.5% 21 27% ^^ 98 9 24 24.5% 4 3.8% 48 49% 98 10 30 28.5% 6 7% 1 1% 1 1.5% 49 46.6% 105 11 20 23% 8 8.3% 2 2.6% 38 44.6% 85 12 19 19.8% 7 9% 4 6% 40 41.1% 96 13 29 37.1% 21 31% 5 14% 2 16.7% 19 24.3% 78 14 16 23.5% 10 28% 1 26 38.2% 68 15 16 21 58.5% 9 75% 36 12 676 110 STANFORD REVISION OF BINET-SIMON SCALE the school work drops considerably below the level normal to the intelligence quotient in question. The chief causes are ill-health, irregularity of attendance, or the possession of moral or volitional traits unfavor- able to school success. Correlation Between Intelligence Quotient and Grade Progress We have made this comparison for the entire number of subjects, but since there is little opportunity for children below 8 years to become retarded we have included in the following table only those with a mental age of 8 years or more. Grade II is regarded normal for Mental Age 8, Grade III for Mental Age 9, etc. The 8-year mental-age group includes all mental ages from 7 years, 7 months, to 8 years, 6 months, and so on. The table of grade distribution by mental age shows that nine-year intelligence is found all the way from Grade I to Grade VII, inclusive; ten-year intelligence from Grade II to Grade VII, etc. Twelve-year intel- ligence, which here ranges from Grade III to Grade VIII, would doubtless have been found in the high school also, if tests had been made there in any con- siderable number. Table 30 shows the number and percent who, accord- ing to mental age, are retarded or accelerated 1, 2, 3 or 4 years. The table includes only the Mental Ages 8 to 16, inclusive. Table 30 is given to facilitate comparison with the data of others. It should be emphasized, how- ever, that the method of expressing the degree and amount of retardation and advancement in years for children of all ages taken together is misleading. RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 111 TABLE 30 Amount of Acceleration and Retardation in the Grades When Mental Age is Taken for the Basis Grade Below Mental Age Normal Grade for Mental Age Grade Above Mental Age 4 Yrs. or More 3 Yrs. 2 Yrs. 1 Yr. 1 Yr. 2 Yrs. 3 Yrs. 4 Yrs. Total Number 4 13 69 184 275 106 22 3 1 676 Percent .5 1.9 10.2 27.2 40.6 15.6 3.2 .4 .1 It overlooks the fact that a given amount of retarda- tion or acceleration is not equally significant at the different ages. Here, as in the case of mental age, a deviation of one year at the age of 8 is as serious as a deviation of two years at the age of 16. Obviously, when we measure age-grade progress of school chil- dren in terms of years of deviation from normal grade we are using a unit of measure which has no fixed value. Important as this is in the statistical treatment of the retardation problem, it has been consistently ignored in all of the very numerous studies of age- grade progress, from the pioneer work of Ayres on down. As a result, all of these studies are mislead- ing, particularly in the comparisons made as to the ^^ amount" of retardation or acceleration at the vari- ous ages and in the various grades. Reverting to Table 29, it may be pointed out that, after the age of 7 or 8 years, misplacement by one grade is not especially significant, as that could easily happen from any one of a number of causes, such as early or late entrance, illness, a little more or a little less than average industry, etc. But in 112 cases, or nearly 16 percent of all, there is a misplacement of 112 STANFOKD REVISION OF BINET-SIMON SCALE two grades or more. Eighty-five of these, or 123^^ per- cent of all, are cases of grade retardation below mental age; 26, or nearly 4 percent of all, represent grade acceleration beyond mental age. It is interesting to note that school retardation of 2 years or more (reck- oned on the mental-age basis) is about three times as common as acceleration of 2 years or more. On the basis of chronological age, the proportion of grade ac- celeration to grade retardation is even less than this. Our present task, however, is to find an explanation of the rather surprising disagreement between grade progress and mental age as determined by the scale. Taking up first the 26 children whose grade status is two or more years ahead of their mental age, we find that 19 of these are chronologically over-age for their grade. Ten of the 19 are from two to four years over- age. In other words, those who are accelerated in school on the basis of mental age are usually retarded on the basis of chronological age, certainly an inter- esting and instructive paradox. The explanation, however, is obvious. The school tends to promote children by age rather than ability, and although the very dull are allowed to become somewhat retarded, this retardation is ordinarily less than would be war- ranted by their actual mental retardation. For ex- ample, there are six children of Mental Age 10 in the sixth grade. Two of these are 14 years of age, two are 15, and one is 16. Of the two children of Mental Age 11 in the eighth grade, one is 17 years old, three are 16, and five are 15. Only two are normal age for the grade. Turning now to the 85 children who are retarded two or more grades below the norm for their mental age, we find that 23 percent are, on the chronological RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 113 age basis, actually accelerated, and that over half of the remainder are in the grade where they belong by chronological age. Only 8 percent of those who are retarded two years or more according to mental age are retarded as much as two years by chronological age. This confirms the suspicion that promotion is largely governed by chronological age and helps to explain why children of any given mental age are dis- tributed over such a wide range of grades. There are, of course, other factors which sometimes cause chil- dren to be enrolled in grades too low for their mental age, e. g., irregularity of attendance, illness, and lack of industry. Unfortunately, our blank for supple- mentary data did not call specifically for information on these points. Tables 29 and 30 gave the school progress of the children on the basis of mental age. Tables 31 and 32 give it on the basis of chronological age. Ages below 8 are disregarded, since at this time retardation and acceleration have had little opportunity to occur. Comparison of Tables 29 and 31 reveals the striking fact that, on the whole, the grade location of school children does not fit their mental age much better than it fits their chronological age. Except in the upper years, children of a given mental age are scat- tered over nearly as wide a range of grades as children of the corresponding chronological age. Plainly, the efforts made at school grading fail to give groups of children of homogeneous mental ability. That this is largely due to the incorrect grading of children of inferior and superior intelligence is easily shown by taking those whose intelligence quotient is practically normal, say between 96 and 105, and find- ing how these distribute themselves in the grades. 114 STANFOED RE^aSION OF BINET-SIMON SCALE TABLE 31 Age-Grade Distribtttion of Children Aboa^ 7 Years of Age (by Chronological Age) Grade Chron. Age I II III IV V VI VII VIII Total 8 Number Percent 23 24.5 5 4.5 49 52.3 20 21.2 2 1.2 16 14.6 3 2.7 17 20 2 2.3 22 28.2 1 1.2 5 6.2 1 1.2 9 9.5 94 9 Number Percent 32 29.3 5 6 2 2.6 53 48.4 109 10 Number Percent 20 23.5 7 9 5 6.2 2 2 41 48.2 85 11 Number Percent 18 23 8 10 7 7.4 2 2.6 1 2.1 28 36 78 12 Number Percent 26 32 11 11.6 1 1.3 36 44.4 81 13 Number Percent 35 37 19 24.3 2 4.2 1 9 31 32.8 95 14 Number Percent 19 24.3 9 19.1 1 9 37 47.5 78 15 Number Percent 16 Number Percent 35 74.5 9 8.2 47 11 Total 678 TABLE 32 Number and Percent of Children Retarded or Accelerated 1, 2, 3, OR 4 Years for the Ages 8-14 Combined Retarded In Grade for Age Accelerated Total 4 3 2 1 1 2 3 Number 4 15 55 173 275 89 9 1 620 Percent .6 2.4 8.8 27.9 44.3 14.3 1.4 .1 100 RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 115 This method gives the correlation relatively freed from the constant tendency of teachers to over-pro- mote the dull and under-promote the superior chil- dren. Table 33 gives this distribution. TABLE 33 Grade Distribution by Chronological Age op Children with Intelligence Quotient between 96 and 105 Chron. Grade Age I II III IV V VI VII VIII Total 8 9 25 4 3 3 2 1 38 9 10 25 38 10 10 2 19 32 11 5 1 17 26 12 9 1 1 10 20 13 15 6 14 31 14 8 4 15 30 15 16 7 1 11 1 Total 31 42 41 28 31 33 26 24 227 116 STANFORD REVISION OF BINET-SIMON SCALE Of the 227 children appearing in the above table, only 4 who are below the age of 14 are more than one grade removed from the place where they belong by chronological age. All the two-grade displacements are in the direction of retardation. That is, the child with an intelligence quotient between 96-105 is never found (in our data) two grades advanced in school; and the chances are about 50 to 1 that if he is under 14 years of age and tests between 96-105 he will not be as much as two years retarded. (Of 198 children with ages 8-13, 4 are retarded two years.) At ages 14 and 15 selection has taken place and the proportion of retardation is naturally much larger. Another interesting comparison may be made by taking the extreme intelUgence quotients and finding the location in the grades for the exceptionally dull and exceptionally bright children of each chronological age. We have done this for the intelligence quotients above 120 and below 80. The results are shown in Tables 34 and 35. Of the 68 children appearing in Table 34, full supple- mentary information is available regarding 34. Of these, not one is doing less than ^'average" work in the grade attended, while 23 are graded as doing either ^' superior'^ or '^very superior" school work. Of the 8 who are advanced two grades beyond chrono- logical age, we have supplementary information re- garding 5. Of these 5, every one is graded as doing either '^superior" or ^^very superior" work, and every one is ranked by the teacher as either "superior" or '^very superior" in intelligence. It will be noted that of the 54 children 7 years old or above, 15 are in the grade where they belong by chronological age, while 3 are even retarded one year RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 117 TABLE 34 Age-Grade Distribution of Children with Intelligence Quotient 120 OR Above Chron. Grade Age Kgn. I II III IV V VI VII VIII Total 6 2 11 1 14 7 6 3 2 11 8 1 7 2 5 1 1 11 9 6 10 1 3 3 1 6 3 2 8 11 1 1 8 12 3 6 13 1 1 4 14 15 16 Total 2 17 5 10 11 6 11 4 2 68 118 STANFOnD TIEVISION OF BTNKT-RTMON SCALE by chronological i\y:,v. That is, IS, or one-third of all those 7 years old or older having an intelligence quotient of 120 or above fail to reap any advantage {as far as pro- niolurn is concerned) from tJieir very superior intelli- gence. They nrc all doiiij;- 'Scry superior" to 'Sivcr- afz;i^" school Avork and would (loubllcss (Hniiinue the sninc record if accorded t he vxivn i)roniot ions warranted by their intelligence ({uotient. The n^luctance of teachers to give sucli promotions is probably due both to inertia and to an \niwiHingness to part with excep- tionally satisfactory ])upils. Of the 12 children appearing in the above table, all of whom have between two-thirds and four-fifths in- t(*Hig(Mict» (intelligence (|uotient 05 to SO), only two are in the grade where they bi^long by chronological age. Jioth of these W(M'e doing "very infiM'ior" s(^hool work and neither was promoted the following year. Six of the 42 are only one year retarded. Su|)pltMnentaiy data are available for only four of the six. Two of these four are doing 'Wery inferior" work, two '* in- ferior" work. Of tlie 18 an ho are retarded two years, supplemtMitary data are available for 1 1, four of whom are said to be doing "average" work, four "inferior" work, and three 'Wery inferior" work. Of the 10 re- tarded three years or more, we have supplementary data for 10, three of whom are doing '^average" work, four "inferior," and three "very inferior." It is inter- esting to note that two of the tlu'ce w^ho are doing "average" work are fonr years retarded: one is 13 years old and in the tliird grade, the other is 14 years old and in the fourth grade. This is really what we should (^xpect of high-grade feeble-minded children of i;> and 14 years. RELATION OF INTELLKHONCE TO BCIIOOL SUCCEBB 119 'I'AItLIO ;{5 Aaro-GuADK Dihtuibution of CiiiummN with Intki.kkjknok (^i/o- TIKNTS liWLOVV 80* Ofiron. Orado Ak<5 1 II III IV V VI VII VIII ToUl 8 2 3 2 9 2 4 1 5 10 2 4 1 4 11 2 5 1 3 12 2 6 13 1 2 1 2 11 14 1 1 1 3 15 16 17 1 2 1 3 3 2 Total 5 7 7 8 2 4 5 4 42 * (35 of the 42 have intolllgenccj quotients between 70-79.) 120 STANFORD REVISION OF BINET-SIMON SCALE The foregoing is suggestive because indicative of what three-quarter intelUgence can do. A child of this degree of deficiency is usually two to four years below grade for his age, and his work is usually ''infer- iors^ or ''very inferior. '^ Rarely is he found in the grade where he belongs by chronological age and he never does better than "inferior" work there. We learn less from Table 34 of what pupils of intelli- gence quotient 125 can do than we do from Table 35 of what pupils of intelligence quotient 75 can do. The reason is that the school does not often give the superior child a chance to work up to his proper level of per- formance. Compared to their possibilities, children of exceptionally superior intelligence are usually retarded, just as we found exceptionally inferior children almost always above the grade where they belong by mental age. Most of the apparently much-retarded children are really accelerates; many of the exceptionally accelerated children are really retardates. TABLE 36 Correlation Between Intelligence Quotients and Intelligence Estimated by the Teachers Teachers' I Q Estimates Below 80 80-89 90-109 110-119 120 and up Total Very Superior Superior Average Inferior Very Inferior 2 10 11 8 2 3 37 22 4 7 60 184 34 6 4 29 39 3 17 7 16 111 277 67 18 Total 31 68 291 72 27 389 RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 121 By the Pearson formula the correlation contained in Table 36 is 0.48. This is about what others have found, and is both large enough and small enough to be significant. That it is moderately high in so far corroborates the tests : that it is not higher means that either the teachers or the tests have made a good many mistakes. We note first that 24 children in the above table are placed two steps higher in the teachers' estimates than the intelligence quotient would suggest, while only 13 are displaced two steps downward. This discrepancy would indicate that there is probably some factor caus- ing teachers to overestimate the intelligence of those whose test performance is low. On looking into the matter we find that of the 24 children misplaced up- ward by the teachers' estimates, 14 are from two to four years over-age for their grade. These cases are, therefore, sufficiently explained. It is the teachers who were at fault, not the scale. In judging the intel- ligence of these children they forgot to make allowance for the over-ageness. Finding them about on a par in intellectual maturity with other children of their classes, they judged them equally intelligent. Of the remaining 10 children in this group, three were in the kindergarten, where the teacher has little opportunity to form an opinion as to a child's intelli- gence. In another case, that of a boy with an intelli- gence quotient below 80 who was ranked ^^ average," the teacher had contradicted her own estimate by adding an explanatory note which made it clear that the boy was probably a borderline case or even feeble- minded, though possessed of some ability to profit by drill work suited to children a year or two below his age. In four other cases the intelligence quotient was 122 STANFORD REVISION OF BINET-SIMON SCALE just over the dividing line, making the disagreement between it and the teacher's estimate appear ahnost twice as great as it really was. In only two of the 24 cases was there no information at hand that would explain all, or nearly all of the disagreement. Of the 13 who were displaced two steps downward in the teachers' estimates, we find that five were from one to two years under-age for their grade. Their intelligence had accordingly been judged by a standard which was unfair to them; that is, by a standard based upon the average intelligence of older children. Two were kindergarten children. In another case the teacher, after ranking the child in the ^'very inferior" group, added a note saying that the child was very deaf and that this might account for the apparent stupidity. The test gave this child an intelligence quotient of 95, which was probably not far from cor- rect. Half the disagreement could be accounted for in two other cases by the presence of the intelligence quotient near the dividing line. This leaves 3 cases of two-step downward displacement still unexplained, though we are inclined to suspect that if more facts were available, these, too, could have been cleared up. Similar reasons appear to account for approximately half of the one-step disagreements. When all the ex- plained disagreements were eliminated from Table 9, the correlation rose from .48 to .71. Another way to get at the degree of agreement be- tween intelhgence quotients and the teachers' esti- mates (as the latter would be if freed from the con- stant error due to neglect of age differences) is to com- pute the correlation separately for those children who are in the grade where they belong by chronological age. When this is done the coefficient of correlation, as may be found from Table 37, rises from .48 to .57. RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 123 TABLE 37 I Q Teacher's Estimates Below 80 80-89 90-109 110-119 120 or above Total Very superior 1 3 2 6 Superior 27 20 3 50 Average 6 80 16 1 103 Inferior 4 12 16 Very inferior 1 3 2 6 Total 1 13 122 39 6 181 Still another method of showing how strongly teachers tend to base their estimate of a child's intel- ligence upon the quality of his school work, to the neglect of age differences, is to take their classification of a group of children according to intelligence, and their classification of the same children according to school work, and ascertain the degree of correlation between the two groupings. We have done this, with the result shown in Table 38. TABLE 38 Correlation Between the Teachers' Groupings According to Intelligence and According to Quality of School Work Teachers' Classification Teachers' Classification According to Intelligence according to School Work Very Inferior Inferior Average Superior Very Superior Total Very Superior Superior Average Inferior Very Inferior 4 13 12 45 11 16 212 35 2 3 83 22 1 12 6 15 105 246 85 26 Total 17 68 265 109 18 477 124 STANFORD REVISION OF BINET-SIMON SCALE The correlation is .82 and would probably have been still higher if the supplementary form filled out by the teachers had not contained the specific instruction to estimate the intelligence of a child "in comparison with other children of the same ageJ^ In spite of this injunction, they have obviously ignored age differences and estimated intelUgence chiefly on the quality of the child's school work in the grade where he happened to be. They have failed to realize that quality of school work is no index of intelligence unless age is taken into account. The question should be, of course, not: '4s this child doing his school work well?" but rather: "m what school grade should a child of this age be doing satisfactory work?" A high-grade im- becile may do average work in the first grade and a high-grade moron average work in the third or fourth grade, provided only they are sufficiently over-age for the grade in question. Our experience in testing children for segregation in special classes has time and again brought this peculiar fallacy of teachers to our attention. We have often found one or more feeble-minded children in a class after the teacher had confidently asserted that there was not a single exceptionally dull child present. In every case where there has been opportunity to follow the later school progress of such a child the substan- tial accuracy of the mental test has been confirmed. The following are typical examples of the neglect of teachers to take the age factor into account when estimating the intelligence of the child over-age for his grade : A. R. Boy, age 17, mental age 11, sixth grade, school work "nearly average," teacher's estimate of intelligence ''average." Test plainly shows this child to be a high-grade moron, or borderliner at best. Had attended school regularly 11 years and had made 6 grades. Teacher had compared child with his 12-year-old classmates. H. A. Boy, age 13, mental age 9-6, low fourth grade, school work ''inferior," teacher's estimate of intelligence "average." The teacher KELATION OF INTELLIGENCE TO SCHOOL SUCCESS 125 blamed the inferior quality of school work to "bad home environ- ment." As a matter of fact, the boy's father is feeble-minded and the normahty of the mother is questionable. An older brother is in a reform school. We are perfectly safe in predicting that this boy will not complete the eighth grade, even if he attends school till he is 21 years of age. F. I. Boy, age 12-11, mental age 9-4, third grade, school work "average," teacher's estimate of intelligence "average," social environ- ment "average," health good and attendance regular. Intelhgence and school success are what we should expect of an average 9-year old. D. A. Boy, age 12, mental age 9-2, third grade, school work "in- ferior," teacher's estimate of intelligence "average." Teacher im- putes inferior school work to "absence from school and lack of inter- est in books"! We have yet to find a child of 75 intelligence quotient who was particularly interested in books or enthusiastic about school. C. U. Girl, age 10, mental age 7-8, second grade, school work "average," teacher's estimate of intelligence "average." Teacher blames adenoids and bad teeth for retardation. No doubt of child's mental deficiency. P. I. Girl, age 8-10, mental age 6-7, has been in first grade two years and a half, school work "average," teacher's estimate of intelh- gence "average." The mother and one brother of this girl are feeble- minded. H. O. Girl, age 7-10, mental age 5-2, first grade for 2 years, school work "inferior," teacher's estimate of intelligence "average." The teacher, nevertheless, adds: "This child is not normal, but her ability to respond to drill shows that she has intelhgence." It is true that even feeble-minded children of 5-year intelhgence are able to profit a httle from drill. Their weakness comes to light in their inabihty to perform higher types of mental activity. The following are examples of the under-estimation of intelligence and school ability of children who are under-age for their grade: M. L. Girl, age 11-2, mental age "average adult" (16), sixth grade, school work "superior," teacher's estimate of intelhgence "average." Teacher credits superior school work to "unusual home advantages." Father is a college professor. The teacher considers the child accelerated in school. In reahty, she ought to be in the second year of the high school, instead of in the sixth grade. H. A. Boy, age 11, mental age 14, sixth grade, school work "aver- age," teacher's estimate of intelhgence "average." According to the supplementary information the boy is "wonderfully attentive," "studious," and possessed of "aU-round ability." The estimate of "average intelhgence" is probably due to the fact that he was com- pared with classmates who averaged about a year older. K. R. Girl, age 6-1, mental age 8-5, second grade, school work "average," teacher's estimate of inteUigence "superior," social environ- ment "average." Is it not evident that a child from ordinary social 126 STANFORD REVISION OF BINET-SIMON SCALE environment who does work of average quality in the second grade when barely 6 years of age, and who has an intelligence quotient of about 140, should be judged "very superior" rather than merely "superior" in intelligence? S. A. Boy, age 8-10, mental age 10-9, fourth grade, school work "average," teacher's estimate of intelligence "average." Teacher attributed school acceleration to "studiousness" and "dehght in school work." Our own guess would be that these traits are, rather, indications of unusually superior intelligence- Ill a special study of a group of superior children, tested separately from the present investigation, we have found even more striking examples of the difficulty teachers have in recognizing superior ability. One case was that of a boy aged 10-6 with an intelligence quotient of 148. He was in the sixth grade, doing '' superior '^ work there, and yet was judged by the teacher to have '^no unusual ability." It was learned from the parents that the boy is distantly related to Meyerbeer, the composer, and that at an age when most children are reading fairy stories, he has a passion for difficult medical literature and text books in phys- ical science. The question has suggested itself, whether teachers^ estimates of intelligence vary in reliability with chil- dren of different ages. We have divided our children into three groups, according to age, and have com- puted for these groups separately the correlation between the intelligence quotient and the teachers' estimates. Ages 5, 6, 7 and 8 were placed in one group; Ages 9, 10 and 11 in another; and Ages 12, 13, 14, and 15 in a third. The coefficients were, in order, .48, .60 and .46. It appears, therefore, that teachers probably make fewer errors with pupils of the middle group, though the difference is not great. Such facts as we have set forth in this chapter sug- gest that, while the judgment of a teacher regarding a RELATION OF INTELLIGENCE TO SCHOOL SUCCESS 127 child's school success and intelligence may, if properly safeguarded, afford valuable data to supplement the results of the intelligence test, the assistance is more likely to be in the other direction; more often it is the test which can keep us from being misled by the erroneous judgment of the teacher. SUMMARY 1. The correlation between intelligence quotient and the quality of the school work as judged by the teachers is .45. An examination of the marked cases of dis- agreement between intelligence quotient and school work shows that these are due largely to the teachers' neglect of age differences in estimating quality of school work. 2. The correlation between intelligence quotient and the teachers' rankings according to intelligence is .48. Detailed study of the cases of disagreement justifies the conclusion that they are due mainly to certain constant errors to which teachers are subject in esti- mating a child's intelligence. Here, as in judging quality of school work, the most common error is that of overlooking age differences. Teachers judge in- telligence mostly by the quality of school work in the grade where the child happens to be located. This results in over-estimating the intelligence of older, retarded children, and under-estimating the intelligence of the younger, advanced children. 3. The wider disagreements between intelligence quotient and grade status are confined chiefly to those children who are superior to, or below the average in ability. The explanation for this has been found in the fact that the tendency of the school is to promote children by age rather than by ability. Those who 128 STANFORD REVISION OF BINET-SIMON SCALE have an intelligence quotient between 96 and 105 are hardly ever more than one grade removed from the location which is normal to their mental age. 4. The child with an intelligence quotient of 120 or above is rarely found below the grade for his chrono- logical age, and occasionally he is one or two grades above. Wherever located, his work is nearly always superior, and the evidence suggests strongly that this superiority of school work would continue even if extra promotions were granted. Superior children are seldom allowed to reap the advantage, in school prog- ress, to which their superiority fairly entitles them. 5. The child of 70-79 intelligence quotient never does satisfactory work in the grade where he belongs by chronological age. After the age of 8 or 9 years such a child is usually found doing ''very inferior" to ''average" work in a grade two to four years below his age. 6. Although the disagreements between intelhgence quotient on the one hand and grade progress, quality of school work, and teachers' estimates of intelligence on the other hand would at first seem to justify serious misgivings as to the value of the intelligence scale, these same disagreements, on closer examination, are found to offer the strongest evidence in support of the test method. CHAPTER VII THE VALIDITY OF THE INDIVIDUAL TESTS Criteria of a Tesfs Validity The first task in the construction of an intelligence scale is to select tests which are really tests of intelli- gence, tests which are not too much influenced by age, home environment or school instruction apart from native endowment. There are three criteria which a test must satisfy before it can be accepted as a valid measure of intelligence. In the first place, since we know that intelligence is to a certain extent a function of age, a test to be valid must show an increase from year to year in the percentage of unselected children that pass it. This is the criterion on which Binet chiefly rehed. Nearly all the tests which he finally included in his scale satisfy this criterion fairly well, though some show a more rapid increase than others. This, however, is not sufficient. Many other traits besides intelligence are also functions of age. Height, weight, chest girth, length of forearm, in fact, any physical trait influenced by growth would show a steady increase from age to age in the percentage that pass a given standard. Yet it is easy to show that tests of this kind have no place in an intelligence scale. If 100 unselected 10-year-olds were measured for length of forearm it would of course be found that the average for all considerably exceeds that for 9-year olds; but it would also doubtless be found that this is little if any more true of 10-year-olds who have superior intelligence than of 10-year-olds who have inferior inteligence. That is, although intelligence and 129 130 STANFORD REVISION OF BINET-SIMON SCALE length of forearm are both functions of age, they have no direct relationship to each other. Such a test would not be found coherent with any already existent intelligence scale. Similarly, if a given test in the Binet series does not agree with the scale as a whole, if 10-year children who by the scale have 11-year intel- ligence do not pass it any more frequently than those 10-year children who have 9-year intelligence, then either this test is worthless or the scale as a whole lacks validity. The entire scale must be coherent. But coherency and age-increase in the percentage that pass do not themselves guarantee the validity of a series of tests for the measurement of intelligence. A set of tests made up of a great variety of physical measurements might very well satisfy both of these criteria. If we have no already existing intelligence scale with which to compare an individual test, then we must compare the test with intelligence as other- wise estimated, for example, with teachers' rankings. If children who are ranked as intelUgent succeed with it no better than those who are ranked as dull, then the test is of doubtful validity. For our present purposes the third criterion may be left out of account in judging the validity of individual tests of the Binet scale. It has been amply demon- strated that the scale as a whole gives a fairly reliable index of intelligence. Its results always show a reasonably high correlation with intelligence as judged by teachers or other observers. We have already shown that its correlation with school success is fairly high, particularly when allowance is made for certain tendencies to error in the estimation of school success. Its use with feeble-minded children in institutions has been especially convincing. Long-continued observa- VALIDITY OF THE INDIVIDUAL TESTS 131 tion of such children rarely necessitates any serious correction of its verdict. As for the first essential — the requirement of an increase in the percentage of passes from year to year — ^it is evident from all the available statistics that all the tests which we have included in the revision meet this criterion in a fairly satisfactory way. Some tests show more rapid increases than others, but not one is passed by equal percentages in three successive years. Accordingly, the criterion of most importance for our purpose is the second one — that of coherency. Since we know that the scale as a whole is fairly reli- able, we can measure each individual test against the entire scale. A test which gives results out of harmony with the results of the scale as a whole can not be con- sidered a satisfactory test of intelligence, whatever increase it may show in the percentage that pass it from year to year. This increase might be due to other factors than intelligence, such as school instruc- tion or the incidental experiences which come with increasing age. One way of applying this criterion would be to classify all our subjects by mental age as determined by the scale and then note the number that pass a given test at successive mental ages. This method gives valu- able information, and we have had to rely on it to a large extent in evaluating and placing the tests of the upper-year groups. It has, however, one objectionable feature; the results are more or less influenced by age, apart from intelligence. Children of 8-year mental age, for example, range in chronological age all the way from 53^ or 6 years to 11 or 12 years, and it is conceiv- able that these large age differences might have a considerable influence on the number that pass a given 132 STANFORD EEVISION OF BINET-SIMON SCALE test. The only way to separate the influence of intelli- gence from that of age is to take a large number of unselected children of one chronological age and find the percentage of passes separately for the bright and dull children of that group. Correlation of the Individual Tests with Intelligence Quotient Following the foregoing plan, we have divided the children of each chronological age into thi*ee groups according to magnitude of intelligence quotient. We have placed in the middle or normal group those chil- dren of a given age having an intelligence quotient between 96 and 105, in the inferior group those with an intelligence quotient below 96, and in the superior group those with an intelligence quotient of 106 or above. At most of the ages this gives three groups of about the same size. Had we tested a larger number of children of each age, it would probably have been better to place more children in the middle group, say all between 91 and 110 intelligence quotient. This would have heightened the contrast between the inferior and superior groups. However, such a plan would have placed about 60 percent of our cases of a given age in the middle group and left only about 20 percent, only 12 to 20 children, for each of the other groups. Accordingl}'-, in order to obtain three groups of nearly equal size we have included in the middle group those with an intelligence quotient between 96 and 105. Table 39 shows the results of this comparison. The figures are in all cases the percentages of passes for children of the chronological age in which the test is located. The three columns give these percentages for children with intelligence quotient 95 or below, 96-105, and 106 or above, respectively. VALIDITY OF THE INDIVIDUAL TESTS 133 TABLE 39 a CO .g fl o fin ^ tiD 6D C3 is •*^ t-i +3 o !=^ 7 sa o © te ® . ^ q "to © © » \\ « \ ' « 0^.. Ov ^ •^ ^<<- \>^ .^ ^ ^ '"^J^^^^ "'-''' ^ '^ O^ . s ' ^ ^ 1 I •^OO^ 1^ •^ v^^^. \' p\^ \^ '=^. * . o 4 ^ >- 'J 5^ ^ ^,, -V <9 .j:<^'^ ''^OC^- <.v'''^. '^ sips ^ •^^'■' :^ ^^^ 4 ^^ } ^^""^ :^/ ■■ ■^ A O 0^