Book COPYRIGHT DEPOSIT MANUAL OF MENTAL AND PHYSICAL TKSTS Part II: Complex Processes [SECOND EDITION, REVISED AND ENLARGED] MANUAL OF MENTAL AND PHYSICAL TESTS In Two Farts Part II: Complex Processes A BOOK OF DIRECTIONS COMPILED WITH SPECIAL REFERENCE TO THE EXPERIMENTAL STUDY OF SCHOOL CHILDREN IN THE LABORATORY OR CLASSROOM BY GUY MONTROSE WHIPPLE, PH.D. ASSOCIATE PROFESSOR CF EDUCATION, UNIVERSITY OF ILLINOIS AUTHOR OF "a GUIDE TO HIGH SCHOOL OBSERVATION," "QUESTIONS IN general and educational psychology" "questions in school hygiene" BALTIMORE, U. S. A. WARWICK & YORK, Inc. 1915 Copyrig-ht, 1915 BV Warwick & York. Inc. JUL 24 1915 IGI.A410214 PREFACE TO PART II If it be not thought bad form to preface a volume with an apologj^, I should like to ask the indulgence of those of my readers who have been so complimentary as to express to the publishers and to me their irritation at this delay of two years or more in the completion of the revised second edition of the Manual. The truth is, that the unexpected exhaustion of the V" first edition of the book found me quite unprepared to rewrite the text at short notice, and that the whole subject of mental tests had meanwhile so expanded as to present a task of no small magnitude to one who would seek to deal at all adequately with the material that had become available. In this volume, then, as in Part I, the text has undergone extensive revision and alteration. In a number of instances the addition of new materials, of new methods and of new re- sults has been sufficient to alter the complexion of the tests so decidedly as to amount to entirely ucav presentations of the topics with which they deal. The Kent-Rosanoff Test and the Analogies Test are introduced as totally new material. On the other hand, I have been compelled, reluctanth^ for reasons set forth in the text, to omit consideration of serial graded tests (Chapter XIII). This omission I hope to repair later on by publishing a supplementary volume dealing with Systems of Tests in general. To incorporate this material in the present volume would increase its size unduly and delay its appearance beyond reasonable limits of time. In addition to the acknowledgments for assistance made in the preface to Part I, my thanks are extended to Miss Margaret Cobb and Dr. H. O. Rugg for valuable assistance in the reading of proof. Other special acknowledgments I have tried to make in the course of the text. G. M. W. University of Illinois, April, 1915. ^ c TABLE OF CONTENTS Chapter VIII. Tests of Desceiption and Report Test 31. — Description of an Object [376 Test 32.— Fidelity of Report : Aussage Test [38.3 Chapter IX. Tests of Association. Learning, and Memory Test 33. — Uncontrolled Association — Continuous Method [410 Test 33 A. — Uncontrolled Association — Discrete Method (Kent RosanofC Test) [419 Test 34. — Controlled Association — Logical Relations [437 Test 3.iA. — /Uialogies [455 Test 35. — Controlled Association : Computation [460 Test 36'.— Mirror-Drawing [485 Test 37.— Substitution [490. Test 38. — Memory for Serial Impressions: 'Rote' Memory. ... [515 Test 35.— Memory for Ideas : 'Logical' Memory [570 Chapter X. Tests of Suggestibility Test 40. — Suggestion by the Size- Weight Illusion [589 Test 41. — Suggestion by Progressive Weights [597 Test 42. — Suggestion by Progressive Lines [601 Test 43. — Suggestion of Line-Lengths by Personal Influence. . .[607 Test ,^^. — Suggestion by Illusion of Warmth [611 Chapter XL Tests of larAGiNATiON and Invention Test /,.5.— Inlj-Blots [619 Test 46. — Linguistic Invention [625 Test //7.— Word-Building [639 Test .^S.— Ebbinghaus' Completion-Method [648 Test 49. — Interpretation of Fables [665 Chapter XII. Tests of Intellectual Equipment Test 50.— Size of Vocabulary [673 Test 51. — Range of Information [682 Chapter XII. Note on Serial Graded Tests for Developmental Diagnosis [689 Appendix I. Formulas for Converting Measures (English and Metric Systems) [691 Appendix II. List of Abbreviations [691 Appendix III. List of Materials for Part II [693 Index of Names [695 Index of Subjects .' [699 10 10 44 307 316 323 325 325 327 329 333 INDEX OF FIGURES ?IG. PAGE G4. The Mirror-Dra\Ying Test [488] 122 65. The EfCect of Practise During Six Trials upon the Average Speed of Mirror-Drawing (From Yoakum and Calfee) . [492] 126 60. The Effect of Long Practise upon Mirror-Drawing (From Starch) [493] 127 67. Percentiles for the Substitution Index for Cincinnati Working Children 14 and 15 Years Old— Sheet 1 (After Woolley and Fischer) [510] 144 68. Percentiles for the Substitution Index for Cincinnati Working Children 14 and 15 Years Old— Sheet 4 (After Woolley and Fischer) [51i] 145 69. Effect of Practise Periods of Different Lengths upon Perform- ance in the Substitution Test (From Starch) [512] 146 70. Jastrow's Memory Apparatus [525] 159 71. Development of Memory for Digits (From Smedley) [543] 177 72. Percentiles of Memory for Letter Squares, Boys (Ander- son) [546] 180 73. Percentiles of Memory for Letter Squares, Girls (Ander- son) [547] 181 74. Memory Capacity of 12- Year-Old Pupils, by Grades (From Smedley) [563] 197 75 Size-Weight Illusion in the Feeble-Minded (Doll) [595] 229 76. Guidi's Apparatus for the Warmth Illusion (]Modifled by Whipple) [016] 250 77. Percentiles of Word Building for Boys (Anderson) [642] 276 78. Percentiles of Word Building for Girls (Anderson) [643] 277 79. Percentiles for Completion Test No. 4, Ages 14 to 17 Combined (Eraser) [660] 294 INDEX TO TABLES TABLE PAGE 50. Comparative Accuracy of Sworn and Unsworn Statements (Stern and Borst) [397] 31 51. Effect of Time-InterA'al on Report (Dallenbach) [401] 35 52. Dependence of Report on its Form (Stern and Borst) .... [402] 30 53. Effect of Practise on Coefficients of Report (Borst) [405] 39 54. Words Written in Tln-ee Minutes by Normal Children ( Pyle) [413] 47 55. AVords Uttered in Three Minutes by Epileptics (Wallin) . [414] 48 56. Distribution of Terms in 'Uncontrolled' Association (Jastrow, Nevers, Calkins, Manchester) [417] 51 57. Influences that Affect 'Uncontrolled' Series of Words or Drawings (Flournoy) [418] 52 58. Average Distributions for the Kent-Rpsanoff Test (Compiled from Eastman, Kent, Rosanoff and Strong) [426] 60 59. Dependence of Distribution in the Kent-Rosanoft" Test on Age (Isabel Rosanoff and A. J. Rosanoff) [427] 61 60. Individual Responses of Children and Adults in the Kent- Rosanoff Test (Isabel Rosanoff and A. J. Rosanoff) . . . [428] 62 61. Types of Associative Response in Normal and Feeble-Minded Children (Otis) [429] 63 62. Distribution of Responses as Conditioned by Intelligence (Isabel Rosanoff' and A. J. Rosanoff) [432] 66 63. Performance in the Part- Whole Test (Norsworthy) [441] 75 64. Correct Associates Written in 60 Sec. Part-Whole Test (Pyle) [441] 75 65. Performance in the Genus-Species Test (Norsworthy) ... [444] 78 66. Correct Associates Written in 60 Sec. Genus-Species Test (Pyle) [444] 78 67. Correct Associates Written in 60 Sec. Opposites Test (Pyle) [448] 82 68. Correct Associates of a Possible 20. Opposites Test (Nors- worthy) [449] 83 69. Correct Associates and Times. Opposites Test (Squire) .. [449] S3 70. Errors and Times. Opposites Test (Carpenter) [449] 83 71. Speed in Sec. for Correct Responses in the Analogies Test (Fraser) [458] 92 72. Average Scores of 29 Normal and 6 Insane Subjects in the Addi- tion Test ( Specht and Bischoff ) [474] 108 73. Efficiency in Addition : Five 10-Minute Periods (Schulze) . [477] 111 74. Efficiency in Addition and Multiplication within an Hour (Bur- gerstein) [478] 112 75. Additions per Pupil, with and without a Rest-Pause (Burger- stein-Schulze) [479] 113 76. Efficiency in Computation within a School Session (Laser) [480] 114 5 b MANUAL OF MENTAL AND PHYSICAL TESTS TABLE PAGE 77. Effect of Practise on Speed iu Mirror-Drawing. College Stu- dents (Whipple) [489] 123 78. Times, in Sec, for Mirror-Drawing (Yoakum and Calfee) . [490] 124 79. Times, in Sec./ Used in the Star Test by 36 College Girls, 16 College Maids and 69 Bedford Reformatory Women ( Weidensall) [495] 129 SO. Scores in the Star Test for Three Groups of Bedford Reforma- tory Women (Weidensall) [495] 129 81. Substitution Test. Number of Symbols Written. Form B. Group Metho.d (Whipple) [505] 139 82. Substitution Test. Speed in Seconds. Form TJ. Individual Method (Whipple) [505] 139 83. Correct Substitutions Made in 60 Sec. Digit-Sj^mbol Test (Pyle). ". [505] 139 84. Correct Substitutions Made in 60 Sec. Symbol-Digit Test (Pyle) [506] 140 85. Substitution Index, in Sec, Cincinnati Worliing Children ( Woolley and Fischer) [507] 141 86. Average Number of Substitutions Made by 37 White and 30 Ne- gro Girls in a Pennsylvania Reformatory (Baldwin) . . [507] 141 87. Substitution Test. Distribution of Gains and Losses in Speed (Whipple) [509] 143 88. Substitution Test. Bright and Dull Boys. Form B. Individual Method (Whipple) [513] 147 89. Differences, iu Sec, Index of Substitution, Form C, between Bedford Hills Reformatory Women and Cincinnati Working Girls 14 and 15 Years Old (After Weidensall) [515] 149 90. Use of the 'Footrule' Method in Scoring the Memory Test (Spearman) [527] 161 91. Norms of Memory Span for Digits, as Conditioned by Age (Smedley) ! [537] 171 92. Development of Memory for Digits (Smedley) [538] 172 93. Dependence of Memory Span for Auditory Digits on Age (Jacobs) [538] 172 94. Dependence of Memory for Auditory Digits on Age (Ebbing- haus) [538] 172 95. Memory for Letter Squares, in Relation to Age and Practise • (Winch) [539] 173 96. Memory for Letter Squares, Score for 10 Trials (Ander- son) [539] 173 97. Memory for Concrete Words, in Relation to x\.ge and Sex (Pyle) [539] 173 98. Memory for Abstract Words, in Relation to Age and Sex (Pyle) [540] 174 INDEX TO TABLE < TABLE PAGE 99. Average Number of Sentences Correctly Repeated (After Squire and Carpenter) [540] 174 100. Memory for Pictures of Objects (After Squire and Car- penter) [540] 174 101. Net Efficiency of Various Memories, in Relation to Age (Pohl- mann; [542] 176 102. Distribution of Boys' Scores in Letter Squares (After Ander- son) [548] 182 103. Distribution of Girls' Scores in Letter Squares (After Ander- son) [549] 183 104. Percentage of Accuracy in Memory for 2-Place Numbers (Sclniyten) [550] 184 105. Sex Differences in Memory Span for Digits in College Fresh- men (Wissler) ,...., [550] 184 106. Memory for 9-Terni Series of Different Kinds (Lobsien) . . [553] 187 107. Memory for Related and for Unrelated Words (Nors- wortby) [554] 188 108. Dependence of Memory upon Form of Presentation (Pobl- mann) [556] HX) 109. Relation of Memory for Auditory Digits and Intelligence (Ebbingbaus) [560] 194 110. Relation of Memory for Digits and School Standing ( Smedley ) [562] 196 111. Memory Span for Digits in the Feeble-Minded (Johnson) . [564] 198 112. Comparative Memory Capacity of Normal and Feeble-Minded Children (Norsworthy) [564] 198 113. Correlations with Amalgamated Results of Immediate Memory Tests (After Burt) [566] 200 114. Correlations between Memory Tests and Other Tests (After Wyatt) [566] 200 115. Recall of Different Members of a 7-Term Series (Binet and Henri) [566] 200 116. Marble Statue Test, Scores by Age and Sex (Pyle) [577] 211 117. Dutch Homestead Test. Words Written and Underlined (Whipple) .....[577] 211 118. Cicero Test. Distribution of 36 College Students (Whipple) ; [577] 211 119. Average Percentage of Loss in Third Reproduction (Hen- derson) [578] 212 120. Story of The Fire. Scores for Epileptics by Mental Age (After Wallin) [583] 217 121. Force of Suggestion (Gilbert) . [593] 227 122. Reactions of Feeble-Minded Children to the Size-Weight Illu- sion (Doll) . [594] 228 8 MANUAL OF MENTx\.L AND PHYSICAL TESTS TABLE PAGE 123. The Progressive- Weight Suggestion. 24 Cases (Binet) .. [600] 23-4 124. Averages for Progressive - Weight Suggestion by Age (Giroud) [601] 235 125. Percentage of 'Yields' to Contradictory Suggestion (Binet and Henri) [610] 244 126. Suggestibility to Warmth. Resistance-Coil Method (Okabeand Whipple) [617] 251 127. Suggestibility to Warmth as Related to Age. 187 Cases (Guidi) [618] 252 128. Average Number of 'Names' Given to Ink-Blots (Kirkpat- rick) [624] 258 129. Scores of Seven Adults in Developing Sentences (Sharp) . [629] 263 130. Averages by Age and Sex, aeobmt-Test (After Anderson and Pyle) [644] 278 131. Averages hy Age and Sex, aeirlp-Test (After Anderson and Pyle) [645] 279 132. Completion Text No. 4. 15-Minute Limit (Termau and Childs) [656] 290 133. Completion Text No. 4. 10-Minute Limit. Results for Boys (Fraser) [657] 291 134. CoiTipletiou Text No. 4. 10-Minute Limit. Results for Girls (Fraser) [658] 292 135. Completion Text No. 4. 15-Minute Limit. Results for 'Citizens' of the George Junior Republic (Whipple and Fraser) . . [663] 297 136. Distribution by Percentages in Interpretation of Fables (Ter- man and Childs) [670] 304 137. Norms for Interpretation of Four Fables (Terman and Childs) [671] 305 138. Average Vocabulary in Relation to Scholastic Status (Kirk- patrick ) [678] 312 139. Relation of Vocabulary to Age : Method of Terman and Childs [678] 312 140. Vocabularies of 20 College Students : Terman and Childs' List (Whipple) [679] 313 141. Distribution of Corrected Vocabulary Index, Seventy College Students (Whipple) [679] 3L3 142. 0^'erestimation of the Vocabulary Index. Seventy College Stu- dents (Whipple) [680] 314 143. Dependence of Range of Information on Academic Status (Whipple) [684] 318 144. Dependence of Range of Information on Academic Status (Smith) [685] 319 145. Dependence of Range of Information on Sex. (Whipple and Smith) [685] 319 PART II COMPLEX PROCESSES CHAPTER VIII Tests op Description and Report The two tests which are described in this chapter have certain features in common which demarcate them, on the one hand from the tests of perception and attention of the previous chap- ter, and on the other hand from the memory tests of the succeed- ing chapter, though, in many respects, they resemble these tests. The essential idea in both of the present tests is to determine capacity, not merely to attend and observe, or to recall what has been observed, but to put the results of this observation into linguistic form. If the observer gives his account of the experi- ence at the time of his observation, this constitutes description ; if at some time subsequent to his observation, this constitutes report. It is evident that this giving of an account of an experience, particularly if the experience be somewhat complicated in form, is a more complex psj^chical process than those under discussion in the tests of attention and perception. This greater com- plexity makes the reduction of the observer's performance to exact quantitative terras a matter of greater difficulty, but, on the other hand, the activity called forth is more akin to that demanded in everyday life, and it is for this reason that these tests have been felt to possess a peculiar value, particularly in the study of individual differences in mental constitution and mental efficiency. Again, language occupies so strikingly promi- nent a place in our mental economy that tests which seek to bring out the observer's ability to cast experience into linguistic form are, on that account, well worth while. This is particu- [375] 9 10 [376] DESCRIPTION AND REPORT larly the case in the second form of test, that of the report, which, in connection with the "ps3'Chology of testimony," has of late had a prominent place in psychological research. TEST 31 Description of an object. — The description test first came into prominence through the work of Binet, who urged that the study of individual psychology may be best advanced by resort to the experimental examination of complex, rather than of simple mental processes, and who considered the description test of special value in this connection, Binet made preliminary tests with Henri in 1893 (3), and worked at the test later by himself (1, 2). His method has been followed, though not in exact de- tail, by LeClere (7), Sharp (9), Monroe (8), and Cohn and Dieffenbacher (5). Materials. — Cigarette. Cancelled 2-cent postage stamp. Lithograph, entitled "Hindoos." I'he cancelled stamp was used by Monroe, the cigarette by Binet. The lithograph is substituted for tlie different pictures that have been used by other investigators (Binet and Henri used Neuville's "The Last Cart- ridge," Binet a picture representing Fontaine's "Le Laboreur et ses En- fants," Miss Sharp "The Golden Wedding" and "The Interrupted Duel"), because of the impossibility of securing these particular pictures, or of the difficulty of using them under the conditions that prevailed in the original experiments ( Bluet's school children were well acquainted with the fable froju Fontaine, for example). If it is desired to extend the list of materials, E may employ other objects used by Binet (2), such as a box of matches, a peimy, a leaf, etc. For group tests, there should be at least one picture for every 5 ;S's, one cigarette for every 2 *9's, and a stamp for each S*. For group tests, it would be desirable to secure a set of stamps whose cancellation marks were approximately the same. The stamps should be trijnmed off in such a manner as to show the full border of the stamp and a narrow margin of the paper upon which it was attached. The lithograph is one of a series called Leutemann's Types of Nations, catalogued by E. Steiger & Co., New York. It may be purchased, lilce all other material cited in this book, of C. H. Stoelting Co., Chicago, 111, Method. — (1) For the picture-test, supply S with writing materials; place the lithograph upright before him, about 75 cm. distant. Instruct him : "Write a description of this picture so that one who had never seen it would know all about it." Allow 10 min. TEST 31 : DESCRIPTION OF AN OBJECT [377] 11 (2) For the cigarette-test, give the following instructions, and no others : ''I'm going to put on this table before you a small object. I shall leave it there under your eyes. I want you to write a description of it; not to draw it, but describe it in words. You will have about 5 min. Here is the object." If S is busy at the end of the allotted time, or has written but a few lines, the time may be slightly extended. (3) For the stamp-test, proceed in a similar manner, save that >S"s are not forbidden to draw the stamp, if they wish to. The instructions may run: ^'Describe this postage stamp so that a person who had never seen one would know all about it." AlloAV 10 min., or more if needed. Variations of Method. — The problem of assigning an appro- priate title to a picture or of asking appropriate questions con- cerning it may be regarded as a variation of the description test. For suggestions as to this test see below, under Notes. For young children, and, indeed, for older ones under many conditions, it is better that E should write from /S"s dictation, perhaps stenographically. Treatment of Data. — In general, the results of the descrip- tion test are not intended to be submitted to exact quantitative treatment, but are to be inspected for the purpose of forming an opinion of >S^'s general mental type and capacity. The papers may, however, be treated quantitativel}^, by (1) counting the number of w^ords written, or (2) counting the number of lines written. E may, further (3), record in general terms the readi- ness and ease wath which S undertakes the description, and (4) may rate his paper as a whole, with respect to its comparative merit, on a score of 10 for a satisfactory or adequate description. (5) The description may, perhaps, be classified also with respect to its general type or character, following the classification adopted by Binet, Le Clere, and others as explained below. (6) It is possible, following Cohn and Dieffenbacher, to score de- scriptions more formally and precisely after the manner pro- posed for reports (Test 32). (7) Descriptions of the postage stamp may also be catalogued with respect to the items men- tioned, as was done by Monroe. 12 [378] DESCRIPTION AND REPORT Results. — (1) The description of an object is inadequate, because it is almost invariably simplified, i. e., a considerable number of its features, even important features, are unmen- tioned. Thus, in one of Binet's photographs, of the 22 objects or features that were mentioned at all, only 9.4 were mentioned, on the average, in each description. (2) This simplification or reduction in the description is the result of what might be termed a process of selection. Certain features are mentioned in practically all descriptions, others are mentioned only occasionally. By tabulating the number of times each feature is mentioned, one may discover some of the princi- ples which condition this selective process. Thus, in Binet's picture of the "Laborer," the old man is mentioned 36 times, his . sons 30, his bed 29, the seated woman 27, etc., until we come to relatively unimportant objects that may almost escape mention at all, e. g., a stick in the hands of one of the children — only 4 times in 36 descriptions. When pictures are used, persons are more often mentioned than furniture or other details of the set- ting of the scene. Similarly, in the stamp-test, tabulation indicates, according to Monroe, the following order of frequenc\y of mention: (1) word-inscriptions, (2) color, (3) number-inscriptions, (4) por- trait, (5) substance, (6) form, (7) use, (8) perforated edge, (9) size, (10) cancellation, (11) ornamentations. The item use de- clines with age : all others are mentioned more frequently as age increases. (3) Dependence on sex. Monroe states that girls generally mention more items than boys, and "seem to surpass boys in their knowledge of the postage stamp." It is not clear, however, whether this seeming superiority is due to better observation, to greater industry or to greater zeal and conscientiousness. Cohn and Dieifenbacher similarly find the descriptions by girls more comprehensive than those by boys. (4) Dependence on age. Cohn and Diefifenbacher tested school children 7 to 20 years of age with a colored picture (Tuss in Boots'). There was no clear augmentation of the range of description after 10..5 years in the case of the boys. At the age of 8, the description is predominantly an enumeration of objects, TEST 31: DESCRIPTION OF AN OBJECT [379] 13 though not a single color was mentioned by boys of that age. Esthetic and interpretative features are rarely noted before the 10th year, when a reflective element is firsL apparent. Increase of age is characterized by an increase in organization and sys- tematization of the descriptions. Actual errors are relatively uncommon : the few that are met with (fidelity is 97.2 per cent, among boys, 98.7 per cent, among girls) are often verbal mis- takes, the remainder true errors of apprehension. (5) Individual differences. In 150 accounts of the photo- graph, Binet found no two alike. This wealth of individuality makes the description-test at once valuable and difficult — valu- able as an indication of the variety of mental constitution, diffi- cult as to quantitative or comparative treatment. As an extreme illustration, one may contrast the following descriptions of a postage-stamp — the first b}^ a girl of 8, the second by a boy of 16. (a) "The postage stamp has a picture in it. The postage stamp costs two cents. It says united states postage ou it. The man has hair braided In back of his head. The Boarder is round. It has arms on it. The shape is square. The color is red. The man is White. You can get these to the postice [post-office] for two cents. There are lines around the boarder. The back of the stamp is white. It has nomber 2 on each side of it. The man has long hair." (b) "comments on the accompanying U. S. of AMERICA 2 CENT POSTAGE STAMP. "I. Its meaning: The Postage stamps have glorious history. In the past 57 years they have been more and more useful until now they are not only absolutely necessary, but constitute one of the great helps in the study of Geography, and one of the noblest pleasures for thousands and millions of people ; Kings and Queens as well as children in the most miserable social condition. "2. This Postage Stamp has the red color and is now next to the one penny stamps of Great Britain the most extensively used stamp used in the world. If I am not wrong its circulation in the past and present is the next largest of all others. The one penny stamp, I think has the first place. "3. Its surroundings are very interesting. It is mounted on a piece of paper, remainder of an envelope, which fact easily indicates that it is used in the most cases for letter correspondence. I notice [Continues in this and the fourth paragraph a description of the stamp itself.] "5. Some particular observations. I had 500-GOO of them at home which my cousin had the kindness to send me. Of course they are of no special value, but yet they teach my little brothers the important lesson that such a little thing, like a stamp, will do all the necessary things for the transportation of a letter or other mail matter from the Atlantic to the Pacific. It is very interesting to me that with the inarch of civiliza- tion the great Postal system of the World has increased its actions more 14 [380] DESCRIPTION AND tlEPOR'f and more until it is now one of the cliief functions under the sun. How much this single stamp has done I cannot say, but I know that some stamps, precisely like this, have done great service to the country." (6) Types. Notwithstanding this diversit}-, investigators have sought to classify descriptions into a limited number of tj'pes. Thus, Binet proposes four types — the descriptive, the observational, the emotional (poetic, imaginative), and the erudite — each present in varying shades and degrees. ( a) The describer, or enumerator, as one might term him, merely cata- logs the features of the ob.iect before him, with little regard for their interrelations, or for the meaning of the object as a whole. Example : "The cigarette has the general form of a cylinder, cut at one end by an inclined plane where the paper is folded. It is stuffed with a rather dark brown tobacco. The paper is striped lengthwise. The paper is somewhat bruised. The tobacco projects about 0.5 centimeter from one end." (6) The observer, though not necessarily more intelligent or clever than the describer, places more emphasis upon the interrelations of the several features that he mentions, interprets what he sees, conjectures and indi- cates the significance of the object as a whole. This type is also men- tioned by Mrs. Bryant in her 'deseription-of-a-room' test (4). Example : "A long, white, round object, composed of a pai)er cylinder, about 1/1' or % centimeter in diameter, filled with what is probably Oriental tobacco. It is about 7 centimeters long and must weigh about 6 grams [really 2 g.]. It is a badly rolled, uneven cigarette, and has been handled since it was pasted. In two places, to the right and left of the middle, the paper shows streaks as if it had been twisted. Other hori- zontal depressions indicate that there has been some pressure exerted upon the cigarette. I don't see the line where it has been stuck, but it must be badly fastened." (c) The emotional, imaginative, or poetic ^S? is less accurate in observa- tion, but introduces emotion, sentiment or imaginative intei*pretation in his description. Example : "It is a cigarette. It is thin, long, somewhat wrinkled. Its shape suggests a kind of elegant ease. Is it the cigarette itself or the memories that it awakes that remind me somehow of a scape-grace? The cigarette, there, all by itself on the table, makes me think of the bad student that goes off in the corner by himself to smoke. But I must write about the cigarette itself, and banish the idea of the smoker," etc. (d) The erudite S tells what he knows, what he has been taught, or interjects bits of personal information about the object. This may indi- cate the presence of an unusual fund of information, or it may indicate sheer laziness, in that it is often easier to write what one knows than actually to describe from direct inspection. Example : "We have before us here a cigarette. Let us see how it is made. In the first place, the exterior envelope is of light paper, called silk-paper. Then, inside is the tobacco. Tobacco is a product that grows almost everywhere in warm or temperate climates. The leaves of this shrub are gathered, and, after a treatment which lasts four years, are turned over to the public in the form of powder, that is, snuff, or in shreds, as in the present instance," etc. TEST 31: DESCRIPTION OP AN OBJECT [381] 15 Miss Sharp did not attempt a classification into types, but noted that ^''s observation "may be primarily directed to the particular objects or details of the picture, to the general arrangement of the objects, that is, the composition of the picture, or to the meaning of the picture, the story which it convey s, — the details observed being such as lead up to this inter- pretation, or explain and apply the interpretation that is given first. The different ways in which the same picture appeals to the various indi- viduals indicate differences in mental constitution." The results of LeClere's test are not directly comparable with those of other investigators, because his instructions were not to describe the object (gold watch), but to "write something that comes into mind as you look at it." He distinguishes in the contributions made by 30 girls, aged 13 to 17 years, seven types, viz. : description, observation, imagination, moral reflection, erudition, pure or simple emotion, and esthetic emotion. He does not find, however, that any one of his ^S's contributes a paper that may be classified in any one of these types, nor does any paper give evi- dence of a 'complete mind,' in the sense that all seven of the types are represented therein. In general, older or relatively more intelligent chil- dren write more varied or complex papers, i. e., approach the theoretically 'complete' type of mental constitution. Mention may be made here of the use of pictures in the Binet-Simon scale witli the simple question: "What do you see in that picture?" or "What is that picture about?" Credit is given the child according as his replies indicate mere enumeration or a comprehension of the total mean- ing of the scene depicted. Mile. Descoeudres (G) has extended the scoring of this form of description test by assigning scores to replies of different qualities and also by noting the number of ideas expressed. She gave a credit of 1 for simple enumeration, 2 if a phrase or sentence was used and 3 if the replies showed interpretation of meaning. Application of three pictures (not those used by Binet) to 14 backward children showed that rank in quality correlated distinctly with rank in quantity and also with estimated intelligence (.84, P.E. .02). Notes. — The attempt to use the description-test for classifica- tion of ^'s into types of mental constitution is of obvious inter- est. The drawing, from sucli a classification, of inferences as to the mental make-up of the >S's is as obviously hazardous, for ^ may write his description in the vein that he thinks is wauled by E. Thus, Binet had reason to think that several S's that he had classed as poetic or emotional were actually, in their every- day life, of a very matter-of-fact and unsympathetic disposition. In general, the drawing of inferences from the work of a?'s would become safer in proportion as the descriptions were in- creased in number and variety, i e., an /S' who wrote in an emo- tional vein in four descriptions of four different objects has, pre- sumably, a real emotional constitution. What may be regarded as a modification of the description test is the test employed by Squire (10), in which children of 16 [382] . DESCRIPTION AND REPORT various school grades were shown a series of 5 pictures and asked in each case (a) to supply an appropriate title to the pic- ture, and (6 ) to ask an appropriate question about the picture. The titles protfered by the children were classed under live rubrics : "mere enumeration of objects, description of pictures, unification in terms of action of principal figures, superficial unification in terms of relation to principal object and complete comprehension evidencing imaginative insight." With regard to the first problem, Mrs. Squire concludes: (1) No six- year-old cliild can be expected completely to comprehend a situation pre- sented pictorially. (2) Neither can a seven-year-old child be expected to give an adequate title. (3) The eight-year-old children are inclined to interpret meaning in terms of action, and a few are able to give super- ficial titles. (4) In the ninth and tenth years the titles given are mostly descriptive, but put tersely, rather than in disjointed statements. (5) By the twelfth year the majority of the names given will pass for titles, though a large proportion still deal with superficial aspects. (6) There are many cases of complete coinprehension in the thirteenth year. With regard to the second problem, replies may be classed as failures, irrelevant, minor or essential. There are no failures after the eighth year, while the percentage of 'essential' questions rises from 6 at age 7 to 58 at age 13. It is scarcely necessary to add that these conclusions obtain only for the particular pictures employed by Mrs. Squire, whose article should be consulted by those who seek to repeat this form of test of comprehension. REFERENCES (1) A. Binet, Psvchologie individuelle. La description d'un objet. AnPs, 3: 1896 (1897), 296-332. (2) A. Binet, L'etude experimentale de I'intelligence. Paris, 1903. Pp. 309. (3) A. Binet and V. Henri, La psychologic individuelle. AnPs, 2 : 1895, (1S9G), 411-4G5. (4) Sophie Bryant, Experiments in tes-ting the character of school children. J. Anlhrop. In.sl. of Great Britain and Ireland, 15 : 1886, 338-349. (5) .7. Cohn and J. DiefCenbacher, Untersuchungen iiber Geschlechts-, Alters- luid Begabungs-Unterschiede bei Schiilern. Bcikeftc zur ZAngPs, 2 : 1911, pp. 213. (6) Alice Descoeudres, Exploration de quelques tests d'intelligence chez des enfants anormaux et arrieres. ArPs (/), 11: 1911, 351-375. (7) A. LeClere, Description d'un objet. AnPs, 4:-. 1897 (1898), 379-389. (8) W. Monroe, Perception of children. PdSe, 11 : 1904, 498-507. (9) Stella Sharp, Individual psychology: a study in psychological method. AmJPs, 10 : 1899, 329-391. (10) Carrie R. Squire, Graded mental tests. JEdPs, 3: 1912, 363-380, etc., especially 373f. TEST 32 : FIDELITY OF REPORT [383] 17 TEST 32 Fidelity of report {Aussage test).— Capacity to observe, or range of observation, may be tested by methods previously de- scribed (Tests' 24 and 25) ; native retentiveness or capacity for recall may be tested by methods such as those that are described in subsequent sections ; capacity to describe what is seen may be tested as has been indicated in Test 31, but there exists a type of- activity, that of reporting a previous experience, which in a way combines these several activities, in that it demands both atten- tive observation, retention, recall, and an ability to marshal and formulate the items of experience in a verbal report (Aus- sage). In studying the 'psychology of testimony," interest has been developed of late in the direct examination by experimental methods of the capacity to report as such, and it has been found that reports may exhibit varying degrees of fidelity or reliability, more or less independently of the capacity that the reporters possess to observe or to retain experience; in other Avords, reports may contain discrepancies or inadequacies which are due, not only to misdirected attention, mal-observation and errors of memory, but also to lack of caution or of zeal for accu- rate statement, to scant}^ vocabulary, to injudicious phrase- olog}', or, of course, to deliberate intent to mislead.^ Historically, the idea of subjecting capacity of report to test seems first to have been definitely proi)Osed by Binet (3). Since ^It is true tliat no hard and fast line can be drawn between the report- test and the test of range of apprehension, or between it and the ordinary memory-test ; in tlie main, however, range of appi'ehension implies a brief exposure followed by simple enumeration of the objects seen, so that what is tested is capacity to grasp or observe, rather than capacity to retain or to formulate. And the stock memory-test measures the amount of mate- rial that can be reproduced ; in It the learning is usually by heart, and the reproduction is largely mechanical. In the report-test, the object is more complex, the time of scrutiny much longer than in the observation- test, while stress is placed as much upon quality as upon quantity of re- production, especially upon the fidelity of reproduction as conditioned by such personal factors as timidity, cautiousness, assurance, skill in verbal formulation, etc. Again, the typical memory-test comprises a direct verbal reproduction of verbal material, while the ty]iical Aussage test comprises a verbal presentation of material originally experienced as visual scenes (pictures, events, etc.), with or without some verbal features. Nevertheless, in the interrogation, the report-test does closely resemble an ordinary test of memory. 18 [384] DESCRIPTION AND REPORT then, the study of the psychology of testimony has found its most enthusiastic and active expositor in Stern, who has written an extensive monograph (31) on the subject, and in whose period- icals {Beitrdge zur Psychologie der Aussage and Zeits. f. arige- wandte Fsijchologie) most of the work of subsequent investi- gators has, directly or indirectly, appeared. The applicability of this line of work to many practical problems, particularly in the field of jurisprudence, is too obvious to need further com- ment. GENERAL METHODOLOGY OP THE REPORT-TEST 1. Choice of material. Of the several types of material that have been elaborated for the study of the report, e. g., the picture- test, the event-test, the rumor-test, etc., the first mentioned has many advantages for our present purposes.- Two types of pic- ture-test are prescribed; the first closely patterned after that employed by Binet in his study of suggestibility in school chil- dren, the second more in accord with the stock picture-test, as developed by Stern, Borst, Wreschner, Lobsien, and others. 2. Choice of evposiire-Umc. For pictures^ times ranging from 5 sec. to 7 min. have been used, though 45-60 sec. is most usual. The principle which has controlled the choice of expos- ure-time for the two tests that follow is to select such a period as will permit an average S to examine each detail of the object once. 3. Choice of time-interval. For the sake of brevity, the in- structions that follow prescribe a report directly after the ex- posure. If circumstances permit, E will find it of interest to extend the interval to several minutes, or even hours or weeks. The effect of lengthening time-interval has not as yet been satis- factorily determined. 4. Choice of form of report. There are two distinct forms of report, (1) the 'narrative' (Bericht, recit) , (2) the 'interroga- tory' {Verhor of Stern, Priifung of Wreschner, interrogatoire of -For a discussion of these advantages, of the several methods in detail, of the chief results, and for a general review of the whole field of the psychology of testimony, the reader is referred to an earlier discussion by the author (34). Suggestions for further tests will likewise be found therein. TEST 8i2 : FIDELITY OP repout [385] 19 Borst, forqage de memoire or questionnaire of Binet ) .^ The nar- rative is a free account, delivered by S, either orally or in writing, without comment, question, or suggestion by E: the interroga- tory is a series of prearranged questions; the replies to these questions constitute the 'deposition' (Vehorsprodukt) . The constituent parts of the narrative or the deposition may be termed 'statements' or 'items.' Each form of report has its ad- vantages and its disadvantages : both should be employed when- ever possible. 5. Choice of form of interrogatory. An interrogatory is 'complete' when its questions cover all features of the experience exhaustively, and are propounded to all S"s in the same order and manner : an interrogatory is 'incomplete' when its questions are restricted to such as refer only to those items not mentioned by >Sf in his narrative. The interrogatories that follow are de- signed to be complete, but E may, by appropriate selection, con- vert thein into the incomplete type. 6. Choice of questions. The form of questioning very mate- rially affects >S"s deposition, particularly if the questions are of the type known as 'leading' or 'suggestive' questions. To some extent any question is suggestive, in so far as it implies that its recipient knows something. If we follow Stern, at least six types of questions may be framed, viz. : determinative, com- pletely disjunctive, incompletely disjunctive, expectative, im- plicative, and consecutive. A determinative question is one that is introduced by a pronoun or interrogative adverb, and is the least suggestive form of question, e. g., "Wlaat color is the dog?" A completely disjunctive question is one that forces the reporter to choose between two specified alternatives, e. g., "Is there a dog in the picture?" An incompletely disiunctive question is one that offers the reporter a choice between two alternatives, but does not entirely preclude a third possibility, e. g., "Is the dog white or black?" In practise, for many re- porters, especially for children, this form is virtually completely disjunc- tive, since a certain amount of independence is demanded for the choice of the third possibility, c. g., for the answer "The dog is brown." An expectative question is one that arouses a moderately strong sug- ^The terminology of the report-experiment has developed in Germany and France. I have been obliged to coin English equivalents — a task not always easy because the foreign terms have not been chosen with special care to secure consistency or to accord with legal phraseology. For this reason, the foreign equivalents are included here and elsewhere in the discussion. 20 [386] DESCRIPTION AND REPORT gestion of the answer, e. g., "Was there not a dog in the picture?" (This is the form used by Binet to induce moderate suggestion.) An implicative question is one that assumes or at least implies the presence of a feature that was not really present in the experience, e. g., "What color is the catV" In practise, it is clear that a determinative question might become implicative if the reporter had completely for- gotten the item to which it referred. (The implicative question was used by Biuet to induce strong suggestion. ) The consecutive question is any form of question that is used to augment a suggeston that has been developed by previous questions. 7. Choice of metliod of grading. Treatment of data. In general; the adequacy of a report depends both upon its quantity aud its quality: quantity is measured by the number of items mentioned or the number of questions answered (in absolute or in relative terms) and is referred to as the range of report {Unifang, etendne) ; quality is measured by the fidelity of the statements made, and is referred to as the accuracy of report [Treue, fidelite). We have also at our command useful indications of the posi- tiveness or degree of assurance that S places in his report. Be- sides (1) complete uncertainty ("I don't know" or "I have for- gotten"), we may distinguish (2) hesitancy (''I think" or ''I believe"), (3) positive statement or assurance of ordinary de- gree, and (4) attestation or attestable assurance, i. e., the high- est degree of assurance, as indicated by ^"s willingness to take his oath that the statement is correct. On this basis, the data may be subjected to treatment for the computation of a number of 'coefficients of report,' by the aid of the following simple formulas : Coefficients op Eeport* Let P = number of possible items, n = number of items reported (or replies made), c = number of items reported with certainty (in- cluding attestation), *The fourth formula is used by many writers, in place of the fifth, for accuracy of report ; as here indicated, however, the indeterminate cases ("I don't know") are omitted from the denominator in computing accu- racy. Next to range and accuracy, the most important coefficient is probably warranted assurance (8th formula), as a high ratio indicates a good wit- ness, Avho reports a large number of items both correctly and with assur- ance. TEST 32 : FIDELITY OF REPORT [387] 21 a = number of items whose correctness is attested under oath, n[N)=-- number of items re^^orted in the narrative, n{D)— number of items reported in the deposition, nir)= number of items that are rightly reported, c(r)'= number of items that are certain and right, a{r)'— number of items that are attested and right, a{tv)= number of items that are attested and wrong Then (1) n (2) n/P -- range of report, absolute (C77?«ffl«(7;, etendue'} range of report, relative, (3) n(N)/ii{D)=: spontaneity of report, (4) n (r) / u = range of knowledge (Urn fang des Wissens, etendue du savoir) , (5) n{r)/G--= accuracy of report {Ti'eue, fidelite), (C) c/n=i assurance (suhjective Siclierheit^ assurance) , (7) c(r)/c^ reliability of assurance {Zuverldssigkeit der S i c her he i t, Sicherheitsherechtigung, fidelite de la certitude) , (8) c(r)/H— warranted assurance [SicherheitJler Person, assurance justifiee), (9) ■c(r)/M(r)= assured accuracy {Versicherte Richtigkeit, justesse certifiee), (10) a/n == tendency to oath or attestable assurance {tendance an serment) , (11) .a (r) /n = Avarranted tendency to oath {tendance au ser- ment veridique) , (12) ff(«7)/7i= unwarranted tendency to oath {tendance au fa ux-tem oignage ) , (13) a{r)/a= reliability of oath {fidelite du serment). The cleterniination of P, and hence of relative range of report, is often beset with clifficnlty ; the most practical working rule is to rank as 'one item' any combination of features that forms a single natural working group, the details of which would escape individual observation under ordinary conditions. Again, P may be taken as the number of separate items mentioned by a competent S in describing the picture or test-object by direct observation. Or, as Hegge (18) proposes, P may be computed by adding all the specific items mentioned in the reports of any one of a nuniber of .S"s. Obviously, the magnitude of P will tend to increase with the number of (S's imtil a point is reached beyond which additional reports fail to affect it appreciably. 22 [388] DESCRIPTION AND REPORT A similar difficulty arises in deciding what items and how many should be the subject of questions in the interrogatory. In general, the coeffi- cients computed will have value only for a given picture or event and oidy when obtained by a given interrogatory, and the interrogatory must be constructed empirically, on the basis of actual preliminary trials, never a priori. Although different errors unquestionably have different degrees of im- portance (to forget a man is more serious than to forget the color of his necktie), no satisfactory plan for arbitrarily 'weighting' different items has been devised. The psychologically best method of grading is unquestionably to classify the data statistically according to various categories — such as persons, objects, colors, sizes, etc. — and to compute range, accuracy, assurance and the S' in a good light. At the end of 30 sec, remove it and keep it well concealed. Direct ^8' at once : "Now tell me everything you saw : describe it so clearly that if I had never seen the card 1 should know all about what was on it." The narrative is given orally by 8, and recorded verbatim by E, without comment, query, or suggestion. Reread the report to S, and ask him to indicate what statements he is so sure of ^These objects are not exact duplicates of the Binet group, and the card is somewhat larger. The exposure-time and the questions of the inter- rogatory have been correspondingly modified. TEST 32: FIDELITY OF REPORT [389] 23 that he would swear to their accuracy. Underline these state- ments. Proceed next with the interrogatory. If possible, ask S the following questions in the order given."' Record his replies by number, verbatim, and underline all attested replies. Interrogatory for the card of objects. (1) Did you notice a com. ^ (2) What kind of a coin is it? (What denomination?) (3) Does it show 'heads' or 'tails?' (4) Is it bright or dull? (5) Is it in good condition, or scratched and marred? (6) What is engraved on it ? (What does it say?) (7) How is it fastened to the cardboard? (8) Did you notice a button f (9) What is its shape? (10) What is its color? (11) Is it the same color all over? (12) Is it made of cloth or of some other substance? (13) How many holes are there in it? (14) How is it fastened to the cardboard ? (15) Did 3'ou notice a small picture {print) near the top of the cardboard? (10) What shape is it? (17) What does it represent ? (What is it about?) (18) How many persons are there in it? (19) What is the lady doing with her right hand? '•8 may intei-fere with this program, either by anticipating the answers to some questions, or by committing errors, c. g., describing an essentially different scene in the larger photograph ; in such an event, E must devise other questions to follow up the cues thus given. Thus, if to Question 14, 8 replies "By a thread," ask further questions, e. g.. "Do the threads pass through the holes or around the whole button?" "Draw them." "What color are they," etc : It Is probably better to question S concerning objects that he fails to mention in his narrative, save that, naturally, if the first question in each group, "Did you notice ?" Is answered negatively, the remaining questions about that object are omitted. Many children fail spontane- ously to recall one or more objects, but can nevertheless answer correctly questions about them, once the object is suggested. 24 [390] DESCRIPTION AND REPORT (20) What is the other person doing? (21) Where is he sitting? (22) AVhat is he looking at? Describe it exactly. (23) Is the name of the picture printed on it? (24) Bid you notice another picture f {A photograph?) (25) What shape is it? (26) What does it represent? (What is it about?) (27) How many persons are there in it? (28) How are they dressed? (29) Where are they standing? (30) How man}^ animals are there in the picture? (31) Is the cart on wheels or not? (32) Are there any words printed in the picture? What are they? (33) What did you see in the background (in the back of the picture?) (34) What did you see in the foreground (in the front of Ihe picture?) (35) Is the picture taken in summer or winter? How do 30U know ? (30) Did you notice a stamp f (37) Is it American or foreign? (38) How much is it worth? (What denomination?) (39) What color is it? (40) Wliat is on it? (What picture or printing is on it?) (41) On what part of the cardboard is it? (42) Is it a new one or has it been used? (Describe the can- cellation mark.) (43) Did you notice a ladel (sticker, paster?) (44) What color is it? (45) What shape is it? (Is it perfectly rectangular? Draw it.) (46) Is there any printing on it ? What? (47) Is there any border around the printing? (48) How is it fastened to the cardboard? (49) How is it placed on the cardboard — right-side up, slant- ing, or how? (50) W^hat color is the cardboard? TEST 32: FIDELITY OF REPORT [391] 25 Variations of Method. — ( 1 ) To shorten the experiment, omit the narrative and take only the deposition, but first ask >S' to name the objects seen. Kecord the number. (2) Mature /5"s may be tested in small groups, though this is not recommended. Both narrative and deposition must then be written by the S'H. For comparative purposes, the same pro- cedure must be followed for all >8"s, since oral and written re- ports cannot be assumed to be equivalent. (3) To induce a moderate degree of suggestion, E may recast the questions of the above interrogator}^ into an expectative form and add others, e. g., in place of No. 14 : ''Is not the button fastened to the cardboard by a thread?" In place of No. 30: "Isn't there a little dog besides the horse?" In place of No. 42 : "Isn't the postage-stamp cancelled ?" Or, for additions : "Isn't there a seventh object on the cardboard ?" "Draw it." "Are there not four wheels on the cart?" etc. (4) To induce a strong degree of suggestion, E may recast the questions given into an implicative form, and add others as desired, e. g., in place of No. 9 : "Draw the button so as to shoAV the place where it is broken." In place of 30 : "Are both horses of the same color?" In place of 42 : "Describe the cancellation- mark on the stamp." In addition to 40 : "What else does the label have on it besides 'Glass. Handle with care.'?" Or, in place of 21 : "Is the little boy's mother putting her arm around him as he sits in her lap?" For additional questions, devise a number such as: "Is the lady's necktie dark brown or blue?" etc. Results. — (1) With regard to the number of objects spon- taneously recalled, the following results indicate the outcome found by Binet" with 23 children 9 to 12 years of age and by an experimenter from the author's laboratory (12b) with 34 school children in the fourth grade of an Ithaca (N. Y.) public school : Number of objects 6 5 4 3 Average. French children 4 1(» 8 1 4.78 Ithaca children t4 10 6 4 5.00 Counting 1 for each right answer and 0..5 for each partly right 'For a detailed presentation of these results, see his book (3, nn. 255- 329). 26 [392J DESCRIPTION AND REPORT answer, the Ithaca children scored in their depositions (possible score = 50) an average of 30.5, with a maximum of 43.5 and a minimum of 6.5. (2) Bearing in mind that the actual objects differed some- what, the reader may compare Binet's results and our own with regard to the order and frequency of omission in the narrative : Name of object stamp label button coin small picture large picture French cbiiclren 10 9 4 3 2 Ithaca children 8 6 6 7 3 4 (3) In tests of older children with written narratives, Binet found little difference in the total number of objects mentioned, but marked differences in the wealth of details and the precision of their formulation. (4) The objects have distinct individualit}^, i. e., though ^ may forget the color or the value of the stamp, yet if he recalls the object at all, it is as a stamp, not, for instance, as "some square, greenish-colored thing." In other words, 8 recalls a thing, not a number of meaningless attributes.^ (5) >8"s may report very precisely and with assurance objects or features of objects Avhich are totally incorrect, e. g., they may draw the thread fastening the button, and take oath as to its presence. Hence, testimony given with precision and detail and with the highest degree of assurance maj^ be absolutely false:, (6) aS^'s may recall one feature of an object exactly, but rail entirely in their description of another feature of the same ob- ject, e. g., recall that the label is red, but err as to its shape. It follows that, in testimonj^, a witness whose assertions are veri- fied in many details may, nevertheless, err in his statements with regard to some other detail that happens not to be susceptible of verification. (7) If >S* fails to mention an object in his narrative, but re- calls it immediately in the interrogatory, his further character- ization of it may be quite as accurate as that of other /8"s who had recalled it spontaneously. "In the aiithor's study of range of Aasual apprehension, however, there appeared numerous cases of the character thus denied by Binet, for exam- ple, a nickel was recalled only as "something bright and round in the upper corner of the cardboard," TEST 32 : FIDELITY OF REPORT [393] 27 (8) III comparing different types of questions, Binet fonnd 26 per cent, error for indiiferent, 38 per cent, for moderately suggestive, and 61 per cent, error for strongly suggestive ques- tions. B. REPORT-TEST WITH A COLORED PICTURE Materials. — Set of four colored pictures : "Australians," "A Disputed Case," "AVasliington and Sally," and 'The Orphan's Prayer."^' Watch. Method, — Give S instructions analogous to those in the pre- ceding form of report-test, but without specifying the time of exposure. Expose the picture for 20 sec. Secure an oral narra- tive and deposition as directed above. Suggestions for interrog- atories for two of the pictures follow. Interrogatory for "A Disputed Case.'" (1) How wide is the picture (horizontally) ? (2) How high is the picture (vertically) ? (3) Is there any border : if so, what color? (1) HoAV many persons are there in the picture? Take tlie person on your right : (5) Is he young, middle-aged, or old? (6) What is his posture, — sitting, standing, or lying down? (7) What is he doing? (8) What is his facial expression? (1)) Is he bald or has he abundant hair? ( 10) What color is his hair? (11) Is he smooth-faced or has he a moustache or a beard? (12) AVhat color is his beard? "All four pictures may be procured through C. 11. Stoelting Co., Chicago, 111. The "Australians" is a large lithograph, one of a series called Lente- niann's Types of Nations, catalogued by E. Steiger & Co., New York City. It is recommended for use with large groups, numbering from 10 to 50 or inoi-e *S"s. The "Hindoos" lithograph prescribed in Test 31 may be used \Yith this for cheek tests, as it is of the same dimensions and of similar character. The "Disputed Case" (No. 1235 of the Taber-Prang Art Co.'s collection) is recommended for use save for very young children or for large groups. "Washington and Sally" and "The Orphan's Prayer" (Nos. 699 and 1207, respectively, of the same collection) may be used for subsidiary and check tests. 28 [384] DESCRIPTION AND REPORT (13 (14 (15 (16 (17 (18 (19 (20 (21 (22 (23 (24 Does his moustaclie conceal his mouth? l>oes he wear eye-glasses or spectacles? Has he a hat on? What kind? What color? Where is his right hand? Where is his left hand? What color is his coat ? What color is his shirt? Has he a collar on ? AVhat color is his necktie? "^Miat color is his vest ? What color are his trousers? Does he wear slippers or shoes or boots? Take the person on your left:^° (25-44) Eepeat questions 5-24. (45 (46 (47 (48 (49 (50 (51 (52 (53 (54 (55 (56 (57 (58 (59 (60 (61 (62 (63 (64 (65 What kind of light or lamp is used? Where is it placed? Where is the ink-well? Is there not a pen in it ? What color is the dog? Is there a table or bench? How long is it (really) ? What color is the table cloth or covering? Is the fringe of the same or of a different color? Kame the objects on the table. How many chairs are there in the room? Is the rocking chair on your left or on your right? Is there an umbrella? Do you think it is jet-black or dark-blue? In Avhat position is it? Name the objects in front of the table on the floor. Is there a satchel or dress-suit case in the room ? Which ? Is it open or shut? What do the pictures on the wall represent ? How many windows are visible? Can vou see any detail of outdoor scenerv through them? "If it is desii-ed to economize tiiiie, omit questions 25 to 44. TEST 32 : FIDELITY OF REPORT [395] 29 How many hats are there in the room? Describe and locate them. Can you recall the time indicated by the clock on the wall? What object is on your extreme right? Are there any books in this part of the room ? What color is the wall ? Where is the newspaper? How long did you see the picture? Interrogatory for the "^Australians." How many persons are there in the picture? How many animals? What kind of animals? What is the person on your left doing? What is the object behind him? What is the person in the middle of the picture doing? Has this person a beard or not ? Is the man who is in charge of the dog holding him by a leash (guiding rope) or by taking hold directly of the scruff of his neck? What are the persons in the background doing? Do the persons in the foreground Avear anything beside the loin-cloth? What color is their skin? AV^hat color is the dog? AVhat is the most peculiar thing that you noted in the appearance of the men in the picture? What objects lie in the immediate foreground? Is there any Avater represented in the picture? Is the white man standing on the left or on the right? Is the sun represented in the picture as shining from your right or from your left ? How do you knoAA' ? (18) How long did you see the picture? Variations of IMethod. — Test the effect of varying the time of exposure, of extending the time-interval between exposure and report, of repeating the report (narrative or interrogatory), without further exposure, two or more times at intervals of sev- 30 [396] DESCRIPTION AND REPORT eral days or weeks/^ of confronting S with, the picture for care- ful criticisms of the report he has submitted. Though it is not advised as the best metliod, the substitution of written for oral narratives and depositions will permit an instructive class ex- periment. Typical Results. — The following narrative by a college senior, a man of varied experience, mature, much traveled, and well trained, though of mediocre native ability, shows clearly the tendency of an adult 8 to describe a situation, a meaningful whole, rather than merely to enumerate details, as do many chil- dren. Indeed, the detail here is distinctly subordinated to the interpretative rendering. The narrative tells what the picture is about, rather than what it is. "The picture, about 10x10 inches, represents a scene that ^YOuld be typical of a rural justice of the peace and a man who has come to ask his advice on some subject. The Justice sits before his desk, an old manu- script before him, one hand on his head as if he had not yet given his decision. The office is filled with books and on one of them in the left of the pictiu'e rests his top-hat. The visitor seems to be troubled very much ; his clothing denotes that he is of a different station in life. He has placed his carpet-bag on the floor and his hat near it, as a sign of great mental strain, which seems to increase as he awaits the decision. On the wall to the right is a double map of the world, showing, perhaps, that the Justice is a man of wisdom and a source of information to his neighbors. The room, furniture, the manner of dress would have denoted a time long be- fore ours. The men seem to be about 65 or 70 years of age." In his deposition, this student rendered an unusually full list of answers : the reply — "I don't know" — is given only twice (Questions 34 and 72). The range of report is, therefore, large, but the fidelity is relatively small, since all the statements that follow are erroneous ones from his report (those italicized are also attested statements) : The picture is 14x14 inches. The man on the right is MM, wears spec- tacles, lias Ms riglit hand on a paper, wears a collar, a purple tie, black trousers, and slippers. The man on the left is thinking hard, has a troubled expression, wears a sandy moustache ; he has his right hand in his pocket, his left on his Icnec ; he wears a light-colored vest and brown trousers. The room is lighted by a candle which stands on the pile of books. There is a pen in the ink-well. The table is 14 feet long, has a light-colored cloth top u-ifli fringe of a different color. There are three chairs in the room, the rocker being at the left. The unihrella is dark blue in color, and lies on the floor. There is a coat on the floor in front of the table; there is a basket on the table. The satchel is shut. One loindow is "See Ref, 34 for further suggestions. TEST 32 : FIDELITY OF REPORT [397] 31 visiNe. There is a chair at the extreme right of the picture. The wall is white. (The cuspidor aud the newspaper are not recalled.) General Kesults op Tests of Report. — (1) Accuracy. The cliief single result of the Aussage psychology is that an errorless report is not the rule, but the exception, even when the report is made by a competent /^ under favorable conditions. Thus, in 240 reports. Miss Borst found only 2 per cent, errorless narratives and 0.5 per cent, errorless depositions. These errorless reports are commonly characterized by very small range, i. e., they are reports of ;S"s who are extremely cautious and state only what they are certain of. For certain tj-pes of material, particularly estimates of time, space, number, etc., not only are erroneous reports the rule, l3ut the most common single answer is more likely wrong than right (Dauber). The average S, when no suggestive questions are employed, exhibits a coefficient of accuracy of approximately 75 per cent. (2) Range and accuracy. There is no general relation of range to accuracy, though, for a given 8, it is doubtless true tliat tliere is an inverse relation between these two coefficients.^^ (3) Range and other constants. There is no general paral- lelism between range of report and other coefficients which de- 4)end upon degree of assurance. TABLE 50 Comparative Accuracy of Sicorn and Vusirorn Statements (Stern and Borst) STERN STERN STERN BORST Range Errors Range Errors Range Errors Range Errors Positive statements— Sworn statements Unsworn statements-.. Certain statements Uncertain statements- (100) 76 2-1 13.6 11 20 (100) 68 32 19 7 (100) 70 30 23 14 (100) 60 40 97.5 2.5 11.0 8.2 15.5 10.1 44.0 Note. — All figures are in per cents. The results, save those of the third and fourth columns, refer to narratives, not depositions. '-The reason for this lack of general relation bet\A-een range and accu- racy is presumably that there are two kinds of good witnesses — the one possesses good capacity of observation, recall and report, and hence exhib- its a large range and a high degree of accuracy ; the other is cautious, and therefore restricts his range, which may be poor at best. 32 [398] DESCRIPTION AND REPORT (4) Accuracy and attestation. Generally speaking, attesta- tion does not guarantee accuracj'; on the contrary, though the number of errors is nearly twice as great in unsworn as in sworn testimony (according to Stern, 1,82 times, according to Borst, 1.89 times as great), there still remains as high as 10 per cent, error in sworn testimony. These relations are shown clearly in Table 50. (5) Dependence on SGX. In all of Stern's work, both in narra- tives and depositions, with pictures, or events, or estimations of times and distances, whether under oath or not, the reports of men have been more accurate (by from 20 to 33 per cent.), though less extended, than those of women, and a similar sex- difference has appeared in some tests of school children. This superior accuracy of boys becomes more evident when the report is difficult to make. Stern's conclusions, however, have not been confirmed by Wreschner, Breukink, or Miss Borst. Wreschner found that among adults women did better than men. Breukink found that men students reported slightly more than women, but with less accuracy, especialh^ when colors were concerned. His men, however, proved more resistant to suggestive questions. Miss Borst. similarly, declares women to be superior to men, but an inspection of her results shows that the superiority of women consisted in the fact that they returned a larger number of cor- rect statements, while the men did not make less accurate state- ments in their more limited reports. A recent and as yet unpub- lished investigation conducted by Boring (6) in the author's laboratory, in which groups of boys and girls and of men and women reported upon the events displayed in a moving picture leads to the conclusion that relatively little sex-difference exists between boys and girls (with a tendencj^ in favor of the boj's), Avhereas a quite marked and certain superiority^ of men over women exists among adult *S"s. More specifically, Borst found that in the narrative the range of men was 70 per cent., and in the deposition 83 per cent, of the range of women, while the accuracy of men in both forms of report was approximately 96 per cent, of the accuracy of women. There is a similar discrepancy between Stern and Borst with regard to the tendency to attestation ; the former found that men swore to 71 per cent, and women to 85 per cent, of their report, whereas the latter found TEST 32 : FIDELITY OP REPORT [399] 33 that men swore to Gl per ceut. and women to but 59 per cent, of their report. Boring found evidence that boys tend to exceed girls in range of report, tendency to oath and unwarranted tendency to oath, while girls undoubt- edly exceed boys in reliability of oath. With adults, men apparently ex- ceed women in range of report, and they undoubtedly exceed them in range of knowledge, assurance, warranted assurance, assured accuracy and reliability of oath. Women possess a very decidedly greater unwar- ranted tendency to oath. The fact that sex-differences in report are more pronounced in'^adults than in children accords with what we know of sex-differences in general. (6) Dependence on age. Most experimenters conclude that the reports of children are in every way inferior to those of adults, that their range is smaller, their inaccuracy greater, and their warranted assurance and reliability of assurance much lower because their assurance is too great. Stern concludes that during the ages 7 to 18 the range, especially the range of knowl- edge, increases as much "as 50 per cent., but the accuracy, save in the deposition, does not increase as rapidly (20 per cent.). This development of capacity to report is not continuous, but . characterized by rapid modification at the age of puberty. Nearly all experimenters have commented upon the excessive suggestibility of children before the age of puberty. Cohn and Dieffenbacher detected improvement in fidelity up to 15 years in boys, but up to 20 in girls. Stern has endeavored to analyze in part the development of the child's capacity to report, and has distinguished four stages: (1) the very young child enumerates only isolated objects or persons (Binet's enumerator type) ; (2) at about the eighth year, actions are reported more carefully; (3) during the years 9-10, attention is for the first time paid to spatial, temporal and causal relations; (4) in a still later period there appears the capacity to make a qualitative analysis of the constituent features of the objects reported. Cohn and Dieffenbacher think that there should perhaps be added a fifth period, from 16 years on, when the report shows evidence of reflec-tive and interpretative consideration. The question as to whether the testimony of children is so imperfect as to warrant absolute exclusion from court proceedings has given rise to much discussion. Thus, Baginsky, the German specialist in chil- dren's diseases, declares that children are the most dangerous of all wit- nesses and demands that their testimony be excluded wherever possible. Gross, the leading German authority on criminal law and criminal psy- .chology, however, asserts that a healthy half-grown boy is the best pos- sible witness for simple events, that children make different errors, but no worse ones than do adults, while, in respect to freedom from prejudice, erroneous interpretation, emotion, intoxication and the like, a child is better fitted than an adult to give an accurate report. Lipmann contends, quite on the contrary, that the unreliability of chil- 34 [400] DESCRIPTION AND REPORT tlreu's testimony is due in part to an uncritical filling out of gaps in memory, and in part to an unskillful distribution of the attention (though the child's attention is well enough concentrated on what he does report). Heindl says that children are perfectly good observers, perhaps even more objective than adults, but that they cannot translate their observations into verbal reports skillfully. The work of Boring Avas specially directed toward this controversy. He found men superior to boys in all coeJflcients, save assurance, assured accuracy and tendency to oath, in which there was no decided difference. Women exceed girls unquestionably in both range and spontaneity of report ; women display a greater tendency to oath and a greater war- ranted tendency to oath, but they also display a greater unwarranted tendency to oath and a lesser reliability of oath, L e., they seem to be less cautious than girls. (7) Dependence on intelligence. There is no conclusive evi- dence upon the relation between good report and general intelli- gence. Winteler found no difference in range of knowledge and lidelity of report between the three most intelligent and the four least intelligent in his classes of 10-year-old boys. (8) Dependence on social status. That intelligence may, however, play a positive role is suggested by the conclusions of Breukink that physicians, professors and teachers give more extended and more accurate reports than nurses and laboring men, and that the cultured group is much less open to suggestion than the uncultured and much less liable to take oath to their answers to suggestive questions. (9) Defectives. The reports of defectives, paralytics, epilep- tics, the insane, etc., show, as one might expect, a xerj high de- gree of inaccuracy, even when the pathological condition is not seriously developed. Such persons are also highly suggestible (de Placzek). Dupree points out that the reports of such per- sons are peculiarly dangerous when their deficiency is latent or concealed. Gregor found that paralytics were not very bad re- porters when the conditions were all favorable, but that they fell olf decidedly under less favorable conditions — long time- interval, suggestion, etc. (10) Dependence on time-interval. Lengthening the inter- val between experience and report tends, on the whole, to reduce range and accuracy, but there is nothing like the loss in efficiency shown in typical curves of forgetting for nonsense syllables and similar material; indeed, for some ^'s the report may be TEST 32 : FIDELITY OF REPORT [401] 35 improved in some respects after several days have elapsed. I)al- lenbacli's figures (Table 51) may be taken as fairly typical. TABLE 51 Effect of Time-Interval on Report {Dallenhach) NARRATIVE DEPOSITION INTERVAL Items Recalled Per Cent. Error Questions Answered Per Cent. Error 765 735 750 569 10.5 14.3 18.0 22.4 880 855 854 801 14.1 5 Days 18.2 15 Days 20.7 45 Days 22.4 From his earlier tests, Stern computed a fairly constant decrease of accuracy with time, amounting, on the average, to a loss of 0.33 per cent, per day over the period of three weeks which he studied ; siinilarlj', Borst computed a decrease in accuracy of 0.27 per cent, per day during a period of six days. x^hough range and accuracy seem thus to suffer with the lapse of time, /assurance, as shown bj' the number of certain and attested statements, lis not, it seems, equally affected, but shows either a surprising constancy, '^•, if anything, a tendency to increase. I'rom this it may be concluded that assurance and tendency to oath are due to S"s 'personal equation' (rather than to the freshness of his memory. It would follow, of course. That warranted assurance and warranted tendency to oath decline with the lapse of time. In explaining the improvement found in some reports after lapse of time, Schultz contends that perseveration is one of the disturbing factors in reports made shortly after the experience ; in so far as perseveration is a tendency that weakens with time, there would thus be less inaccuracy from this source of error in later reports. Jaffa asserts, more positively, that narration directly after an event by no means gives the best result ; rather the memory of the event is organ- ized and consolidated several weeks later and then affords a far more faithful picture of the event than an account after a brief interval. It seems doubtful, however, whether such a view can be accepted as a gen- eralization, however true it may be under some conditions. That the lapse of time occasions various and complex modifications is also indicated by the work of Cohn and Dieffenbacher, who compared direct descriptions of one colored picture (Test 31) with narratives and depositions upon another colored picture with an 8-day interval between presentation and report. Here, while there was a positive correlation between range of description and range of narration, there appeared dis- tinct differences in the nature of the two accounts, e. (/.. acts and interpre- tations are more prominent in reports than in descriptions. (11) Dependence on form of report. All authorities agree that Ihe use of the interrogatory, whether <:)f the complete or in- 36 [402] DESCRIPTION AND REPORT complete form, increases the range and decreases the accuracy of the report.l Thus, in comparison with the narrative, the range of the inteirogatory may be 50 per cent, greater, while the inac- curacy (of tlie inconijilete interrogatory) may be as much as 550 per cent, greater. In general terms we may say that about one- tenth of the narrative is inexact, but about one-quarter of the deposition. Typical statistics are given in Table 52. Colin and Dieffenbacher believe that reliability should always be compaled from the narrative and deposition combined, because onh' thus can differences in the two forms of report l)e eliminated. TABLE 52 Dependence of Report on its Form (Stern and Borst) RANGE ACCURACY Narrative Deposition Narrative Deposition Stern 25.5 40.5 52.1 65.6 Per cent. 94 89 Per cent. 67.1 Borst — 83.0 Note. — In comparing these figures, it should be remembered that Stern used an incomplete, and Borst a complete interrogatory. According to Ereukink, the use of written instead of oral re- ports apparently tends to increase the number of indefinite an- vswers, but to decrease the number of erroneous answers. (12) Dependence on the type of question. The work of Stern, Lipmann, Kinet and others shows that the introduction of leading or suggestive questions decidedly decreases the accu- racy of report in children and may affect seriously the testimony of uncultured adults, or even of competent adults unless the con- ditions are favorable. Stern (33) estimates 50 per cent, error for 7-year-old children and 20 per cent, error for 18-year-old >8"s in replies to suggestive questions. Most experimenters have found women less resistant to suggestive questions than men. Cohn and Dieffenbacher find relatively slight differences in the suggestibility of boys and girls, though the boys tend, on the whole, to take a somewhat more critical attitude. They find that TEST 32: FIDELITY OF REPORT [403] 37 the decrease in suggestibility with age is more marked in girls than in boys, and that greater suggestibility in dull as compared with bright pupils is evident in girls, but not in boys. These investigators call attention to the fact that a suggestive question that is introduced too abruptly (so that its very form attracts attention) is apt to arouse immediate resistance. (13) Dependence on contents or features. Not all the fea- tures of the original experience are reported with the same fre- quency or with the same accuracy. In general, we may say that persons and their acts, objects, things and spatial relations are reported with considerable accuracy (85-90 per cent.), whereas secondary features, especially quantities and colors, are reported with considerable inaccuracy (reports on color have an error of from 40 to 50 per cent.) In his subsidiary test with geometrical forms of different shapes, sizes and colors, Dallenbach found the errors most frequent with color, next with position, next with size, and least Avith shape, and this regardless of time-interval. Of the colors, errors were most frequent with green and least frequent with yellow tone. On the reliability of different classes of /S's with respect to different features, see further the tables of Cohn and Dieft'enbacher (11a, pp. 86f.). (14) Dependence on the ideational type of the reporter. The best reports are given by observers of a mixed ideational type, e. g., acoustic-motor or visual-motor (Borst) : even in a picture- test, the purely visual-minded observer is inferior, though less open to suggestion (Lobsien). A characteristic analysis of reports, for the piu'pose of classifying re- porters Into ideational types has been given in the description-of-an-object test (No. .31), in "which Binet distingnishes four types of reporter — the obsei'ver, the describer. the emotionally-minded, and the erudite. Miss Borst was unable to use this classifieation, however, with her »S''s. Another classification of reporters according to mental type was at- tempted by Miss Borst, who, after a preliminary tachistoscopic test, com- pared the reports of 'fixating' and 'fluctuating' iS"s, and concluded that -S"s whose attention is of the 'fixating' type have uniformly the greater war- ranted assurance of report. There was no relationship found with extent of report. (15) Qualitative analysis of errors. Stern finds four kiuds of errors in the narrative: (a) errors of apprehension (observa- . tion), like overlooking, misapprehending, underestimating, over- 38 [404] DESCRIPTION AND REPORT estimating, etc. ; {h ) real errors of mernorv, like forgetting, fill- ing in of gaps, gradual amplification, etc.; (c) errors of imagi- nation, 'retouching' the recollection, unintentional blending of imagined experiences with the one reported, or the harmless 'playing' with the report [Fahulieren) often seen in children, and (d) errors of judgment (will), like lack of caution or self- criticism. Schultz (29) has also attempted a qualitative analysis of the material gathered by Aall in an event test. The following are the main points upon which stress is laid : (a) Whether an item is reported depends both upon the mental state at the moment and also upon the objective complex in which the item occurs. Attention is attracted by novelty and by the logical significance of the impression. (&) But there is a certain 'spread' of attention such that details that are trivial and accessory may also be included with those that yS Is aiming to observe. (c) The novel attracts attention, but it is also difficult to observe cor- rectly. Optimal conditions are given when a familiar thing (easy to understand) is in an unfamiliar setting (motive of novelty). (d) An event which suddenly breaks into consciousness and disturbs the set of the moment is a source of difficulty until a new adaptation for it is secured. (e) "Perseveration plays an important role in the mistakes of wit- nesses." Its falsifying effect decreases with time, and thus reports that are separated by a time-interval from the event may be better than imme- diately given reports. \, (f) There takes place a process of logical elaboration, the effect of which is to em])hasize the kernel of the episode and to minimize unessen- tial details (principle of conscious economy). (//) This tendency also operates to distort reports so as to make them conform to what the witness regards as the natural course of events. Por- tions of the episode uuperceived or not understood are filled out or re- arranged in accordance with this principle. Characterizations of persons especially show this tendency. (/t) JMany yS''s show a distinct tendency to embellish or roimd out their reports into good literary form, and may thus unwittingly distort their statements. (0 If the experience moves S emotionally, his reports are strongly col- ored and ]uay suffer decided modification, particularly reports upon verbal items ( quotations ) . ( /) Experiments so arranged as to cause S to believe that his report is serious and responsible (not a mere classroom test) produce a different conscious attitude and reveal the presence of new factors, both inciting and inhibitory ; in general, the effect is to augment the value of the testimony. (Jc) That a witness should be motivated by a desire to awaken a cer- tain judgment upon a case need not be an undesirable condition. TEST 32 : FIDELITY OP REPORT [405] 39 (16) The effect of repeating a report. When ^ is called upon to make his report several times, the effect of this repeti- tion is complex, for (1) it tends in part to establish in mind the items reported, whether they be true or false, and (2) it tends also to induce some departure in the later reports, because these are based more upon the memory of the verbal statements of the earlier reports than upon the original experience itself, i. e., the later reports undergo distortion on account of the flexibility of verbal expression. si:^ (17) The effect of practise: educahility. On the basis of Miss Borst's work (Table 53), St would api>ear that simple prac- TABLE 53 Effect of Practise upon Coefficients of Report (Narrative) (Borst) NUMBER OF REPORT (TESt) Range Accuracy Assurance Warranted assurance Reliability of assurance Assured accuracy Tendency to oath Warranted tendency to oath__ Unwarranted tendency to oath- Reliability of oath I II III IV 39.0 39.0 42.3 40.3 86.6 87.7 92.9 88.2 96.6 96.4 97.8 97.9 84.0 87.0 91.0 88.0 87.5 89.4 92.6 89.8 97.0 98.0 98.4 98.6 43.0 59.8 62.8 61.9 40.2 53.2 58.5 57.5 2.8 6.6 4.3 4.4 93.0 88.8 92.5 93.0 42.0 90.0 98.6 89.0 90.3 99.2 72.1 66.5 5.6 91.7 I Note. — The effect of practise in these tests is somewhat obscured by the fact that the tirst and third tests were made after a 3-day, the others after a 9-day interval. tise, without special coaching or conscious effort to improve, facilitates the report. ) In her work it will be noted that the tendency to oath and^warranted tendency to oath are both par- ticularly improved, while there is appreciable improvement in the other coefficients, save assurance and assured accuracy. On the other hand, some doubt is cast upon generalizations from Miss Borst's work by the reports made by Baade and Lipmann for the Commission of the Institute for Applied Psychology ap- pointed especially to investigate this problem of the educability 40 [406] DESCRIPTION AND REPOPvT of report. Baade shows that with regard to reports made upon verbal statements (quotations) the ^"s (196 girls, aged 12-13 years) shoAved no demonstrable improvement, either as a result of the threefold repetition of each experiment (physical labo- ratory demonstrations) or as a result of the succession of three different experiments. There was an influence of earlier upon later experiments, but this influence was sometimes favorable and sometimes unfavorable. Lipmann, who scored the estimates of duration and size, found, on the whole, some improvement in these estimates due to the succession of experiments, but only a \evy slight improvement due to the repetition of given experi- ments. Other experimenters have reported results more nearly in accord with Miss Borst's conclusions. Breukink, for instance, found that if >S"s are allowed to see the picture after reporting, the practise increases fidelity of report, especially in the deposi- tion and in resistance to suggestive questions. Again, the very interesting Metliode der Entscheidungs- tind Bestimungsfragen (questions in form of : ^'Do jon know thus and so?" and ''What is thus and so?" respectively) has led Franken to declare that such training as this method induces, causes an improved cau- tiousness in asserting positive knowledge. Other experiments by Marie Diirr-Borst (1906) indicate that improvement in the capacity of children may be best secured by appeal to zeal, interest, enthusiasm and desire for improvement, Avhereas more formal training of an intellectual type — sugges- tions for systematic observation, specific training in sense-per- ception, etc. — is much less effective. REFERENCES A. The most important single source is Stern's Beitragc zur Psychol- Ofjie der Aussage, Leipzig. 1903-6. Laclc of space forbids the Itemizing of tlae numerous titles : besides extended reviews, communications, reports of lectures, etc., this periodical contains important articles by Stern, .Jaffa, Cramer, Lobsien, Lipmann, Borst, Bogdanoff, Rodenwaldt, Oppenheim, Kosog. Wendriner, Giinthei", Gottschalk, and others. B. The following are other important references. See especially Nos. 11, 21, 32, .34 and 35 for bibliographies and general reviews. The new literature is summarized annually in PftBii. (1) W. Baade, Aussage tiber physikalische Demonstrationen. (Mit besonderer Beriicksichtigung der Frage der Erziehbarkeit der Aussage.) TEST 32 : FIDELITY OP REPORT [407] 41 1 Abli. Die Metliodik cler Yersuclie imd die lulialte der Textaussagen. ZAiKjFs, 4 : 1911, 189-311. (2) A. Baginsky, Die Kiiideranssage vor Gericbt. Berlin, 1910. Pp. 41. (3) A. Biuet, La suggestibilite. Paris, 1900. Pp. 391. (4) A. Binet, La science clu temoignage. Aid's, 11: 1904 (1905), 128-137. (5) A. Binet. Psycbologie individnelle. La description d'uu objet. AiiPs, 3: 1S9G (1897), 29G-332. (G) E. G. Boring, Capacity to report upon moving, pictures as condi- tioned by age and sex. To appear probably in PsRev. (7) Marie Borst. Recbercbes experimentelles sur I'educabilite et la fidelite dii temoignage. ArPs{f), 3 : 1904, 233-314. (8) Marie Diirr-Borst, Die Erziebung der Aussage luid Anscbauung des Scbnlkindes. EPd, 3 : 190G, 1-30. (9) M. Borst et E. Claparede, La fidelite et I'educabilite du temoig- nage. A)x-h. des sciences pliysiques et iiatwclles, April 7, 1904. (10) H. Breukink, Ueber die Erziebbarkeit der xUissage. ZAvgPs, 3: 1909, 32-87. (11) E. Claparede, Psycbologie du temoignage. (General review.) ArPs(f), 9: 1910, 228-232. (11a) J. Cobn und J. Dieffenbacber, Untersucbungen tiber Gescb- lecbts-. Alters- und Begabungs-Unterschiede bei Schiilern. Beihefte zur ZAugPs, 2: 1911. Pp. 213. Q2) K. M. Dallenbacb, (a) Tbe relation of memory error to time- interval. PsR,2Q: 1913,323-337. (&) Tbe effect of practise upon visual apprebension in scbool cbildren. JEdPs, 5 : 1914, 321-334, 387-404. (13) J. Dauber, Die Gleicbformigkeit des psycbiscben Gescbebens und die Zeugenaussagen. FsPs, 1 : 1912, 83-131. (14) E. Duprce, Le temoignage: etude psycbologique et medico-legale. Rev. d. deux. Blondes, 55 : 1910, 343-370. (15) A. Franken, («) Ueber die Erziebbarkeit der Erinnei'rungs- aussage bei Schulkindern. ZPdPs, 12: 1911, 63-5-642. (b) Aussage- versucbe nacb der ]Metbode der Entsclieidungs- und Bestimmungsfrage bei Erwacbsenen unci Kindern. ZAngPs, 6 : 1912, 174-253. (16) li. B. Gerland, Zur Frage der Zeugeuaussage. Arkr, 39: 1910, llG-119. (17) II. Gross, Zur Frage der Zeugeuaussage. ArKr, 36: 1910, 372- 382. (18) T. Hegge, Zur Frage der Bewertung von Aussagen bei Bildver- sucben. ZAiigPs, 6 : 1912, 51-59. (19) R. Heindl, Die ZuverUissigkeit von Signalamentsaussage bei Scbulkindern. ArKr, 43 : 1909, 109-132. (20) S. Jaffa, Ein psycbologiscbes Experiment im kriminalistisehen Seminar der Uuiversitat Berlin. Beitrdge zur Psych, der Aussage, 1 : 1903, 79-99. (21) O. Lipmann, Neuere Arbeiten zur Psycbologie der Aussage: Sammelbericbt. JPsN, 3 : 1904, 245-2-19. f22) O. Lipmann, Die Wirlaing der Suggestivfragen. ZPdPs,S: 1900, 89-90. (23) O. Lipmann, Die Wirkung v. Suggestivfragen. ZAngPs, 1: 1907- 1908, 44-92, 382-415, .504-546 ; 2 : 1908, 198-242. Also publisbed separately, Leipzig, 1909. (24) O. Lipmann, Pedagogical psycbology of report. JEdPs, 2: 1911, 253-261. (25) M. Lobsien, Ueber Psycb. der. Aussage. ZPdPs, 6: 1904, 161-209. 42 [408] DESCRIPTION AND REPORT (2G) L. Maurei", Beobaohtungeu iiber das Anschauungsvermogen cler Kinder. ZPclPs, 5 : 1903, 62-85. (27) II. Miinsterberg, On the witness stand. N. Y., 1908. Pp. 265. (The several sections of tliis boolv have also appeared in magazine form, chiefly iu McChn-e's Magazine.) (28) de Placzek, Experimentelle Untersuchungen iiber die Zeugen- aussagen Scliwachsinniger. ArJtr, 18 : 1904, 22-6.3. (29) G. Schultz, Znr Aiissagepsycliologie : Prinzipielle Ei'ortermigen im A]isc]iluss an ein zweifaches Experiment. ZAngFs, 7 : 1913, 5-17-574. (30) C. und W. Stern, Erinnerung, Aussage und Liige in der ersten Kindheit. Leipzig, 1909. Pp. 160. (31) W. Stern, Zur Psychologic der Aussage. Experimentelle Unter- suchungen iiber Erinnerungstreue. Zeils. f. d. ges. Strafrechtioissen- schaft, 22: 1902. (Also published separately, Berlin, 1902.) (32) W. Stern, (a)Literatur zm* Psj^chologie der Aussage. ZAngPs, 1: 1907-8, 429-450. {'b) Bibliographie zur Psychologie der Aussage, 1908- 1910. ZAngPs, 4: 1911, 378-381. (c) Psychologie der Aussage, 1911- 1913. ZAnyPs, 7: 1913, 577-596. (33) W. Stern, Abstracts of lectures on the psychology of testimony. AiJtJPs, 21 : 1910, 270-275. (34) G. M. Whipple, The observer as reporter: a survey of the 'psy- chology of testimony.' PsBu,6: 1909,153-170. (Also amiual summaries of literature on the topic, PsBii, 1910 on.) (35) J. II. Wigmore, Professor Miinsterberg and the psychology of evidence. Illinois Law Review, 3: Feb., 1909, 399-445 (with bibliog- raphy of 127 titles). (36) A. Wreschner, Zur Psych, d. Aussage. ArOcsPs, 1: 1905, 148-183. CHAPTER IX Tests of Association^ Learning^ and Memory A generation ago, the members of the 'English School' of psy- chologists exalted 'association' as a fundamental principle or law of mind comparable in its scope and importance with the law of gravitation in the material world. Whether this extreme position be held or not, it must be admitted that the more com- plex phases of mental activity are more readily understood if certain basic conditions of mental elaboration are posited, par- ticularly the conditions: attention, retention, and association. Disregarding the first of these, which we have already discussed, we find in retention the sine qua non of the development of human mental activity, and we find constantly at work in the conscious life of the organism a tendency for the establishment of connections between its concurrent and its successive psj'cho- physical activities. In so far as the conscious organism acquires new capacities for response, there must be retention and organ- ization, liearning, retaining, recalling, associating, these are terms obviously descriptive of a series of related activities, and on this account, tests which deal with them are here assembled. Association and memory, taken together, have undoubtedly been the occasion of more numerous and more elaborate experi- mental investigations than anj'- other phase of mental life. Learning, in the narrower sense, has, perhaps, received some- what less attention, though of late the importance of its appli- cation to pedagogical problems has stimulated work upon it. The experimental study of associative activity can be, and has been, undertaken for quite varied purposes, e. g., to examine the time relations of mental phenomena, to study individual differ- ences in thought-processes, as conditioned by age, sex, training, physical condition, and the like, to analyze the diurnal curve of psychophysical efficiency (as in Kraepelin's use of computa- tion), to diagnose mental content, and even to reveal obscure [409] 43 44 [410] ASSOCIATION^ LEARNING AND MEMORY mental tendencies and motives or intentionally' withheld infor- mation (diagnostic association tests). Space forbids the ex- I>loitation of all the tests that have been developed in these fields, but a study of the more common tests of learning, associa- tion and memory that have been selected for treatment here as being most applicable to the experimental study of school chil- dren will serve to indicate the lines along which variant meth- ods may be developed and employed. The earlier tests in this chapter investigate the nature and efficiency of those associative connections that the subject has already established at the time of the test, either when the asso- ciative processes are allowed free rein (uncontrolled associa- tion) or when they are placed under certain restrictions (con- trolled association). The tests of learning that follow investi- gate the subject's capacity to establish new associative connec- tions, under relatively novel conditions. The memory tests, in a soinew^hat different way, investigate his retentive capacity or his ability to reproduce a series of symbols or a series of related ideas. The classification of tests of association, learning and memory is, of course, somewhat rough; it is difficult to draw sharp distinctions between each type or to delimit precisely the mental processes that are brought into operation, as is illus- trated, for example, in the obvious overlapping of tests of memory, of memory-span, of report, of range of attention and range of apprehension. TEST 33 Uncontrolled association — continuous method. — The essence of this test is the requirement to write or pronounce an extended series of words not in the form of sentences. Our interest lies, first, in the difference of facility exhibited by different ^'s in the production of such a series of terms; secondly, in the nature of the terms given by S's of different sex, age, or social condition ; and thirdly, in the nature of the mental processes underlying the, word-naming process. Cattell and Bryant (4) make brief mention of the test; Jas- trow (6, 7), and later Miss Nevers (10), Miss Calkins (3), Miss TEST 33 : UNCONTROLLED ASSOCIATION [411] 45 Tanner (11) and Miss Manchester (8), employed it for the study of the couimnnit.y of ideas of men and women, Flournoy (5) for the study of the effect of environment, present and immediately past, upon the course of association, and Binet (1) for the study of individual differences in intellectual processes. In a modified form (test of 60 words in 3 min.) it appears in the Binet-Simon Scale (Ch. XIII). Materials. — Stop-watch. Blank forms containing numbered spaces for 100 words. [The seconds-clock.] Method.— Give oken by S. With mature >Sf's, it may be pos- sible to get nothing more than scant abbreviations for the more rapid portions of the series, but these may be filled out subse- quently. The points at which S makes distinct pauses maj" be noted on the form. At the 100th word, stop the watch and record the time. If time permits, and /S* can do so, it is advisable at once to go over his series, not only to fill out the list of terms, but also to make marginal notes of all the intermediate links and subsidiary associative ijrocesses that he can recall. Variations of Method. — (1) For group tests, E should pro- vide each (S with a blank. He may allow 3 min. for writing, and rate speed in terms of number of words written (method fol- lowed by Pyle, 10), or each S may record his own time for writing 100 words by the aid of the seconds-clock for group tests. The latter procedure is recommended rather than the former. "When aS" does the writing, the method resembles that of the users of it discussed below, but the standard method of oral naming is best. (2) Instruct aS' to keep his eyes closed during the test. This variant is to be preferred for individual testing, at least with adults ; its effect is commonly to reduce the speed of naming and to lessen the number of terms suggested by objects visible in the room where the test is administered. 46 [412] ASSOCIATION, LEARNING AND MEMORY (3) Vary the test by demanding short lists, say of 20 words each, referring to the several categories indicated in the table which follows, e. g., ''Name words pertaining to clothing." ''Name abstract terms." "Name adjectives," etc. Note the time needed for each such list. (4) E niay omit the instruction to write or to speak as rap- idly as possible, and allow 8 to work at his leisure. This method, which was followed by Miss Nevers, is perhaps more satisfactory for the subsequent qualitative report upon the series, but de- prives the test of whatever quantitative merits it possesses, be- sides tending to yield results of a distinctly different nature that are not comi)arable with those otherwise obtained. (5) When working with j'ounger >S"s, E may Avith advantage limit the length of the series. Thus, Flournoy demanded but 10 words, while Binet recorded tlie time for three series of 20 words each, and occupied the intervals in reviewing with 8 the terms of the preceding series. This method is less fatiguing, and en- ables immature S'^ to give a more satisfactory account of their associative connections, but it does not test /Sf's capacity as rigorously as the longer list. (6) E may secure a very limited measure of uniformity in the earlier portion of the series by starting all >S"s from the same word. For this, the words quich and play are recommended. Here it is of interest to observe the lines of divergence in asso- ciation taken by different ^'s. (7) Another variation is that of Flournoy, who, in addition to the word test, gave 45 >S's instructions to make 10 drawings of any sort. Treatment of Data. — In the standard form of test, ^'s speed is indicated directly by his time for naming 100 words. In the group test, it is customary, similarly, to rate ^'s speed in terms of words written in 3 min. It is not possible, however, to regard the times obtained from these two forms of the test as inter- changeable, since the second form includes writing and this, as is demonstrated below, tends, even in the case of mature /S"8, to slow the rate of performance. In so far, too, as ^'s differ in their speed of writing, this fact enters as an unavoidable disturbing factor in the group test. TEST 33 : UNCONTROLLED ASSOCIATION [4:13] 47 For qualitative comparison of the lists, E may, by inspection, supplemented by S's, explanation, catalog the words, either in the 7 categories used by Binet, or in the 25 categories used by Jastrow, Miss Nevers and Miss Manchester. Both classifications are embodied in the results below. Results. — (1) Some idea of the relation between perform- ance in the group test (words tvritten in 3 min.) and age and sew ill normal ^'s may be secured from the averages published by Pyle for a limited number of cases and under less precise in- structions than those above recommended. These results are set forth in Table 54, where it will be observed that on the whole the number of words increases Avith age year by year, and that girls at nearly ever}- age somewhat excel boys in their scores. TABLE 54 Words Written in Three Minutes by Normal CJiildren (Pyle) SEX AGE 8 9 10 11 12 13 14 15 16 17 18 ADULT Male Cases 33 60 66 66 77 'so 57 38 36 16 21 64 Male Aver 23.0 26.9 29.7 33.3 34.2 33.9 33.3 40.0 33.3 42.8 48.9 42.2 Male Av. Dev 7.5 7.6 9.0 11.4 10.9 14.6 13.2 14.8 14.6 12.3 16.6 13.8 Pern Cases 37 82 88 65 90 66 61 46 46 38 29 86 Fern Aver 23.7 31.0 32.2 36.8 36.6 38.3 39.1140.2 40.9 41.6:47.1 38.3 Fern Av. Dev 8.2 8.9jl0.8 12.1 15.416.8 12.9 13.8 14.1 14.0|13.9 13.1 (2) In tests of college students the average time for writing 100 words ranges between 5 and 6 min. Jastrow reports an average of 130 sec. for oral and 308 sec. for written lists of this length. Since writing an equal number of words from dictation took 212 sec, he concludes that about 1.14 sec. was used, on the average, in thinking the association between one word and the next. (3) Table 55, derived from Wallin's studies of mentally de- fective epileptics (12), shows that the test of uncontrolled asso- ciation (here the number of words spoken in 3 min. under cer- tain special instructions necessitated by the nature of the S's) is of some value for mental classification, since the average re- 48 [414] ASSOCIATION^, LEARNING AND MEMORY suits show a steady increase with increase in mental age (Binet-Simon diagnosis) when due allowance is made for the small number of oases tested in certain aaes. TABLE 55 Words Uttererl in Three Minutes l)y Einleptics (WaUin) Binet-Simon Age VI VII VIII IX X X] XII XIII Average Words Spoken- 16.0 25.5 21.5 33.4 43.6 51.3 59.9 65.0 (4) Inspection of the lists printed both hj Jastrow and by Binet shows that 8's follow what might be termed a series of themes: a number of terms are written, all of which cluster about a common central idea ; through one of these terms access is given to a new central idea, which in turn becomes a theme for the next series of terms. Thus, in the series liand, face, lip, chest, Icnees, calf, cow, horse, pig, etc., the transition from the parts-of-the-body theme to the animal theme is effected by the common term calf. (5) In some ^"s, the controlling theme is an auditory se- quence, which occasions long series of rimed or alliterative terms, e. g., run, pun, fun, etc., or hen, hand, head, harp, etc. (6) In this test, the most common tvords, i. e.^ those most easily got at, or those that lie, as it were, on the surface, are given first. After these are delivered, the task grows more diffi- cult; deeper and more remote-lying terms must be actively sought for. Closely related to this is the fact that, at least in the lists of younger aS"s, practically all the terms are nouns.^ This is particularly the case in the short series conducted by Binet, so -that, as he remarks, the test, as he conducted it, is virtually equivalent to a request to write 20 common nouns. (7) In vieAv of the vast number of words available, it is at first surprising to note the degree of community present in lists -Cliildrcu often interpret the instructions to mean tliat only nouns are wanted. If E stops to explain that other parts of speech are permissible, the result is sometimes more confusing yet, as they may then seek to name some of every part of speech. TEST 33 : UNCONTROLLED ASSOCIATION [415] 49 of 100 terms given by a limited number of persons. Tlius, Jas- trow found tliat in 50 lists (5000 words), only 2024 words were different, only 126G words occurred but once, while the 100 most frequent Avords made up three-tenths of the whole number. These most frequent words are, as has just been said, names of common objects : in Jastrow^'s 50 lists, the following were the most frequently used words: hook (40), horse (37), girl (35), man (34), hoy (33), tahlc (30); then follow chair^ tree, coio, paper, dress, etc., in somewhat lesser frequency (8) For the classification of the words given by 20 12-year- old pupils, Binet found seven categories adequate, viz.: [a) names of objects in the room where the test was held, (&) parts of the person or clothes, (o) objects or persons in the school, {(I) objects recalled from the home, (e) objects seen in the streets (horse, tree), (f) objects seen in fields or on country excursions, {g) unclassified nouns. Here there is no place for abstract terms, many of which were found in series given by American pupils in Jastrow's tests. Jastrow's own classification is indi- cated in Table 48, where it Avill be seen that his 25 categories are much more elaborate and extended than those employed by Binet. (9) Dependence on sex. The question as to sex difference in spontaneous trains of ideas such as are evoked in this test has been answered differently by the tests conducted at different institutions. The comparison of Wisconsin men and Wisconsin women was made by Jastrow, the 1894 test of Wellesley women by Miss Nevers and with no instruction as to speed, the 1890 test of W^ellesley women by Miss Calkins 1)ut with the same in- structions as those of Jastrow, the test of 75 men and 75 women at the University of California in 1905 by Miss Manchester after Jastrow's method. The categories of particular interest are those printed in italics. Jastrow's results in this and other tests led him to believe that "women repeat one another's words much more than the men." He found that "the class lo which women contribute most largely is that of articles of dress, one word in every eleven belonging to this class. The inference from this that dress is the predominant category of the feminine (or of the 50 [416] ASSOCIATION, LEARNING AND MEMORY privy feminine) mind is valid with proper reservations." Since the women exceed the men in the enumeration also of foods, amusements, arts, and educational matters, but fall below them in naming implements and utensils, professions, and especially in abstract terms, Jastrow concludes, ''that the feminine traits revealed in this study are an attention to the immediate sur- roundings, to the finished product, to the ornamental, the indi- vidual, and the concrete, while the masculine preference is for the more remote, the constructive, the useful, the general, and the abstract" (6; pp. 5G4-5) . Most of these conclusions are flatly opposed hj the Wellesley results of 1894, but the employment of Jastrows' methods in the 1896 test produced less marked diver- gencies. It is particularly to be noted that writing at a faster rate (1896 test) caused a marked decrease in the number of ab- stract terms, and brought the terms relating to 'interior furnish- ings' up even beyond those of the Wisconsin women ; on the other hand, the frequency of terms for 'wearing apparel' was not affected by this change in method. The three sets of 25 lists each (25 men and 25 women) pro- cured by IMiss Manchester at California show complete agree- ment with Jastrow's results in the following aspects : men lead in naming (1) verbs, (2) implements and utensils, (3) occupa- tions; women lead in naming (1) wearing apparel, (2) build- ings and building materials,- (3) interior furnishings, (4) edu- cational terms, (5) arts, and (6) amusements. Miss Manchester generalizes these differences as follows: (1) "The dynamic aspect of objects is more attractive to men, while the static or completed aspect appeals more to women." (2) "Time as a factor enters more largely into the surface ideas of men ; space is more often a prominent feature of the surface ideas of women." (3) "Men are interested in far-reaching relations ex- isting between things ; women give more attention to the minute analysis of things themselves." (4) ''The range of the surface ideas of men, as a group, is slightly greater than that of women." -In explanation of tliis seemingly unusual superiority of the women it sliould be said that the things named are not distinctive building mate- rials or operations, like mortar, cement, mortising, etc., but such common terms as floor, door, gate, church, etc. TEST 33 : UNCONTROLLED ASSOCIATION [417] 51 TABLE 56 DistriMition of Terms hi 'Uncontrolled' Association (Jastrow, Nevers, Calkins, Manchester) {Each: column represents 25 lists of 100 words each. Those from Cali- fornia are based upon 15 lists reduced to the same basis.) CATEGORIES < < m Z z mZ Z g z o og Hoo m - H Z oz fe z feS 3 H J2§ o o H O ^ ^ ^^ Hoo tn - W Z ►J s w o 1. Animal kingdom 2. Wearing apparel and f abrics- 3. Proper names 4. Verbs 5. Implements and utensils 6. Interior furnishings 7. Adjectives 8. Foods 9. Vegetable kingdom 10. Abstract terms ^ 11. Buildings and building ma- terials 12. Parts of body 13. Miscellaneous 14. Geographical and landscape features 15. Mineral kingdom 16. Meteorological and astro- nomical 17. Stationery 18. Occupations and callings— 19. Conveyances 20. Educational 21. Other parts of speech 22. Arts 23. Amusements 24. Mercantile terms 25. Kinship 254 178 214 187 146 129 224 82 118 97 194 153 84 92 81 197 134 302 258 279 169 121 115 82 139 89 190 90 119 212 177 102 208 266 300 53 179 81 78 88 121 110 83 90 101 131 97 113 101 101 105 117 121 140 86 101 105 91 62 66 91 97 197 180 123 97 80 102 114 70 74 96 96 58 30 85 76 86 87 109 60 86 58 54 69 71 47 60 35 24 62 52 44 50 19 34 76 59 74 102 96 5 108 103 164 33 61 59 79 17 30 53 25 45 17 30 29 13 14 18 17 32 9 12 42 223 96 141 114 132 84 234 56 91 280 106 34 162 142 54 26 26 33 79 167 41 44 102 15 18 Burt and Moore repeated Jastrow's test both Avith children and adults of both sexes, with results that roughly confirm Jas- trow's. "The females are more personal and subjective iu their interests; the males are more impersonal and objective. . . . 52 [418] ASSOCIATION^ LEARNING AND MEMORY ^Vomen alter their theuies and topics far more frequently than men; men, on the other hand, show a greater variety of associa- tive connections between one idea and another within the same theme. The course of ideas is also more frequently disturbed in women by the various signs of 'complexes' (systems of asso- ciated ideas characterized b}^ strong emotional colorings)." These discrepancies raise the issue, as Miss Tanner has pointed out, whether this test can be expected to reveal funda- mental native diilt'erences in mental constitution of the two sexes, or whether it reveals merely acquired traits, social traditions, individual habits, educational and other environmental influ- ences. The lists written by college students might be expected, for example, to be considerably affected by their recent occupa- tions, courses of study pursued at the time, etc. The more direct comparison of the speed of the two sexes in naming terms of different kinds which we have suggested (Vari- ation of Method, 3) does not appear to have been attempted by any of these investigators. (10) This influence of environment upon the lists of associa- tions is indicated particularly in Flournoy's brief tests (10 words and 10 drawings), the results of which are summarized in Table 57. TABLE 57 Influences that Affect 'VncontrolJcd' Series of Words or Drawings (Flournoy) DRAWINGS WORDS Traced to present surroundings Per cent. 13.8 1.9 Per cent. 29.0 Traced to the immediate past 8.2 Due to the milieu 15.7 2.4 39.2 37.2 Traced to recent personal experiences Traced to personal habits 3.9 9.2 Expressing individuality 41.6 42.7 13.1 Unexplained 49.7 TEST 33a : UNCONTROLLED ASSOCL\TION [419] 53 REFERENCES (1) A. Biuet, L'etucle experimentale de riutelligence. Taris, 1903. I'l). 309. Especially chs. ii to iv. (2) C. Biu-t and R. C. IMoore, The uieutal differences between the sexes. JEFd, 1 : 19 12, 273-2S4, 355-388. (3) Mary W. Calkins, Comnumity of ideas of men and women. PsR, 3 : 1S9G, 420-430. (4) J. jNIcK. Cattell and Sophie Bryant. Mental association investi- gated by experiment. Miml, 14 : 1889, 230-250. (5) Th. Flouruoy, De Taction du milieu sur I'ideation. AnPs, 1 : 1894 (1895), 180-190. (G) J. Jastrow. A study in mental statistics, isleio lieuiciv, 5: 1891. 559-568. (7) J. JastroAT, Community of ideas of men and women. PsR, 3: 1890, G8-71, 430-1. (8) Genevieve S. jManchester, Experiments on the nnreflective ideas of men and women. PsR, 12 : 1905, 50-66. (9) Cordelia Nevers, Dr. Jastrow on community of ideas of men and women. PsR, 2 : 1895, 363-7. (10) W. H. Pyle, The examination of school children. New York, 1913. Pp. 70, especially 24-27. (11) Amy Tanner, The community of ideas of men and women. PsR, 3 : 1896, 548-5.50. (12) J. E. W. Wallin, Experimental studies of mental defectives. EdPsMon, No. 7, Baltimore, 1912. Pp. 155. TEST 33A Uncontrolled association — discrete method (Kent-Rosanoff test). — This test resembles the preceding one in that it deals with association of the free, unrestricted or uncontrolled type, but it differs from it in that *S' is called upon to respond with a single term on\j to each of a series of words presented by E. This form of response has been, of course, the object of an ex- traordinary amount of investigation, particularly with refer- ence to its time-relations. But in the special arrangement of the test developed by Kent and Kosanoff no attempt is made to measure the time-relations, and the search for devices for the logical classification of the responses (a decidedly prominent feature of many laboratory and clinical studies in association) is limited to a simple empirical sorting of them into 'common,' 'doubtful,' and 'individual' responses, on the basis of prepared frequency tables. Tabulated lists of the frequency with which different re- sponses are made to the stimuli presented in association tests 54 [420] ASSOCIATION, LEARNING AND IVIEMORY were constructed by Cattell and Bryant (4) as early as 1889, and since then have been developed to some extent by Gertrud Bal- ing (20) in 1908, and by Reinhold (16) in 1910, while Bovet (1) has outlined several methods by which such tables might be handled in figairing a 'coefficient of banality.' Nevertheless, the frequency tables published by Grace Kent and A. J. Eosanoff (10) in 1910, taken in conjunction with the supple- mentary reports upon their applicability made by Eosanoff Avith the assistance of Eastman (5) in 1912 and of Isabel Eosanoff (18) in 1913 and the recent study by Miss Otis (15), constitute so important and well-standardized a development of the idea of measuring commonplaceness, or normality of re- sponse by means of empirical tables of distribution, as to war- rant the introduction of their test as a special and specific method of testing association. Whether the Kent-Eosauoff test merits the rather extrava- gant encomiums that have been awarded it by some writers^ appears to me extremely doubtful; it has certainlj^ discarded whatever advantages might be secured by resort to introspec- tion and to the making of time measurements; it sets up an arbitrary standard of normality, valid at best only in the gross and when the test is conducted by certain fixed and probably far from ideal conditions for exploring individuality in mental connections. Materials. — Prepared forms comprising a printed list of 100 stimulus words- with spaces for recording responses, their times and their indexes. The Kent-Eosanoff frequency tables. [Stop-watch.] ^Woocl worth and Wells (2.5), for examiole, talk of the free association test as having achieved, and being likely to retain, a place "in the fore- most rank among the methods of individual psychology," and assert that the form of it developed by Kent and Eosanoff gives "perhaps the best objective correlate of temperament at present to hand," and that it has "established a definite standard of normality" — statements that are hard to understand in the light of the results established to date. 'Sixty-six of these terms are taken from the series published by Sommer in his Diagnostifc dcr Geisteslcranlclieiten ; the remainder have been se- lected on the basis of preliminary experimentation in such a manner as to cover a variety of situations without being especially liable to call up personal experiences. TEST 33a: uncontrolled association [421] 55 Method.— Seat aS' in a. room free from distracting influencei^ and with bis back to E. Instruct bim as follows : ''I am going to read to j^ou, one at a time, a series of 100 words. Just be- fore eacb w^ord I sball call out 'ready.' As soon as you bear tbe Avord tbat follows tbe 'ready' signal, you are to respond by saying tbe first w^ord tbat comes to your mind otber tban tbe word tbat I bave just spoken. Your response must be a single word, and you must say it just as quickly as you can." If S, despite tliese instructions, repeats tbe stimulus word, be is cautioned not to do so, and tbe same stimulus is given again after several otber stimuli bave been used. If be con- tinues to repeat tbe stimulus word during some 25 trials, E sbould forego furtber attempts to prevent tbis form of re- sponse. If S responds by a sentence or pbrase, a compound word or a grammatical variation of tbe stimulus-word, be is similarly warned of tbis infringement of tbe instructions, and tbe stimulus words are similarly repeated later in tbe test- ing. In any event, tbe original response as Avell as tbe subse- quent one bad best be noted in tbe record, tbougb tbe second ones sbould be used in computing tbe results.^ If any response seems incoberent, devoid of any apparent connection witb tbe stimulus, ask 8 wby be responded as be did, and make a note of bis explanation. As tbe test is somewbat wearisome witb cbildren, it is ad- visable to introduce a rest-pause of a minute or so after tbe 50tb word, or even after tbe 25tb, 50tb and 75tb words. Variations op Method. — ^(1) Use tbe stop-watcb to measure tbe time elapsing between tbe stimulus and tbe response. Start tbe watcb just as tbe stimulus is uttered ; stop it wben 8 utters bis response; record tbe time in tentbs of a second. Tbis varia- tion of metbod is strongly advised, despite tbe reasons advanced by Kent and Eosanoff for neglecting tbe measurement of tbe association time. Tbe experience of otber users of tbe Kent- Rosanoff test sbows tbat tbe times are frequently valuable ad- juncts in diagnosis. E must remember, of course, tbat tbe time "Willi guite young cliilclren, say 4 or 5 years old, it will be impossible to follow tliese iiistrnctions precisely. Thus tbe Rosanoffs (IS) were obliged in such cases to i)ermit responses of a sentence form and to take tbe main word in tbe sentence as tbe desired single word. 5G [422] ASSOCIATION; LEARNING AND MEMORY does not ahvavs measure the speed of the assoeiation that is recorded, since between stimnlus and response more than one mental process may intervene Avhich is not reported by ^ and which may be quite unrevealed in the word he utters. (2) Make the test without instructions for speed, but with explicit instructions to adopt a quiet, leisurely attitude in which the association is allowed to develop in whatever way it niaj'. This method of conducting the association test yields responses that often differ widely from those obtained under instructions for speed, and it must be understood that the coeflScients obtained from the frequency tables then possess no necessaiy correspondence with those obtained when the stand- ard mfethod is followed. (3) Follow the suggestions just cited in Variation 2, with the additional proviso that >S may respond by a phrase or com- pound word in case that be the first verbal association that rises in his mind. It is instructive to compare the responses obtained under this Aufgalm with those obtained by the stand- ard method. Here, again, the calculated coefficients are not directly comparable with those established by Kosanoff with his tables and his instructions, though the method is, in the author's opinion, a better one for determining the degree of individuality in associative tendencies. Treatment of Data. — To determine the coefficient of com- monplaceness compare the responses for each one of the 100 terms with the responses listed in the Kent-Eosanoff frequency tables; record the several 'index-values,' then average them to obtain the coefficient. Thus, if to taMe 8 responds chair ^ the index is recorded as 267, because 267 of the 1000 persons tested by Kent and Eosanoff gave this response : if the association be taMe-horcl, its index is 9; if it be tahle-tlack, the index is 0, because no one of the 1000 persons chanced to give that re- sponse. The association tahle-hlack and any other association which is not found in the frequency tables is termed an indi- vklual response, while any association found in the tables, whatever its index may be, is termed a common response. Any response that is a grammatical variant of a term listed in the tables is classed as a doubtful response, e. g., the association TEST 33a: uncontrolled association [423J 57 tahlc-iiil-ij is doubtful since onlj^ taMe-ink is found iu the list for that stimulus word. The i)ereentage of common, individual and doubtful re- sponses are then comi)uted. If no response is obtained (c. (j., ''nothing," "don't know"), E maj^ find it necessary to add a fourth class (''failures") to the three classes just mentioned. The term common response is not entirely synonymous with the term normal response^ because, obviously, the responses gathered from 1000 persons do not exhaust' the possibilities of perfectly 'natural' associations. To meet this difficulty Kent and Kosanoff have given in their appendix statements that cover in a general wa^^ the responses that are to be deemed 'normal' for each of the 100 stimulus words and also still more general rules to cover associations to any stimulus word. These expla- nations (10, pp. 126-142) must be kept in mind whenever the question arises whether the responses of a given *S', however individual the^- may be, are yet within the bounds of normality. In certnin of the words Jn their list this restriction of the indexes to those resiioiises actually secured from the 1000 persons occasions a per- fectly oljvious and rather unfortunate artificiality. Take, for example, the word city (No. 79). As a response to it, 12 different cities have been named, and with the most diA'ers frequencies, e. g., New York, 99 ; Cleve- land, 1. If >S' chanced to respond IiidianapoUs, he would have to be cred- ited with an individual response, index 0. I would suggest that in this case all names of cities be counted together and the name of any city be given tlie resultant index, 124. Similar situations arise with other stimu- lus words, like doctor, square, child, ocean, etc., and with respect to the grammatical variants of many of the responses. Thus, for instance, the association mun-woinan has an index of 394, that man-v:omcn an index of 0. There are numerous such cases in which a very slight modification of the response alters enormously its index value, so that changes in the association which would appear psychologically indifferent remove re- s])onses from tlie realm of the commonplace and credit them with indi- viduality. Another criticism that might be raised against the use of the frequency tables in tlie inanner prescribed is that the tendency toward common- l)laceness may be almormally raised by the chance giving of a very few responses whose index value is unusually high. To give the three re- sponses table-chair, dork-lif/ht, fioft-hnrd. alone, will give .S' 10.j9 points of commonplaceness, even if he should happen to give an entirely individual i-esi)onse to every one of the 97 remaining terms. To meet this difficulty, .S"s may be compared simply with respect to the number of common and of individual responses they have given, or, as suggested by Bovet. by computing as an index of banality the number of terms to which the most common response is given and as an index of originality the num- lier of unique responses. Still other methods have been suggestetl by Bovet. 58 [-124] ASSOCIATION, LEARNING AND MEMORY In addition to computing the coefficient of commonplaceness, and the percentage of common, doubtful and individual re- sponses and failures, E may attempt a further classification of the responses, and, indeed, he will find it quite desirable to do so whenever the number of individual responses distinctly exceeds the ordinary number, or whenever other features of the responses indicate the possibility of some anomaly' in the asso- ciative processes. For this purpose the classification and analy- sis published by Kent and Eosanoff in conjunction with their frequency tables will serve satisfactorily.* The following explanation may serve to assist in the use of this classification. (a1 While couuiion responses are as a rule also normal responses, there are certain ones of them that may be termed non-specific responses, which, if present to an unusual degree, may have some pathological signifi- cance. A non-specific response is one which has so wide an application as to be a possible associate for almost any stimulus word, e. p., such nouns as tiling, article, ohiect, or such adjec-tives as good, small, useful, pleasant. ^^"ithin the individual responses, in addition to individual non-specific responses (like those just mentioned, but not in the frequency tables), there may be distinguished : (?>) Responses hp soioul, that lead to neologisms, /. c., construction of new words, e. g., man-man ion, anger- Angoria. (c) Neologisms without sound relation, as dark-tinhrigJit, deep-dept- al)U ness. (d) Repetition of the preceding response. ((') Repetition of a response five times or over (stereotypy), as the response parent to the stimuli man. mountain, mutton, sliort, woman, cold, etc. if) Repetition of the preceding stimulus. (g) Response by derivatives, /. e., grammatical variants of the stim- ulus word, e. g.. short-shortness, sweet-sweetened. (h) Individual non-specific responses. (i) Responses l>y sound, but with actual words, e. g., man-manners, short-shortliand. (/) Word complements, /. e., responses in which an addition to the stinuilus word forms a word, name or compound term iu common use, G. g., liahy-hood, thirst i/-hlood, green-Paris. (Jc) Responses hij particles of speech, as articles, numerals, pronouns, auxiliary verbs, adverbs of time, place and degree, conjunctions, preposi- tions and interjections, e. g., chair-down, eating-sometimes, soldier-pours, whisTiey-no. (I) Association to preceding stimulus, meaning a response not foimd ^The prolonged discussion concerning the most feasible and psycho- logically justifiable system of classifying responses iu association tests is too lengthy for consideration here. For some account of recent classifica- tions the reader may consult Wells (22) and Kelley (9). The idea of Kent and Rosanoff has been to forego logical classification in favor of a strictly empirical and objective system. TEST .j3a : UN(:ONTUOLLED ASSOCIATION [425] 59 ill the freaiiency tables for the word tliat it follows, but found in them as a response for the preceding stiniulns, as in the pair of responses, thief- iii(jh t, lion-pockcthoolc. {m) Association to yrcrediug rcfS"s has also been given to facilitate comparison. The point upon which most emphasis has been placed is the relativelj'' small percentage of individual responses (G.8) given by normal >S"s. (2) Normal times for free association with the Kent-Rosanoff series have been reported by few experimenters. Miss Otis merely states that the times proved significant and valuable and that defective children were both slower and more variable than normal children. Goett (6), who used Jung's list, found that with normal children the mode was in the 2d second and 60 [426] ASSOCIATION, LEARNING AND MEMORY TABLE 58 Average Distributions for the Kent-Rosauoff Test {Com piled from Eastman, Kent, Rosanoff and Strong) EXPERIMENTERS S'S COMMON DOUBTFUL INDIVID'AL Kent and Rosanoff Kent and Rosanoff Kent and Rosanoff Strong Eastman and Rosanoff 1000 normals 8-80 91.7 1.5 247 insane___ adults 70.7 2.5 32man.-dep. adults 75.8 3.0 16 man.-dep. adults 78.6 1.0 253 delinq't 11-17 84.0 2.2 6.8 26.8 21.5 19.7 13.2 was not much affected by age, while with mental defectives the mode was in the 3d second. Kelley's tests of 12 college stu- dents, with a list of 100 terms decidedly more difficult than the Kent-Eosanoff' list, revealed a skewed curve with the mean slightly- higher than the median, the median slightly higher than the mode; the mode was 1.0 sec, the minimal time 0.5 sec, the maximal time 3.5 sec The author has found the average times for college students with the Kent-Kosanoff list to lie between 1.00 and 2.75 sec An overage less than 1.5 sec may be con- strued as a fast association time. (3) Dependenee on age. That children give distinctly fewer common associations (and hence more individual associations) than adults is the general conclusion of all experimenters (Eeinhold, Saling, Wreschner, Ziehen and the Eosanoft's), though Eeinhold did not find the number of common responses to increase steadily from year to year, and the Eosanoffs be- lieve that the differences between children and adults ai'e jjrac- tically obliterated after the age of 11. Their results, expressed in per cents., are shown in Table 59. Graphs of these distributions will be found in the original text (p. 49). The increase in the frequency of individual responses at the age of 15 is attributed to the presence of a number of retarded pupils who were still members of a grammar school at this age. Failures to respond (sixth column), which include replies of "^'don't know," are found to be due usually to lack of familiarit}' with the stimulus words. While this conclusion i§ borne out by their detailed table of failures (18, p. 47). it TEST 33a: uncontrolled association [427] Ci TABLE 59 Dependence of Distrihution in the Kent-Rosanoff Test on Age {Isal)el Rosanoff and A. J. Rosanoff) COMMON RESPONSES DOUBTFUL RESPONSES INDIVIDUAL RESPONSES FAILURES AGE TO RESPOND Specific Non-Specific 4 40.4 1.1 3.8 25.3 29.4 5 55.1 2.0 4.4 21.4 17.1 6 62.2 2.7 - 3.2 18.6 13.3 7 ■ 64.9 4.0 3.5 20.0 7.6 8 68.4 5.8 3.1 18.0 4.7 9 75.1 5.5 1.7 14.2 3.5 10 72.9 8.4 2.3 14.3 2.1 11 82.0 7.1 1.7 8.6 0.6 12 83.8 6.6 1.3 7.6 0.7 13 81.1 8.4 1.8 8.5 0.2 14 84.1 6.3 1.4 7.7 0.5 15 78.7 7.6 2.0 10.8 0.9 Adults 85.5 6.2 1.5 6.8 must be borne in mind that occasionally even normal adults reply ''nothing," and again that young children often make no reply and yet subsequent questioning shows that they have had numerous visual and even verbal associates in consciousness to which for one reason or another they have given no expres- sion. A further analysis of these writers sheds some light upon the relative jireponderance in children (300 cases, 4-15 years old) as compared with adults (86 normal cases, selected records containing not over 10 per cent, individual responses) of cer- tain types of individual responses. Eeference to this analysis, reproduced in Table GO, shows that the greater part of the excess individual responses given by children fall in the categories 'partial dissociation' and 'perseveration,' while the individual, but normal responses (by appendix to the frequency tables) are actually fcAver with children.^ '^In this analysis tlie tei'in 'partial dissociation' embraces what liave been described abo\'e as non-specific resjionses. responses by sound (including neologisms), ^Yord complements and particles of speech, while the term 'perseveration' embraces all varieties of responses to earlier stimuli or to earlier responses and repetitions of responses more than five times. (>2 [42S] ASSOCIATION, LEAUNING AND jMEMOUY These experimeiitei^s argue that "it would seem, then, that the tendency of children to respond with individual reactions more TABLE GO Indiridtiul Responses of OliihJreii aiul Adults in the Kent-Rosanoff Test {Isahcl Rosanoif and A. J. Rusanoff) TYPES OP RESPONSE 86 NORMAL ADULTS 300 CHILI 41.8 20.0 0.3 0.1 8.0 11.1 (i.l 27.8 0.6 43.8 40.4 Normal (by appendix) Derivatives ol stimulus words Partial dissociation Perseveration Neologisms (without sound relation) Unclassi fled often than adults rests in a large measure upon a certain lack of mobility of attention which results in an inability to quickly dismiss from the mind previous stimulus or reaction words and to turn the mind wholly toward the new stimulus word." The author is inolined to believe that Iiere, as in not a few other in- stances in whicli cliil'iren differ from adults in psychological tests, what wo are really brin.Lrini; to lii;ht is an inability of the children to nnderstand the instructions or disinclination to follow them if they are understood. In other words, the regular instructions of the Kent-Rosanoft" test consti- tute an artificial restriction of the natural associative tendencies, as will be shown further on. Adults are able and willing to maintain the proper attitude and follow the rules of the game ; many children are unable or unwilling to do so. Again, as regards the perseverative tendencies, no instructions are given to tlie child to avoid repetition of association, while it is common for cultured adults to avoid repetition from some preconceived notion that they are called upon so to do. JMoi'Oover, \\hile data are lacking to prove this contention, it seems very likely that the amoiuit of 'perseveration' witnessed in responses to an association test is much inlluenced by the speed with which the entire test is conducted, because the faster the succession of stinmli, the greater the 'hang-over' effects of the words, whether stimuli or responses, that have been in consciousness. Although precise statements are wanting, Hosanoff and his co-workers appear to have conducted their tests at a fast pace. It goes without saying that for purposes of comparison be- tween any groups of »V's, the speed of giving the stinmli should be con- stant and that pauses for rest introduced with one group should be intro- duced likewise with the others. A closely similar opinion is expressed by Kakise (S), when he says: "To sum up, these so-called characteristic forms in children and the ab- normal can all be found in normal adults in their natural associations, /. c, when they react according to natural and spontaneous suggestions, TEST 33a: uncontuollki) association [421)J i\:\ as was the case with our experiment, and do not react according to arti- ficial and 'sojjhisticated' associations, /. p., by mere verbal associations, as is the case in the customary experiment with normal observers who are expert enough to obey the 'rules.' " Another attempt to establish relations between age and asso- ciative type has been made by Miss Otis, who tested 200 normal children, aged 4 to 8 years, 40 in each of the 5 ages, and com- pared the results with those for 130 children in the Vineland, N. J., Training School for the Feeble-Miuded, classified for mental age by the Binet-Simon tests. TABLE Gl Types of Associative Response m Normal and Fechle-Minded Children (Otis) Normals. AGE FAILURE TYPE I TYPE I[ TYPE III TYPE IV TYPE V TOTAL 4 1 21 4 11 3 40 5 G 13 1 14 9 43 6 9 6 13 18 39 7 2 1 8 29 40 8 1 1 3 5 30 40 All 1 28 22 11 51 89 202 Defectives. AGE FAILURE TYPE I TYPE II TYPE HI TYPE IV TYPE V TOTAL 2 4 • 4 3 5 1 1 7 4 3 1 1 5 5 1 5 1 7 6 2 3 I 9 1 IG 7 1 11 G 18 8 1 7 19 27 9 1 7 15 23 10 3 14 17 11 4 4 12 4 4 All 2 17 4 1 43 G5 132 04 [430] ASSOCIATION^, LEARNING AND MEMORY In Table 61, T^pe I signifies repetition of the stimulus, Type II non-logical responses (no apparent connection between stimulus and response), Tj'pe III responses b}^ sound (whether by a real word or by a neologism), Type IV multiverbal re- sponses (like tvJiistle — ichen you ivliistle^ doctor — to make you 'better^ tfihle — there's a table), and Type V normal responses (meaning here responses by one word, of Avhich at least 50 per cent, must be found in the frequency tables). A child is classed as belonging clearly to one of these five types only when at least 50 per cent, of his responses are of the kind indicated, but the figures given in Table 61 include cases of 'mixed' types, which have been classed hj the preponderant tendency. So far as normal children are concerned, it appears (1) that at 4 years more than half belong to the types characterized by repetition of the stimulus, (2) that non-logical responses (Type II) are characteristic of 5 years, (3) that multiverbal responses (Tj^pe IV) are very prevalent from 4 to 6 years, (4) that a normal type of response, in the sense here used, is established in 75 per cent, of children by the age of 8 j^ears, though these children by no means respond like adults, or even like children of 12 when their detailed responses are taken into consideration. The relation of speed of association to age is not so clearly made out as one might expect. Nearly all experimenters find that work with the Kent-Kosanoff lists takes longer with chil- dren than with adults. Ziehen concluded that free association times decreased markedly year by year and Wreschner reached a similar conclusion, but both Goett and Rusk report that there is no definite relation betAveen speed and age, while Meumann calls attention to the fact that, though work progresses more rapidly with older children, the more intelligent not infre- quently respond more slowly, and the less intelligent, by reason, seemingl}^, of their relatively less originality and paucity of imagery, frequently respond more rapidly. In the limited num- ber of tests made by the author, children (of about the age of 9) have invariably been distinctly slower than adults. (3) Dependence on sex. The results reported by Burt and by Burt and Moore show a number of inconsistencies : in one group at least ^65 children in the Holt School, Liverpool) the TEST 33a : UNCONTROLLED ASSOCIATION [431] G5 girls slightly exceeded the boys in number of associations writ- ten, while in another group (130 children, aged 13 years, in the Wallasey School) 35 per cent, of the boys exceeded the median of girls. Burt and Moore, in any event, conclude that "the males are far quicker than the females." In their compilation of data from 1000 normal ^*'S Kent and Eosanoft" did not find any considerable differences between the sexes in the nature of the distribution of the responses. (4) Dependence on practise. Both Rusk and Wells (23) find that practise in giving free associations reduces the time. Since this practise is not gained by actual repetition of the same series of stimulus words, the gain in time must lie in facilita- tion of general factors that condition the process of associating. Wells finds that the responses become less emotional, that the number of supraordinate relations is diminished and that of simple language-motor responses is increased, while at the same time there is greater 'particularization' in the responses. Ver- bal connections appear to become 'loosened up,' and general linguistic readiness is augmented. The effect of practise, then, is to develop an easier, simpler and more superficial type of re- sponse. Practise also decreases the times, so that the median speed is reduced to about 1.2 sec. from any amount above that up to 3.0 sec, with the consequence that individual differences in speed are less after practise than before it. (5) Dependence on intelligence. The original data collected by Kent and Rosanoff permit them to compare the responses of 100 persons of collegiate education with those of 100 persons of connnon school education. The comparison indicates (see their Table I, p. 9) more individuality in the responses of those of collegiate education, but the authors deem it unsafe to risk a definite generalization to this effect on account of the wide variability in individual records of both groups. In the case of children Isabel and A. J. Rosanoff compared 21 'bright,' 21 'average' and 21 'dull' children (teachers' esti- mates) and secured the results shown in Table 62, The same investigators contrasted 38 pupils who were peda- gogically advanced with 38 pupils of the same ages who were pedagogically retarded, and found, similarly, that the retarded 66 [432] ASSOCIATION, LEARNING AND MEMORY TABLE 62 Distrihutlou of Responses as Conditioned Vy Intelligence (Isahel and A. J. Rosanoff) GROUP COMMON DOUBTFUL INDIVIDUAL FAILURES Bright — Average— DuU 79.0 75.3 66.9 3.4 2.7 2.3 12.0 12.6 22.0 5.6 9.4 8.8 pupils gave more individual responses (13.7 vs. 9.8 per cent.) and the advanced pupils more common responses (86.9 vs. 81.6 per cent.). They believe that extreme departure from the dis- tribution which is average for the age of the child in question is an indication of a fundamental difference in mental ability; that 'plus-variations' [exceptionally high percentage of common responses?] characterize cases of precocity, while '^minus- variations border on the pathological." These conclusions are distinctly at variance with those reached by Ziehen, by Wreschner and by Meumann, all of whom lind a greater degree of originality, i. e., more individuality, In the associations given by more intelligent children. Other differences cited by Meumann (14, 89-101) are the following: (1) the unintelligent more often misunderstand or misinterpret the stimulus word; (2) they more often fail to respond; (3) they give a greater number of incoherent and seemingly sense- less associations; (4) they more often use very 'superficial' con nections, such as grammatical variants of the stimulus, rimes, simple opposites; (5) they often give responses derived from phrases or verbal connections that they have learned in some school exercise; (6) they often exhibit an apparent precocity by giving responses like adults rather than the more concrete and pictorial associations that are characteristic of most chil- dren of their years; (7) they tend to stick to certain forms of response once they have begun to use them (perseveration). Keinhold, on the other hand, found that in two of four classes the better children showed more, and in the other two less orig- inality than the poorer children : he also argues that no differen- tiation between intelligent and unintelligent children can be TEST 33a : UNCONTROLLED ASSOCLVTION [433] 67 made on the basis of the tendency toward responses by sound, as this tendency is found on repetition of the test to be quite variable and a mere matter of chance attitude or 'set.' Simi- larly, Winteler, who sought to distinguish two types of re- sponse (the one termed the perceptual or describing, the other the comparing or relating type), could discern no relation be- tween intelligence and propensity to use these types. (6) Dependence on family relationsliip. Fiirst, Avho tested 100 persons in 24 families with Jung's test w^ords (7) and classified the responses under various categories, concluded that persons related to one another tend to exhibit more similarity in the use of these types than do persons not related. He also concluded that the associative type of children resembles that of their mother more than that of their father. (7) The feehle-minded. Miss Otis' results with Vineland children have been i^resented in Table 61, where it is shown that repetition of the stimulus (Type I) is a common tendency with low-grade mental defectives, that multiverbal responses (Type IV) are encountered more often and persist till a later age in feeble-minded than in normal children, and that normal responses (Tj^je V) appear later and less regularly in the feeble-minded. Goett deems the test of diagnostic value for ex- amining abnormal children. He states that imbeciles have slower association times, tend to repeat responses and to give an unusually large number of multiverbal and non-specific re- sponses and responses of the 'predicative' type {wood — hum, glass — breaking). The 253 children examined by Eastman and Eosanoff seem to have been at least two 3 ears or more peda- gogically retarded. The results accord quite closely with those of aiiss Otis and of Goett in that they reveal an unusual number of non-specific responses, of repetitions of response, and of the use of particles. In addition, these investigators found a rela- tively large proportion of failures to respond, and a percentage of individual responses much above the average for normal per- sons, though not so great as in the insane (Table 58). On the other hand, really incoherent responses, senseless neologisms, etc., so frequent in the insane, were almost never given by the feeble-minded. These authors point out that the conclusions (!S |4;M| ASSOCIATION, IJOAUiNlNU AM) iMK.MOUV jus( ciU'cl ai'o triio oiilv for (lie jirouj) as a mIioIo; a i^cuxl many of Ihe (•liildi-ou ratod as looblo-iniiulcd oi' dt'liiKiiuMil <;avo noniuil associations. >Vhcii (o (liis admission is added tlic dciuonslra- tion of Kakisc, (o which wc liave alluded, thai nuiltivei'bal re- sponses, repetition of the stimulus and other tendencies sup- posed to be characteristic of abnormal nunds can also be found in normal adults, the value of the Kent-Kosanotf test as a device for diajA'nosis of individual cases is certainly much less evident than some of its friends avouUI have us believe. (8) 77/0 insane. That the insane show a relatively high fre- quency of individual responses is shown by the work of Kent and Kosanotr and of i^^tronji- (Table 58). Kent and Kosanotl' have also shown by further analysis of their nniterial (ID, p. 21)) that there are characteristic ditferences in the distribution of the various forms of individual responses in the several dif- ferent forms of insanity, such as dementia praecox, paranoia, epilepsy, general paresis, manic-depressive insanity. A similar comlusion is reached by Ley and IMenzerath (12). The results obtained by Strong with KJ cases of numic-depressive insanity shoAV good agreement with those obtained by Kt'nt and Kosanolf for [V2 cases of the same sort. To what extent inferences may safely be drawn from jteculiarities in the times of responses, particularly from excessive slowness of reply, with resi)cct to the pres(>iic(» of hiddcd emotional com])Iexes is a niattt'r of nuu-h dispute.'' (9) DcpciKhiicc on instructions. Attention has already been called to the dilVerence in the outcome of the assot-iation test according as ^'' is set to respond as qnicl'lt/ as ])ossible or as /re// as possible.' It should be repeated that the lonclnsions drawn from the Kcnt-Hosanotf test with its freipu'ucy tables hold only when the instructions to res[)ond by a single word as (piickly as possible are strictly followed. In illustration reference nmy be made to the author's own responses, taken under Varia- "A gonoval idea of this iu-ol)Ioiu may l)o gaiiUHl from the rori'roiues Iicre cited from .Tung. I.oy and Mouzoratli and Lovy-Siilil. "Consult Monmanu (1,"^. 420 ff.) for further analysis of possible instruc- tions for this tost. Koels (17) has also called attention to the fact that jS^'s, despite unifonnity of instructions, do adopt different attitudes toward the test and thus give different times and responses. 'nOS'l' .'>.')A : lJN<'<)N'nt()I,M':i) ASSOCIATION |1'>'">| *>-^ lion of Mcdiod No. ;> (leisurely i-esponst? will) |ierniission lo use jtlirases \vlieii sncli did ai)j)car first in consciousimssj. Tliree alteralions arc; jfioniincnt wlien Hie responses are conijjared will) (liose hy llie standard inslruciions. (a) Tlie nnnd)er of individnal responses is decidedly inereased, so tliat the coeffi- cieul of (■oninioni)laeeness falls from 12.8 to 10.5 (reckoned in lernis of (Ik; mean) or from 7.0 to 3.0 (reckoned in lerms of ilie median). Striking examples are the following: STIMULUS QUICK RESPONSE COlilFriCIENT LEISURELY RESPONSE COEFFICIENT soft n('('dl(! I'clifrion whiskey city hfird tin-cad faith rye town 365 160 47 9 258 pedal stickpin ecstasy rotten voleur' (h) There are numerous responses by phrases, and, furtlier- niore, comparison of these ])hi"ases with the single-term re- sponses under standard instruction shows that the single terms v\-ere really ])icked out fi-om the ])hrase that was rising in con- sciousness. That yonng children may not always stop to make (his seleclion is Ihe evident explanation of the tendency seen in them to I'esjjond by phrases, even despite re])eated instnic- tions to tlie conlrary by E. The following examples will make 1 his ])oinl cleai- : QUICK RESPONSE LEISURELY RESPONSE lion command justice child boast order peace father king of beasts yours to command .justice, peace and mei'cy cliild is fatlier to tli(! man (c) There are numerous indications of 'perseverative' tend- encies, esix'cialiy in liie use of the same response for a number "By \\:\y ul' llic ]ilii';isc: "llic cily is i'lill of fhii'vcs" iiiul llicncc !<» Ilic I'^reiicli for thief — iiii cxccili'iil ilhistr;i1ii)ii of Ihe coiiiiilcxily acities measured and to evaluating the results of these tests lies in the selection of the stimulus words themselves, for, if the terms are too difficult, failures appear due to lack of famil- iarity with their meaning or with the meaning of the terms connected with them in various logical relations; while, if they are too simple, no 'thinking' is demanded and the responses are 72 [438] ASSOCIATION, LEARNING AND MEMORY given well-nigh automatically. Moreover, the inclusion in a list of terms of one or more stiinnli that are markedly different from the others in this respect introduces a source of difficulty in administering the test that is hard to meet, especially in group tests. The only solution of these difficulties is to discover by comprehensive testing what might be termed the 'association value' of each stimulus word for ^'s of a given sex, age, intelli- gence, etc., and then to prepare standardized lists of stimuli suited by their like association values to the measure of con- trolled association in specified types of ^'s. Much has been accomplished in this direction, but much still remains to be done. Consideration of the various possible forms of controlled association is limited in what follows mainly to the most-used relations, part-whole, genus-species, and opposites.* A. THE PART-WHOLE TEST Materials. — (1) For individual tests: Split-second stop- watch. Set of 20 cards (and 3 samples), each containing a stimulus word. Paper for recording times, responses and re- marks. (2) For group test : Stop-watch or special seconds clock. Printed form containing the same stimuli and provided with spaces for the recording of the 20 associates. The terms incorporated iu these cards and in tlie form are tliose rec- ommended by AYoodworth and Wells as the result of numerous efforts at standardization. Cards are used, however, instead of the narrow cardboard strip of tliese autliors in order that the time of each response may be measured by itself. The paper form is used to admit of. written group tests. The terms proposed by Pyle for this test are : window, leaf, pillow, button, nose, smokestack, cogwlieel, cover, letter, petal, page, cob, axle, lever, blade, sail, coach, cylinder, beak, stamen. His supplementary list is the same as that of Woodworth and Wells. The 10-word lists used by Kusk were : ear, wheel, beak, inch, platform, mast, branch, kernel, funnel, buckle: aUciiiativcs, moutli, handle, claw, ounce, pavement, sail, stem, core, boiler, knob. Tlie ten terms employed by Miss Norsworthy were : door, pillow, letter, leaf, button, nose, cover, page, engine, glass. The ten terms employed by Wyatt are not specified by him. ^The mixed relations or analogies test is dealt witli separately as Test 3JA. For further details concerning other tests of logical relationship, consult Rusk (15), Watt (20) and Woodworth and Wells (22). TEST M : CONTROLLED ASSOCIATION [430] 73 Method. — Instruct *S' as follows : ''Each one of these cards has printed on it a woi-d. As soon as I uncover a card I want .you to look at the word on it and then, as quickly as you can, say aloud the name of the whole thing of which that word is a ]jart. The word you read is a part : you are to name the whole. For example, if the card should have the word fur on it, you might say cat or seal or fox. We will try these sami>le cards first to make sure you understand." After a warning 'now' remove the cover-card from the top of the pile and take >S"s time for the first samjile — button. Follow with the other samples, leaf and drawer. These cards are dis- played and the time taken just as in the test proper, in order to accustom *S' to the regular i^rocedure. Misunderstandings are, of course, coiTected and cleared away. Proceed with the 20 standard test cards. Record on the blank sheet of paper the times, in tenths of a second, together with ;S"s responses and any comments that suggest themselves. Each card is provided with its own cover-card. They are best re- moved with the left hand and the watch started simultaneously with the right. It is recommended that the split-second watch be used, so that one hand may be stox^ped when S first re- sponds ; then, if his response chances to be wrong, say 'No, give me another,' and take the time of his second attempt with the other hand of the watch. In this event, both times and both responses are recorded. Variations of Method. — (1) Individual testing may also be carried out, especially if none of the terms is likely to cause unusual delay, by giving *S' the printed form provided for group tests and taking his total time for naming orally the entire series of responses. (2) For a group test by the work-limit method^ (which is recommended for 8's who are competent to record their own time) use the printed forms and the special seconds clock, after the samples above mentioned have been displayed on a black- board and discussed with the aS"s. The clock is started at the signal for turning over the forms. Each >S', of course, makes his own written record. =See Vol. I, p. 8, Section (7). 74 [440] ASSOCIATION, LEARNING AND MEMORY (3) For a group test by the time-limit method use the same forms and stop all »S"s at a time-limit so chosen on the basis of preliminary trials with *S's of that grade of ability that the fastest iS* shall reach about the 10th term on the list.^ It is hardly necessary to state that the scores obtained by either group test are not directly comparable with those obtained by the individual method. Treatment op Data. — (1) In the individual test the best indication of speed is furnished by the median. For a measure of variability the semi-quartile variation may be used, /. e., one- half the difference between the 5th and 15th time, when the series of times is arranged in order from fastest to slowest. If the *S"s are competent, the errors will ordinarily be negligible, so that performance may be measured in terms of speed alone. If it should happen that differences in speed are slight, Avhile qualitative differences are well-marked, speed may be neglected and performance rated in terms of quality, as by scoring 1 for each well-chosen associate, 0.5 for each 'partly right' associate, and for wrong associates or omissions. If both speed and correctness need to be considered, some of the methods sug- gested in the cancellation test (No. 26) or in the opposites test (below) may be employed. (2) When individuals are tested by recording the total time for the entire list (Variation 1), errors may again be neglected if few and of slight moment ; if more serious, the time may be increased by adding to it a penalty figured on the basis of the average time taken to utter a correct response to each stimulus omitted or responded to Avrongly. (3) In group tests by the work-limit method, performance may, similarly, be taken in terms of total time, or of correct- ness, or of some combination of time and correctness. (4) In group tests by the time-limit method, the simplest method of scoring is that of crediting 1, 0.5 or for each re- ^The lists of terms iii these tests of controlled association have been so arranged by Woodworth and Wells that the terms lying bet\yeen the 8th and the 10th represent as nearly as possible stimnli of equal difficulty. Thirty see. will suffice for testing competent adults. Pyle recommends 60 sec. for Grades 2, 3 and 4; 45 sec. for other grades (and 30 sec. for adults?). TEST 34 : CONTROLLED ASSOCLVTION [441] 75 spouse, as above explained. If necessary. to compare the per- formances of groups that have had different time-limits, they may be related by computing them all as if 60 sec. liad been assigned, e. (/., by multiplying the score of adults by 2, etc. Results. — (1) Norms of performance in the ])art-Avhole test are supplied chiefly in the data published by AVoodworth and Wells, by Miss Norsworthy and by Pyle. Tests of adult college students by the first-named authors show that the average asso- ciation time for this variety of controlled association (Varia- tion of Method No. 1) may be taken as 1.53 sec, P.E. .06, Avith a range for different individuals of from 1.03 to 2.50 sec. The. median times reported by Rusk for 22 children, aged 7 years 6 months to 14 years 9 months, under Meumann's 'B' instruc- tions (emphasizing quality rather than speed) range from l.G to 5.0 sec. Miss Norsworthy's norms, based on 504 cases, rep- resent results with her list of 10 words, no time-limit, scored in terms of number of correct associates. Pyle's norms are based upon his list of 20 words, scored in number correctly written in a group test, computed on a basis of 60 sec. time-limit. TABLE 63 Perfoiiiiance in the Part-Whole Test {JVorsicorthy) AGE 8 9 10 11 12 13 14 15 16 ADULTS Median P. E 6.5 2.3 7.8 1.3 7.8 1.9 8.7 1.1 8.7 1.2 9.0 0.7 9.0 0.7 9.0 0.7 9.0 0.7 10.0 5 TABLE 64 Correct Associates Written in 60 i^ec. Part-Whole Test (Pyle) SEX AGE 8 9 10 11 12 13 14 15 16 17 18 ADULT 1 Cases ,31 67 70 65 i 76 177 1 62 42 35 1 12 23 66 Male Aver. 5.5 6.5! 7.3, 8.9 8.9 11.1 12.2 14.8 15.9 15.8 19.3 18.5 A. D. 1 3.6 2.9 2.5 2.8 3.4 4.3 4.1 5.5! 5.3 4.0 5.6 3.6 Cases 43 64 88 67 87 71 '63 48 51 38 !28 87 Female Aver. 1 4.6' 5.91 7.810.010.010.812.514.0 16.9 16.2 19.7 19.7 A. D. 2.6 2.4 2.9 3.5 3.7 3.5 3.2 4.5 4.5 4.8 4.6 3.4 76 [J:42] ASSOCIATION^ LEARNING AND INIEMORY (2) Dependence on age. The results obtained bv both Pyle and Miss Norsworthy show that jjerformance in this test un- dergoes a fairly steady improvement from 8 to 18 years. The lack of any correspondence between speed and age reported by Knsk is probably due to the small number of cases examined by him. (3) Dependence on sex. Sex-differences are not sufficiently evident to warrant conclusions, though it may be surmised that girls and women tend to be slightly superior to boys and men. (4) Dependence on intelligence. Wyatt found a fair degree of correlation with intelligence (0.67, P.E. .07 in one group using teachers' estimates, and 0.56, P.E. .08 in another group, using class examinations as the basis for intelligence). (.5) Feehle-viinded. The work of Miss Norsworthj^ shows that mentally defective children are distinctly inferior to normal children in this test: thus the percentage of normal children with a record above the median, above — 1 P.E., and above — 2 P.E., would, of course, be 50, 75, and 91, respectively, but the percentages of feeble-minded children obtaining these three grades of efficiency were but 9, 17, and 27, respectively. That is, only 9 per cent, of the feeble-minded children reached the de- gree of efficiency attained by one-half of the normal children, etc. (6) Other correlations. Wyatt obtained with his Group I a moderatel}' satisfactory' coefficient of reliability, 0.65. His correlations with other tests range from 0.09 to 0.77. The lowest correlation was with. the letter-squares test; the higher correlations appeared with analogies (0.67), the completion test (0.75) and word-building (0.77). B. THE GENUS-SPECIES TEST ^ Materials. — (1) For individual tests: Split-second stop- watch. Set of 20 cards (and three samples) each containing a stimulus word. Paper for recording times, responses and re- marks. (2) For group tests: Stop-watch or special seconds clock. Printed form containing the same stimuli and provided with space for recording the 20 associates. TEST 34 : CONTROLLED ASSOCIATION [443] 77 These 20 terms are those recomiiieiided by Woodworth and Wells. The 10 terms used by Miss Norsworthy are : book, tree, room, toy, iiame, dish, boat, game, plant, fish. The 20 terms proposed by Pyle are : mountain, city, weed, metal, fur- niture, machine, author, planet, river, book, ocean, fruit, country, animal, bird, food, lake, tool, fish, money. His supplementary list is the same as Woodworth and Wells. The 10-word lists used by Rusk were : tree, fish, college, battle, picture, tool, hero, lesson, taste, wrong: aUenuttives, bird, leaf, game, poem, song, toy, hobby, book, smell, virtue. Method. — Instruct S as follows : "Each one of these cards has printed on it a word. As soon as I uncover a card, look at the word on it and then, as quickly as jou can, say aloud the name of some particular thing that belongs in the class that is given on the card. The word you read is the name of a class or genus; you are to name an example of that class, a species of that genus. For example, if the card should have on it the word taste, you might say sweet or salt, or if the word verl), you might name any verb like run or go. We will try three sample cards first to make sure you understand." Follow the procedure outlined for the j^art-whole test in regard to the use of the sample cards, timing, etc. Variations op Method. — Follow the suggestions given for part-whole test, save that here the samples will be 6 my?, dish and game. Treatment op Data. — Follow the suggestions given for the part-whole test. Results. — (1) Tests of adult college students by Woodworth and Wells (Variant Method No. 1) show for the genus-species test an average association time of 1.84 sec, P.E. .07, with a range for different individuals of from 1.20 to 2.63 sec. The medians reported by Rusk for 22 children from about 7 to 15 years of age, with quality emphasized more than speed, range from 1.6 to 11.4 sec. The norms reproduced here from Miss Norsworthy are based on 511 cases and represent performances made with her list of 10 words, no time-limit, scored in terms of number of correct associates. The norms reproduced from Pyle are based on his list of 20 words, scored in terms of num- ber correctly written in a group test, computed on a basis of 60 sec. time-limit. 78 [444] ASSOCIATION, learning and memory TABLE (;5 Perforyaanve in ihe Genus-Species Test {NorsicortJnj) AGE 8 9 10 11 12 13 14 15 16 ADULTS Median___ P. E 5.0 2.0 5.0 2.7 7.0 2.9 9.2 1.9 9.2 0.7 9.3 0.4 9.3 0.5 9.5 0.5 9.5 0.5 10.0 0.0 TABLE 06 Correct Associates Written in 60 Sec. Genus-Species Test (Pyle) SEX AGE 8 9 10 11 12 13 14 15 16 17 18 ADULT Male Female Cases Aver. A. D. Cases Aver. A. D. 29 4.6 3.4 34 5.5 3.6 67 5.7 3.4 65 5.4 2.5 66 6.5 3.7 84 7.8 3.2 62 7.2 3.3 63 8.2 3.7 69 7.1 2.5 81 9.3 2.9 68 10.0 3.8 64 9.5 3.2 64 10.5 3.8 55 11.8 3.2 41 33 18 16 11.1 15.2 14.0 17.3 5.4 4.3 4.1 6.0 40 45 32 25 14.0 16.416.018.3 4.2 5.4 4.9 5.3 65 15.1 4.0 86 15.5 3.8 (2) Dependence on age. Though Rusk can find no definite relation between speed and age in the genus-species test, the figures reported b}^ both Pyle and Miss Norsworthy show a general improvement with age, despite certain exceptions. Miss Norsworthj's test was obviously too easy for ages of 11 and above, so that any tendency toward improvement beyond 11 was obscured. Rusk's negative result is explicable partly by his instructions against haste and partly by the small number of cases he tested. (3) Depetidence on sex. Pyle's averages make it reasonable to assume a slight superiority of girls over boys, since they show this superioritj' in ten of the age groups. (4) Feehle- minded children, according to Miss Nors worthy's results, are distinctly inferior in this test to normal children of the same age : only 9 per cent, reach the median of the normal children; only 16 per cent, reach — 1 P.E.; only 17 per cent, reach — 2 P.E. of normal children of their age. TEST 34 : CONTROLLED ASSOCIATION [445] 79 C. THE OI'I'OSITES TEST Materials. — (1) For individual tests: Split-second stop- watch. Two sets of 20 cards each (exclusive of samples), one set of easy, and one of moderately difficult stimulus words. Paper for recording times, responses and remarks. (2) For group tests : Stop-watch or special seconds clock. Printed forms containing the same stimuli (one for easy and one for difficult words) and provided with spaces for recording the 20 asso- ciates. The opposites test has been extensively used and has appeared in a variety of forms. The most common lists are pi'inted herewith. They demand a few words of explanation. Lists I, II and III represent the standardized lists for easy opposites prepared by Woodworth and Wells : Lists I and II, which are those used by Briggs (his Tests 43 and 44) are presumed to be of equal difficulty and to be so arranged that the last half is just as difficult as the first half ; List III. which is the set of easy opposites here recommended, is a selection of the 20 easiest opposites in Lists I and II. MATEEIAL ITSED BY VARIOUS INVESTIGATORS IN THE OPPOSITES TEST. II III IV long north high good best soft sour summer outside weary white out out quick cloudy far weak white tall patient up good slow big careful smooth after yes loud stale early above above white tender dead sick north light ignorant hot slow top happy doubtful asleep large wet false serious lost rich good like reckless wet dark rich rich join high front up sick advance dirty love front glad honest east tall long thin gay day open hot empty forget yes summer east war calm wrong new day many rare empty come big above dim top male love friend difficult 80 [446] ASSOCIATION, LEARNING AND MEMORY VI VII VIII IX day great succeed tender asleep hot strict animated absent dirty tardy proficient brother heavy sleepy impoverish best late suspicious cruel above first rigid generous big left suave haughty backwards morning sinful silly buy much conservative insignificant come near refined disastrous cheap north pride miser broad open despondent I'csnlt dead round imaginary hindrance land sharp beautiful strength country east injurious innocent tall known diligent busy son something sell remember here stay sure increase less push active preserve mine nowhere venturesome belief List IV, one of the oldest and most employed, appears in several pub- lished articles and texts by Thorndike ; it forms one of Simpson's easy opposites, has been used by Miss Norsworfhy in an extensive study, and constitutes the regular test list prescribed by Pyle. The opposite of this list, i. c, bad, inside, etc., has been used by Bouser, by Miss Nors- worfhy and by Mrs. Squire. List V is proposed by Pyle as harder opposites for use with adults. List VI has been used by Bonser, by jNIrs. Squire and (with two changes) by Simpson. List VII has been used by Bonser, by Mrs. Squire and (with some changes) by Simpson and by Carpenter. It also appears in Thorudike's tests. Lists VIII and IX are two of four hard opposites useil by Simpson. The easy opposites test may be regarded as fairly well standardized so far as choice of material is concerned. But List III. which has been selected as best for younger children, will prove too easy for most /S's of 10 years or over, rwid we have no lists of moderately difficult and very difficult opposites that have been tested by very extensive experimenta- tion. The set proposetl by the author has been selected from the 50 terms iised by Ilollingworth, who. in turn, selected them from a list of 200 tested by Woodworfh and Wells and showing association times of from 2 to 5 sec. The aftemj)! has been made on the basis of the author's trials with college students and with the assistance of Dr. Ilollingworth to select 20 hard opposites that shall be relatively easy to score and that shall be of closely similar difficulty. Method. — Use the easy opposites for younger children, the more difficult ones for children over 10 or thereabouts and for adults. Instruct 8 as follows : "Each one of these cards has printed on it a word. As soon as I uncover a card, look at the word on it and then, quickly as you can, say aloud a word that TEST 34 : CONTROLLED ASSOCL\TION [447] 81 means just the opposite to it. For instance, if the card should have on it the word dirty, you would say clean. We will try three sample cards first to make sure you understand." With the hard opposites follow the procedure outlined for the part- whole test with regard to the use of the sample cards, timing, recording responses, etc. Variations op Method. — Follow the suggestions given for variations of method in the part-whole test, with due regard for changes in sample terms, etc. When the harder opposites are used, particularly with adults, the instructions may be altered to put special stress upon the giving of an exact opposite, i. e., the emphasis may be placed upon quality rather than upon speed of performance. Further, it is well to instruct aS' that opposites formed by the use of the prefixes M/i or in, or of the suffix less, will not be allowed, save when the root of the stimulus word is changed; thus, for in- stance, inharmonious would not be accepted for harrnonious, nor unsafe for safe, but harmless would be accepted for dan- gerous. The split-second watch may then be used to advantage by rejecting responses that are incorrect, and recording the time of various responses made before the proper one is given. To test the eifect of practise, the cards may be shuffled and the series repeated any desired number of times, as in the pro- cedure adopted by Hollingworth in his tests of the effect of caffein. Treatment op Data. — This may follow the directions given for the part-whole test. With older children and adults, how- ever, when the instructions have emphasized quality rather than speed, the scoring of responses should be decidedly rigorous, and it may be well for many purposes to permit only a single correct opposite and allow nothing for 'partly correct' re- sponses.* ■•Mrs. Squire counted as errors all responses that were not accurate and gave no credit for adverbs when adjectives were correct, nor even for approximate opposites, urging that "there is no mental test in which an approximate is less permissible." The important thing is, of course, that E should settle upon the type of mental activity that is to be demanded (speedy approximate re- sponses or rigorous precision) and then adjust instructions and scoring to measure this aspect of the associative process. In using the cards S2 [l-IS' ASSOCIATION. I.i:AI!XIX(i AND MKMOUY l\Ksui/rs. — (1 ) Aoriiis. 'VUc avciii^c (iiuc ol" i'es|)()iis(' of aduK colloji'e sludonis is given bv ^^'<)o(l^\•()l•lll juid >Vc'lls as 1.1 1 soc, V.Vj. .0-1, rangv 0.85 to 1.40, fov llic easy lis(. The uoriiis ob- taiiied by Tyle, Miss Norsworthy, Mrs. Squire and Carpenter are rej»ro(lueed in Tables 07, (18, ()<) and TO, respectively: these ligni-es may not be conii)ared directly Avilh one another on ac- count of differences in materials, scorin<>- and other conditions, bnt they will serve as bases for conclnsions with re<>ard to the dei)endence of performance in the test upon ai>e, sex and other factors. Pyle's norms represent the average number of opposites that could be written in (!0 sec. in a group test, usiug a list foi-mod of tlie opposites of List IV, above. Miss Norsworlhy's tigures refer to tlie number of correct associates to the opposites of List IV (her First List) and to List IV itself (her Second List), given by about till normal children of both sexes. Here the maximal possible score is evidently 20. Mrs. Squire's ligures show the average time in sec. and the average number of correct responses (rigid scoring) for small gronjts of peda- gogically unrotarded children for three separate lists (our Lists VI. VII and the oiiposites of IV, respectively). For the standards iirojtosed by her for each age, see the original article, pp. 5(K)-r)0('). Cari)enter's results are based upon what he describes as practically the same terms as our List VII. The tigures show the errors and the TABLE 67 Correct Astiociatcs Written in GO Sec. Opposites Test {Pylc) SEX AGE 8 9 10 H 12 13 14 15 16 17 18 ADULT Cases 33 65 60 61 72 65 61 40 33 17 22 62 Male Aver. 9.0 S.4 7.510.911.514.514.516.0 18.6 17.6 22.4 22.1 A, D. 3.3 3.0 3.1 2.9 2.9 4.5 4.3 5.2 5.3 3.3 3.2 3.3 Cases 33 56 77 65 74 73 58 49 48 27 26 85 Female Aver. 8.0 7.6 10.911.213.9 14.917.417.3 19.3,21.4 23.4 23.4 A. D. 4.0 2.9 3.1 3.0' 3.6 4.3 3.9 5.1 4.2 4.9 3.1 4.0 for individual testing it will be found convenient to list upon the back of each card the ditTerent words that are given by S's, together with the score previously determined upon for each word. For a more elaborate system of equating specxl and quality of work the reader may eonsidt Simpson (Iti, pp. 14, 10). The work of Woolley and Fischer contains many valuable suggestions for the evaluation of various responses in the opposites test (see 22a, ]>p. 210-221). 'I'KS'r ')4 : coX'ntoi.i.KK association 44!)] ^•^ time ill soc. obtained in individual tests iu which tiie terms wei-e ivad by E and responded to orally j>y H. His records are confessedly "of little value below age '.)." Since tli*; times were taken with an ordinary watch !ind merely exjiress the total time occupied by the test for ea<-h child, it is impossil)le, of course, to derive from them any jirecise idea of the association time, strictly speaking. TABLE 08 Correct AssocUttcs of a Possihle 20. Oppo-ntes Test {Norsworthy) LIST AGE 8 9 10 11 12 13 14 15 16 ADULTS Opp. of IV Afedian 7.4 9.o' 9.912.5 13.5 14.0,14.5 15.0 1,5.5 20.0 P. E. 2.0 2.0| .10^ 3.0| 2.6| 2.5 2.3, 2.3, 2.3 l.(» IV... Median 8.7 9.5 11.5'l3.114.7,16.4 17.818.519.0 20.0 P. E. 1.4 1.7 2.2j 2.9 3.6 2.4 2.0 2.0 2.0 '■" TABLE GO Correct A.sfiociatcs and Times. Opposites Test (Squire) LIST VI LIST VII LIST IV (OPP.) AVEK. FOR ALL AGS Words Time Words Time Words Time Words Time 6 7 8 9 10 11 12 13 8.5 13.1 16.1 17.6 17.1 19.3 19.5 19.4 192.3 155.6 110.3 103.7 87.1 79.3 81.2 72.5 11.3 15.0 17.1 16.5 17.7 19.0 19.2 19.2 143.3 137.7 98.2 101.1 87.0 102.0 85.5 61.5 11.6 14.5 1(3.7 17.7 18.2 19.6 19.2 19.2 120.0 117.6 104.9 98.4 76.2 68.3 63.2 65.0 10.5 14.2 16.6 17.2 17.6 19.3 19.3 19.3 151.8 l.'56.9 104.8 101.7 83.4 83.2 63.2 66.3 TABLE 70 /■Jrrors aiifl Times. (Jpposites Test (Carpenter) AGE 7 8 9 10 11 12 13 14 Cases 7 118 4.1 19 118 4.2 46 108 4.5 50 101 4.7 41 98 4.8 44 82 3.8 58 79 3.6 49 Aver. Time Aver, Errors 71 3.2 S4 1150] ASSdC^lAPlON, l.KAUNMNO AN'P ISIEMORY [-) l>(/icii(Uiicc on (///(■. 'riio cvi(l(Mir(> from (lu'sc I'oiu' (ables, as well as from lUo rcsulls of WooUey and Fisrlior, Ihouiih iio( iiuaranfooiuii in cvi'iy insfanco a uniform im-roaso in (juanlily or (inality of pcrfoi-mance from yoar (o year, un- doubtedly justities tlie nvnerali/.alion that both speed of tiud- iug oi)posites niul accuracy of ilie responses made, improve with age. The failure of this tendency to ajipear in the tables publishi'd by Uonser is appariMilly due (o the iuadtMpiacy of his test material for briugiug' out the abilities of the older pupils. (o) Depcnth'in'c on scr. in all comjiarisous of groups the superiority of females over males is readily noted. Thus, girls surpass boys in 10 of Tyle's 12 age-groups; llollingworth (9) found women faster than men in namiug oi)])osites both before aud after jtractise; Kurt and iMoore report that in one group only -\h- and in another only 42.2 per ceut. of the boys reached the median mark of girls; lionser found girls superior to boys in every school grade, though it is signiticaut that in most of his groups more boys than girls were found in the highest quartile. Wool ley and Fischer T'eport a slight superiority of girls, but add that "it is too small to be considered very sig- nificant." (4) Dependence on school (pride. Bonser found a general progress from grade to grade, coupled with a decrease of varia- bility; his results are somewhat afl'ecied by the fact that his test was too easy for use much beyond the OA grade. On the otlier hand, the curves of distribution published by Chambers show that it is impossible to distinguish 7th grade and 8th grade pupils by their curv(>s of distribution in this test. (5) Dependence on inteHi(/ence. With the exception of Win- teler's conclusion (based on the study of only 8 boys, divided into two contrasted groups) that the op})osites test is less well adapted than other forms of controlled association to reveal differences in intelligence, the general opinion of experimenters is decidedly favorable to its use for this purpose. Mrs. Squire found that pedagogically retarded pupils show a distinctly lower general average performance with more irregularity in speed and quality of Avork than the unretarded. Bonser ob- TEST 'M : CON'I'UOLIJCI) ASSU(JIAT1(>N [4r)l | 85 (jiiiH'd :i ((jnchil i()(i of <)..S5 hclwccn (>|t|)Osilcs ;ni(i I Ik; ;ivcra|^e Hhindiii}; in nil llie tests used by liiin to nicsisiir-c jihilify t(j rea- son, and d('il;ii(!S tliat "tlu* <)i)posit('S test seems to be a lest of rather sii|)eiior- inei-it as .i, siiif^hi test for this J4('iier;il foiiii of iiient;il nbilit.v." In Simpson's interesting^- stndy of two «-on- ti"isted f^roiips of adnlts (17 of superior, and 150 of (piile in- ferior j^-enernl abilityj tlui two j;i-onps were completely sepa- i-ated by both the (;;isy iind tiie hard oj)posit(;s lest, i. c.^ no l»erson in the |;oorer j^roup did as well as tlie jtooresl in the good group. Simpson estimates tlie true correlation witli tlie intelli- gence of ])e()ple in general to be as high as 0.82 for tlie easy, and O.Dti for the hai-d opposites test. At Jiedford Kills Jteformatory, Miss Weidensall j-ejtoits a coi-relation of 0.70, I'.E. .01}, between rank-oi-dei- in ojjposites and the (slimate of int(;lligenc<' of the; women made, by the director of the Industi-ial School of the institution. The <-orre- lation would have been higliei- uiidei- moi-e favorabh; conditions for the testing and the (islimating. Again, the institutional women who had received the most schooling showed almost exactly tlie same ability as the ('incinnati 15-year-old working girls tested by Woolley and Fischer, whereas tlu^ J»edfoi-d Jie- low-Cirade Oroup (schooling less than Gi-ade 5J>) were de- cidedly infei'ior to the Bedford (Jrade (Iroup and to the Cin- cinnati girls. (6) The feehle-minded and delinquent. No feeble-minded child, according to Miss Noi-s worthy's figui-es, i-eached the median performance of normal children; only about one in a hundred were better than — 1 I'.E, and only about six in a liun- di-ed were better than — 2 JM<]. of normal children of their age. Reference has just been made to the comf)arison of delinquent women and school girls. Dr. Weidensall's results show also that the Bedford women, taken as a group, are slightly inferior to the Cincinnati 15-year-old working girls at the upjier quar- tile, 7.5 j)er cent, less accurate at the median and 20.2 per cent, less accurate at the lower quartile. It is of interest to note that the opposites test proved somewhat difficult of compre- hension for these *S"s, so that special explanations had to be contrived and repeated as well before the testing could proceed. so [452] ASSOCIATION^ LEARNING AND MEMORY (7) Dependence on practise. Hollingwortli (9) put 11 men and 8 women through 100 trials with opposites, and also tested their speed in reading the stimuli and the responses from a typewritten sheet. The results were : Initial naming (average of trials 2-6), men 113.5 sec, women 99.7 sec. Final naming (average of trials 96-100), " 36.3 " " 31.2 " Reading directly (average of 5 trials), " 18..5 " " 16.1 " It is seen that extensive practise increases the speed of the associative process markedly, but that even after 95 trials a considerable part of the time taken in the test is occupied by the process of association as over against the time needed in reading and uttering the words. The speed attained by indi- viduals in the test by reading correlates with the speed ob- tained in the regular test of naming the opposites by approxi- mately 0.60. (8) Dependence on fatigue. In the course of his experi- ments upon the effects of caffein Hollingworth (8a) was able to observe the effects of time of day upon the opposites test under unusually favorable conditions as regards elimination of the practise error. In preliminary experiments (tests made at intervals of two or three hours) and also in more intensive experiments (15 trials between 10.30 A. M. and 10.30 P. M.) there appeared a distinct reduction in speed of naming oppo- sites as the day passed, and the fatigue effect was more pro- nounced in this test than in any others that were tried. After the second trial the initial records of the day were never sur- passed. (9) Dependence on race. In opposites, and also in other controlled association tests (genus-species, part-whole), Pyle (13a) found negro children of both sexes less than half as efficient as white children. (10) ReliaMUty. Simpson found internal correlations between his various lists amounting to from 0.53 to 0.93 for his easy and to from 0.60 to 0.97 for his hard list. The test may, therefore, be regarded as possessing a good degree of re- liability, particularly since Simpson's lists contained some words of unequal difficulty and are presumably less well TEST 34 : CONTROLLED ASSOCIATION L^'^^J ^^ adapted for testing than the lists which have here been pro- posed for standard nse. (11) Various correlations. In the ease of the women at Bed- ford Hills Reformatory, whose Avork was decidedly slow on the whole, the correlation between speed and accuracy was so high (0.83, P.E. .03, for those who needed no help in reading or writing) that the scoring was finally done in terms of accuracy alone. Correlations determined by Simpson between the easy and the hard opposites and other tests were as follows : with the Ebbinghaus completion test 0.72 and 0.85, with memory for words 0.65 and 0.84, with the A-test 0.50 and 0.58, with memory for passages 0.50 and 0.70, with adding 0.56 and 0.70, respect- ively. These figures represent ''estimated true correlations for people in general," as based upon raw correlations figured for his own adult /8"s, corrected for attenuation and other probable sources of error.? Thorndike found a very high correlation, 0.90, P.E. about .05, between the capacities of twins in this test. Notes. — Special comparisons of different forms of controlled association have been made, among others, by Watt, Rusk and Winteler. The average association times reported by Watt are 1.864 sec. for part-whole, 1.454 for whole-part, 1.418 for co-ordi- nate, 1.548 for superordinate and 1.859 for subordinate rela- tions. General agreement appears in the conclusions reached by Rusk, who lists the several varieties of association tested by him in the following order, passing from the easiest to the hardest : whole-part and part-whole, co-ordination, free con- cretes, superordination, subordination, free abstracts, causal. Winteler concludes that when superordinate, subordinate, co- ordinate, species-genus and opposite relations are tried with school children, the first takes the most and the last the least logical power. The so-called 'B-method' of Ries, a test in which *S' is given a number of nouns representing causes and asked to name an- 'See the original article for the raw correlations and for correlations with other tests than those here cited. For Bonser's correlations, which refer to various special tests, see his monograph, p. 96. 88 [454] ASSOCIATION^ LEARNING AND MEMORY other noun representing a related effect, yielded in his hands extraordinarily high correlations with estimated intelligence, 0.85, 0.86, 0.91 and 0.94 in different groups. The method is endorsed by Meumann (11, 432f.), who also reports excellent results achieved with it by Oksala, in Finland. Meumann de- clares that the capacity to seek out causes or effects affords a decisive index of degree of intelligence in children of from about 10 to 14 3'ears, and perhaps older. We have made attempts to use this test in the educational laboratory, both at Cornell University and at the University of Illinois, and have en- countered so many difficulties in the preparation of material, and especially in scoring, as to render the method unsatisfac- tory, even for mature college students. Another relatively easy, though strictly controlled associa- tion test is that known as the backward-alphabet test. This has usually been conducted by asking S to name, or to write, as rapidly as possible, the letters that precede f, k, s, p, w, I, e, r, d, o, V, j, n, t, and h. For comparison, and to obtain a rough notion of >S"s familiarity with the sequence of the alphabet in general, this test might be supplemented by another in which /S* was required to state the letters that follow another series of 15 letters.'^ REFERENCES (1) A. R. Abelson, Tests for mental deficiency in cliildliood. The Child, No. 3 : 1912, 1-17. See also : The measnremeut of mental ability of 'backward' children. Bt^JPs, 4: 1911, 268-314. (2) H. A. Aikins, E. L. Thorndike and Elizabeth Hubbell. Corre- lation among perceptive and associative processes. PsR, 9 : 1902, 374-382. (3) F. G. Bonser, The reasoning ability of children of the fourth, fifth and sixth school grades. GohimljiaGonEd, No. 37. New York, 1910. Pp. 113. (4) T. H. Briggs, Formal English grammar as a discipline. Teachers College Record, 14 : 1913, 2,51-343. (5) C. Burt and R. C. Moore, The mental differences between the sexes. JEPd, 1 : 1912, 273-284, 355-388. (6) D. F. Carpenter, Mental age tests. JEdPs,4: 1913, 538-544. (7) W. G. Chambers, Individual differences in grammar grade chil- dren. JEdPs, 1 : 1910, 61-75. (8) H. L. Ilollingworth. The influence of caffein on mental and motor efficiency. ArPs{e), No. 22 (ColumUaConPhPs, 20: No. 4). New York, 1912. Pp. 166. (8a) H. L. Hollingworth, Variations in efficiency during the working day. PsR, 21 : 1914, 473-491. "On the use of this test, see Aikins, Thorndike and Hubbell. TEST 34a : ANALOGIES [455] 89 (9) H. L. Ilollingworth, Articulation and association. JEdPs, 6: 1915, 99-105. (10) E. Jones, Some results of association tests among delinquent girls. PsB, 10 : 1913, 78-79. (11) E. Meumann, Vorlesungen zur Einfiilirung in die exp. Pada- gogik, 2d ed., Vol. II. Leipzig, 1913, especially 418-433. (12) Naomi Norswortby, The psycliology of mentally deficient chil- dren. New York, 1906. Pp. 111. (Much of this material is also given in JPsAsth, 12: 1907-08, 3-17.) (13) W. H. Pyle, The examination of school children. New York, 1913. Pp. 70. (13a) W. H. Pyle, The mind of the negro child. School and Society, 1 : 1915, 357-360. (14) G. Ries, Beitrage zur Methodik der Intel ligenzpriifung. ZPs, 56 : 1910, 321-343. (15) R. R. Rusk, Experiments on mental association in children. BrJPs, 3: 1910, 349-385. (16) B. R. Simpson, Correlations of mental abilities. GolumhiaCon Ed, No. 53. New York, 1912. Pp. 122. (17) Carrie R. Squire, Graded mental tests. JEdPs, 3: 1912, 363- 380, 430-443, 493-506, especially 430-443 and 500-500. (18) E. L. Thorndike, Measurements of twins. Cohwi'biaGonPhPs, 1905. Pp. 64. (19) E. L. Thorndike, Educational psychology, 2d ed., New Yoi-k, 1910. Pp. 248. See also his Principles of teaching. New York, 1906, and his Introduction to the theory of social and mental measurements, 2d ed. New York, 1913. Pp. 277. (20) H. J. Watt, Exp. Beitrage zu einer Theorie des Denkens. AiGsPs, 4: 1905, 289-436. (20a) Jean Weidensall, The mentality of the criminal woman. To appear in EdPsMon. (21) J. Winteler, Exp. Beitrage zu einer Begabungslehre. EPd, 2: 1-48, 147-247, especially 207-239. (22) R. S. Woodworth and F. L. Wells, Association tests. PsMon. 13 : 1911, Whole No. 57. Pp. 85. (22a) Helen T. Woolley and Charlotte R. Fischer, Mental and physical measurements of working children. PsMon, No. 77, 18 : 1914. Pp. 247. Especially 213-227. (23) S. Wyatt, The quantitative investigation of higher mental processes. BrJPs, 6 : 1913, 109-133. TEST 34A Analogies.^ — In Test 33 the associations to be formed are left entirely' to /S"s choice, are unrestricted; in Test 34 the associa- tion is restricted to a single form of relationship throughout any one series. In the analogies test there exists restriction, but the kind of restriction varies from one stimulus to another ^The author is indebted to Professor D. Kennedy Eraser, of Cornell University, for the arrangement of this test. 90 [456] ASSOCIATION^ LEARNING AND MEMORY within the series of terms. The kind of restriction, moreover, is not indicated to aS* in the instructions, but is supplied to him by the test material itself, and must be apprehended by him from that material. Each stimulus in the series consists of three terms; the first and second terms illustrate the relation in question; the third term is the first of a pair which are to stand in the same relation one to another as the first and sec- ond terms. /S"s problem, then, is to find the appropriate fourth term. Because the relation varies from stimulvis to stimulus, the test is sometimes referred to as the 'mixed relations test,' as, for example, hj Woodworth and Wells (6), who say that it tests 'flexibility of mental performance' and also 'skill in han- dling associations.' Burt (2), from whom the term 'analogies' is borrowed, holds that the test involves "perception, implicit or explicit, of the relation and reconstruction of the analogous one by so-called relative suggestion." The test is recommended by these authors, as well as hj Wyatt (7), and it needs little trial to show that it has many possibilities, particularly in view of the chance that it affords of constructing series of stimuli of varying diflficult}-. Materials. — (1) For individual tests: Split-second stop- watch. Three sets of 20 cards each, affording tests of three grades of difficulty. Sample set of 7 cards for preliminary trials. Prepared blank for registering times and incorrect an- swers. (2) For group tests or variant form of individual tests : stop-Avatch or special seconds-clock. Printed forms for each of the three sets of stimuli, provided Avith spaces for recording the responses. The stimuli chosen for these series are taken from a large number of stimuli originally employed by Burt in Avork with the analogies test in England. They have been selected on the basis of fairly extensive trial with children and high-school and college students. If further materials are desired, E will find it more profitable to turn to the two lists of 20 stimiili each that are published by Woodworth and Wells (also repro- duced by Briggs (1) ) than to construct lists of his own, as only by actual trial can the feasibility of a given set of terms for use in this test be demonstrated. Method. — Show S one of the sample cards and instruct him as follows : "On each of these cards there are three words, as on this one. As you see, there is relation between the first and TEST 3J:A : ANALOGIES [^57j 91 the second word. You are also given a third word, and I want you to find a fourth word which shall have the same relation to the third as the second has to the first. Work as rapidly as you can, and say the fourth word aloud as soon as you know what it should be. Thus, in this first card the fourth word is what? In the cards that follow the relation does not remain the same as this one, but varies from one card to another." If this explanation seems sufficient, proceed with the other sample cards, saying: '"I Avill try these sample cards now to make sure that you understand." Follow the instructions given in Test 34, part-whole test, including the use of the warning 'now,' the taking of times during the sample set, the correction of Avrong responses, etc., save that it is unnecessary to record 8's response unless it be a word which is not provided for upon the prepared form. In ordinary testing E should pass to the next card whenever S is unable to give a response within 30 sec. Whether List A, List B, or List C, or some combination of them shall be used will depend upon >8"s age and ability and upon the time at E's dis- posal. Variations of Method. — Follow the suggestions for variant methods given for the part-whole test. Treatment of Data. — This may be based, in general, upon the instructions already given for the part- whole test and for the opposites test. The English investigators have attempted a somewhat finer scoring of quality of response than we have recommended. Thus, Burt scored 1 for each correct response, % for fair and % for poor responses, and for omissions. Wyatt gave 4, .3, 2, 1 or for responses grading from fully correct to omissions. This elaborate scoring is not needed for most of the terms in our lists, for in quite the majority of cases there is but one single correct response. In the other cases the use of the split-second watch in individual testing permits E to wait until the correct response is given and secure a direct measure of the time needed for this response, while the time at which the first (erroneous) response is made can also be put on record as an indirect measure of *S"s general accuracy. Results. — (1) A general idea of the times that may be ex- jjected by the use of these three lists may be gained from the 92 [458] , ASSOCIATION^ LEARNING AND MEMORY results that have been obtained by their use in a limited num- ber of cases (Table 71). The averages exceed the medians on account of the occasional ver^' long association times which appear with nearlj^ all S's. The average time reported for their lists by Woodworth and Wells (about a dozen college and graduate students, using the method of exposing the entire list) is 3.14 sec, P.E. .13, with a range for individual averages of from 2.33 to 4.40 sec. TABLE 71 Speed in. Sec. for Correct Responses in the Analogies Test (Eraser) LIST A* LIST B* LIST C Group Cases Median Aver. Median Aver. Med an Aver. College Adults High-school Girls 8 19 30 1.8 2.0 . 2.4 2.36 2.64 3.16 3.0 4.38 3.4 6.51 "Lists A and B, at the time these figures were secured, contained 25 stimuli each. The omitted ones do not alter the conditions enough, how- e^'er, to invalidate these records as norms. More extended use of the Woodworth and Wells tests by W. V. Bingham, to whom I am indebted for advance figures from the results secured with 200 freshmen at Dartmouth Col- lege, has jdelded the following percentile distribution for the analogies test (average time in sec. per response, based on two trials of 10 responses each) : Poorest 10 20 30 40 50 60 70 80 90 Best 5.49 3.98 3.69 3.31 3.03 2.85 2.67 2.52 2.36 2.06 1.35 (2) Dependence on age. While sufficient data are lacking to present figures for various ages, there is a clear difference in the speed of aS/'s of grammar-school, high-school and college TEST 34a: analogies , [459] 93 standing when List A is used. Similarly, List B, and more espe- cially List C, proves too difficult for younger aS"s. (3) Dependence on sex. Burt found an advantage of 15 per cent, in favor of tlie girls at the Wallasley School, Liverpool, i. e., only 35 per cent, of the boys reached the median perform- ance of girls. In other tests at the Holt Secondary School, however, the average ijerformances of the two sexes were vir- tually identical. No sex difference appeared in our tests of Cornell University students. (4) Dependence on intelligence. Wyatt, working with the time-limit method on groups, found that the analogies test afforded the highest correlations with intelligence of any of the tests he tried, save the completion test. His correlations amount to 0.62 in one and 0.80 in another group. Burt's tests at the Holt School gave a correlation between the results of analogies and intelligence of 0.50 in the individual test and 0.52 in the group test ; his tests at the Wallasey School gave again a correlation of 0.50 (see Burt and Moore). (5) Reliahility. Burt's figures show that the analogies test possesses a good degree of reliability, as its coefficient of inter- nal correlation figured in different trials 0.58, 0.71 and 0.92. Notes. — The analogies test appears to be better suited than other tests of association to bring out individual differences in quickness of adaptation to the task demanded. Thus, in the case of one high-school girl, the average association time for the first half of the list was 4.83 sec, for the second half only 2.19 sec. The inference that this ^ was naturally slow^ in adapt- ing herself to new situations, but was able to work efficiently when once adapted, was afterward confirmed by the reports secured from her teachers of her performance in her school tasks, especially in geometr3\ When /S's are tested by the standard method of securing the time for each response, it is often instructive to plot a rough frequency curve, with the second as a unit. A comparison of the distribution of the times for different /S's, as in the following sample, shows clearly individual differences in steadiness and consistency of performance as well as differences in general tendency toward fast or slow rates of mental activity. 94: [4G0] ASSOCIATION, LEARNING AND MEMORY ISaniple Distribution for Two College Students, List C SECONDS 1 2 3 4 5 6 7 8 9 10+ MEDIAN AVER. Subject D__ Subject G- 1 9 4 3 3 4 1 3 1 2 1 2 1 1 1 3 2.9 5.6 3.91 5.67 It is also instructive to make notes of aS"s general attitude to- ward the test, whether confident or hesitating, hurried or tran- quil, etc., and to compare these attitudes with the quantitative results. REFERENCES (1) T. H. Briggs, Formal English grammar as a discipline. Teachers College Record, 14 : 1913, 251-343. (2) C. Burt, Experimental tests of higher mental processes and their relation to general intelligence. JEPd, 1 : 1911, 93 112. (3) C. Burt, The experimental study of general intelligence. Ciiild Study, 4 : 1911, 14-15. (4) C. Burt and R. C. Moore, The mental differences between the sexes. JEPd, 1 : 1912, 273-284, 355-388. (5) R. S. Woodworth, The consciousness of relation. Essays, pJiilo- sophical and psychological, in honor of William James, 1908, 485-507. (6) R. S. Woodworth and F. L. Wells, Association tests. PsMon, 13: 1911, Whole Number 57. Pp. 85. (7) S. Wyatt, The quantitative investigation of higher mental proc- esses. BrJPs, 6 : 1913, 109-133. TEST 35 Controlled association: Computation. — The solution of simple arithmetical problems in addition, subtraction, multiplication, and division may be considered as essentially dependent upon the accuracy and rapidity with which the appropriate associ- ative processes are executed. Computation is, therefore, a test of controlled association in which the restriction of the associative sequence is complete, in which only a single out- come is correct. But numerous subsidiary activities are, of course, involved. Thus, the solution of arithmetical problems with the aid of paper and pencil demands, besides associative activity, both visual perception and motor activity, while men- tal computation imposes an additional tax by necessitating the holding in mind of the problem itself and of the various steps in its solution. TEST 35: COMPUTATION [461] 05 Because of this implication of perception, movement, atten- tion, retention, and perliaps other forms of mental activity, as well as simple associative activity, the computation test has been employed not merely for the special purpose of studying the nature and course of associative processes, but also for the more general purpose of investigating mental efficiency at large {geistige Leistungsfdhigkeit). Oehrn, for example, who was one of the first to use computation as a mental test, sought to study individual differences in the nature of associative proc- esses; Aikins, Thorndike, and Hubbell, Brown, Burt, Simpson, Hollingworth, and Krueger and Spearman to stud}- the correla- tion of specific mental functions; Thorndike to determine the relative influence of heredity and environment upon mental efficiency ; Reis to compare the ability of normal, paralytic, and hebephrenic children ; Jones to investigate the effect of bodily posture, Vogt the effect of distraction, and Hollingworth the effect of caftein upon mental efficiency; Winch and Starch to investigate the transfer of special drill. But the commonest application of the computation test has been made in the for- mulation of the curve of mental efficiency, or the work-curve (Arhcitskurve), with special reference to the influence of prac- tise, rest -pauses, exercise, and similar factors upon the mental efficiency of adults, and especially of children, during a school day. This use of the test is illustrated in the work of Aral, Bellei, Bischoff, Bolton, Burgerstein, Ebbinghaus, Friedrich, Heck, Heiiman, Holmes, Katzen-Ellenbogen, Kafemann, Keller, Kemsies, Laser, Lindley, Marsh, Martin, Ordahl, Robinson. Schultze, Specht, Teljatnik, Thorndike, Wells, Weygandt, Winch and others.^ Addition, multiplication, and both in alternation, have been more popular forms of computation than subtraction or divi- sion. With all four forms varied types of problems have been used. These variations in the arrangement of the test natu- rally affect its outcome. The most important types of test are illustrated herewith. Beside the types that are shown, Winch \\ major portion of these studies are the direct or indirect develop- ments of the special technique of the adding experiment as formulated by Kraepelin and his followers (see the various volumes of the Psy- cJwlogiscJie Arheiten and Kraepelin's summary (27). 96 [462] ASSOCIATION, LEARNING AND MEMORY and others have emploj^ed miscellaneous arithmetical prbblems,- Reis had his ^'s add mentallj^ for 1 min. by 7's or by 12's, while Holliugworth has used a form of test, also recommended by AVoodworth and Wells, in which a constant number is added (or subtracted) from a given list of numbers. Examples of Material Used in Computation Tests A 2 B 4 1 3 5 2 6 4 C 4 7 2 9 8 2 3 9 2 9 9 5 4 6 5 4 3 17 E 42 F 492 38382655139 +79 +763 6 9 7 4 T D 95799 86967 32687 1 ■ 0" G 93 68 41 2 etc. H 64293643194831457627 i 982 J 61 K 28 +38682725423585791858 —469 —27 X 8 S 9 L 363 M 47 N 948 7986 R 4)799 5 1 3 4 84799 95976 25 52 X6 XS9 X 579 X ^523 7 6 34797 5 2 1 5 97864 98945 P 428423995479253314325 Q 254) 4659234 ( X4 G 3 5 1 3 9 2 1 4 2 1 2 3 6 5 3 87824 68792 79867 88896 97745 39799 48970 89043 3 (3 9 8 6 5 1 2 67354 54628 91176 90253 3 5 2 7 2 =It is hardly necessary to allude to the development and use, particu- larly by Courtis, of special sets of tests for measuring the abilities of school children in the fundamental operations of arithmetic, the solving of arithmetical problems, copying figures and the like. Similar tests for algebra and geometry have been announced recently by other investi- gators. Tests of this variety are, of course, aimed at the determination of specific pedagogical attainments and dift'er, therefore, in scope and method from those here under consideration. For an account of the Courtis tests, with results of their application to a large group of school children, see Courtis (12). TEST 35: COMPUTATION [463] 97 EXPLANATORY AddUion A. Vertical series of 1-place numbers, arranged to avoid repetitions and pairs adding to 10. S's add continuously and drop bacli to units when eacli hundred is reached, or add by pairs, either orally or writing down the unit figure of each sum. The pairs are sometimes taken so that each digit is used twice, thus S adds 1 and 2, 2 and 3, 3 and 4, etc., and sometimes so that each digit is used once, as 1 and 2, 3 and 4, etc. Oehrn, Vogt, and others working under Kraepelin used columns of as many as 7,000 of such digits. Krueger and Spearman used 70, grouped by lO's as illustrated. Marsh used similar columns of 15 or 25 digits. B. Vertical column of 24 1-place numbers, using 1 to 6 only. Used by Jones, who had 8 add aloud while he himself followed with a check list. C. Horizontal series of 1-place pairs of digits. A modification of the Kraepelin 1-place series in order to make possible the examination of the accuracy of each addition. The unit figure of the sum is the only one recorded, as illustrated in the first four problems. Used by Schulze, and apparently also by Ebbinghaus and by Vogt. D. Twenty 5-place numbers. Used by Thorndike. E. Two 2-place numbers. Used by Teljatnik. F. Two 3-place numbers. Used by Kemsies for mental addition. G. Five 2-place numbers. Twenty such problems were given and 2 min. allowed for computation. Used by Tliorndike and by Aikins, Thorn- dike and Hubbell. Four longer columns (25 numbers in each) are used in the Woodworth and Wells constant-increment test. H. Two 20-place numbers. Used by Burgerstein, Laser, Friedrich, and Holmes. The last-named investigator published elaborate rules for the construction of these problems in such a way as to avoid the exten- sion of errors in 'carrying.' She used 4 blanks with 16 such problems on each blank, Suhtraction I. Two 3-place numbers. Used by Kemsies for mental subtraction. J. Two 2-place numbers, to be written on the blackboard (Teljatnik). Multiplication K. Two-place multiplicand, 1-place multiplier. Used by Kemsies for mental computation, and by Ebbinghaus for written group tests. L. Three-place multiplicand, 1-place multiplier. Used by Kemsies. M. Two-place multiplicand, 2-place multiplier. Used by Keller, and by Marsh with the digits 1, 2, 5, and 9 excluded. N. Three-place multiplicand, 3-place multiplier. Used by Keller for written, and by Thorndike for mental computation. O. Four-place multiplicand and multiplier. Used by Thorndike and others both for written and for mental computation. The multiplicand is usually a combination of (5, 7, 8, and 9 ; the multiplier of 2, 3, 4, and 5. P. Twenty-place multiplicand. 1-place multiplier. Used by Burger- stein, Laser, and Friedrich, with the restriction of the multiplier, in most tests, to 2, 8, 4, 5, or 6. Divisi07i Q. Three-place divisor, 7-place dividend. Four blanks of 10 problems each were used by Bellei for an hour's work. R. One-place divisor, 3-place dividend. Used by Kemsies for mental computation. 98 [404] ASSOCIATION, LEARNING AND ]\IE:kIORl' There are certain advantages and certain disadvantages in each of these forms of material. In general, E must select that form of test that best suits the conditions under which he works. Materials. — Stop-watch, preferably split-second. Printed forms, containing problems in addition and multiplication. For group tests, the special seconds clock. Five forms have been prepared for this test : others may be prepared by E as desired. A. Addition test : several thousand digits in vertical columns with a line separating each 10 digits, after Model A. This form may be used with children or with adults, and either for short series or for continu- ous adding, after the Kraepelin method, after the plan of Krueger and Spearman, or after the method of adding pairs. B. Addition test with 36 problems, patterned after Model G, but con- taining 10, in place of 5 numbers each. This can be used also for tests in which a constant number is added or subtracted. C. Addition test, patterned after Model C (Schulze's method), and specially recommended for younger *5"s. D. Addition test, patterned after Holmes, Model H, and virtually identical with the material used by Burgerstein, Laser, and Friedrich. E. Multiplication test, after Model P, as used by Burgerstein, Laser, and Friedrich. Method. — (1) General determination of S's ahiliti/ may be carried on with any one of the forms. The following general principles should be kept in mind: (a) Individual tests are usually more satisfactory than group tests, {h ) Any computa- tion work that is so easy that the mental operations can pro- ceed as fast as the results can be written (as Form C for adults) would better be given individually and arranged so that S may announce the results orallj^ and E check them otf upon a prepared key; and in general, care must be taken that the recording of results shall not fall to S unless it is certain that his associations will neither be delayed nor disturbed by the process of recording, (c) Group tests with competent ;8"s may be most satisfactorily carried on by the work-limit method with the aid of the seconds clock, (d) Group tests by the time- limit method should, as a rule, be terminated at such a time that the fastest >Sf in the group can no more than complete the task. In accordance with these general principles, Forms B, D and E Avill be found adapted for group tests or for individual tests TEST 35: COMPUTATION [405] 00 Avitli the recording of the figuring done by ;S' himself. E can take the time for performing any specified number of the prob- lems, or he may also, especially by using a split-second stop- watch, secure the exact time for solving each problem without interrupting >8"s work until the entire test form is finished. For the constant-increment test S is given Form B, printed side down. He is instructed at the signal 'now' to add a speci- fied number to each number found in the columns when he turns over the form. The numbers commonly used have been either 4 or 17. The test may be repeated with other increments, and these may be so chosen as to secure wide variations in diffi- culty, as by assigning easy constants, like 1 or 2, or more difficult ones than 17. Again, the assignment may be to sub- tract a given number.^ Form C alfords a particularly good test of skill and accu- racy in the addition of units, especially when conducted orally. Record the time for each row horizontally. Adults will make but few errors, and these they may be allowed to jjass over or to correct, whichever way they may prefer. Form A may be given by a variety of methods. In particular, *S' may write down the sums for each section of 10 digits and E record the time for each section until one page of the material has been covered : or ^S may add orally by pairs while E watches for errors upon a prepared check sheet, and also notes upon it the place reached by S at given time-intervals, as at each minute or each half-minute; the adding in this case nmj be done by either of the methods of grouping the pairs mentioned above (Explanatory, Addition, A). (2) For those who wish to arrange an experiment for the special determination of S's susceptiMUty to practise^ fatigue, etc., some suggestions may be found in the following develop- ment of the method of Kraepelin illustrated in the work of t^pecht and of Bischoff. To carry out this experiment fully, .S' adds by pairs, 10 min. per day, on each of 12 successive days.^ -Tor timing worlv by columns a convenient arrangement is to cut them out of the form and paste tliem singly upon small stifC cards. ^It would seem possible to condense this time, either by taking fewer days or by adding during several sittings on a given day, though It Is impossible to predict whether the results would then be comparable to those reported below from Specht and Bischoff. 100 [466] ASSOCIATION^ LEARNING AND MEMORY The pairs are added by the 1 and 2, 2 and 3 method and the unit figure only of each sum is written down by >S*.* A bell-stroke or other signal is given at the end of each minute, and >S^ marks by a horizontal stroke the point he has reached at the signal. On the 1st, 3d, 5th, 7th, 9th and 11th days there is introduced between the 5th and the 6th minute of the adding a rest-pause of 5 min. : on even-numbered days S adds directly through the 10 min. without pause. In carrying ont this special form of addition test, or in fact, in carry- ing out any test wbicli is designed to measure efficiency under various conditions — different 'periods of tlie day, after recesses, after gymnastics, after eating, etc. — it is evident tliat E must bear in mind the possibility that a number of different factors may enter to affect the performance, and that to measure any single factor, like fatigue, the influence of these other factors must be excluded or allowed for. The most serious of these disturbing factors are practise, excitement, ennui and carelessness. A common method for cancelling out practise is to divide (S's into two equivalent groups on the basis of a preliminary test, and to administer one set of problems early to the first, and late to the second group (if, for instance, fatigue is to be investigated), the other set late to the first, and early to the second group. In studying the work curve, some ^^'s have used computation both as the test and as the work to induce fatigue, practise, etc. ; others have used computation as a test of efficiency, but have allowed 8 to follow in the main the regular work of the school session. In the first procedure, computation (usually addition) is pursued more or less continuously for an hour, or even for several hours ; in the second procedure, the com- putation itself occupies but a short time, relatively, say from 1 to 10 min., and is repeated at intervals of an hour or more, while /S meantime takes up his regular tasks, indulges in physical activity, or rests, as E may direct. In illustration, Vogt, Oehrn, and other disciples of Kraepelin, have kept their S"s adding continuously for several hours ; Holmes used 4 periods of adding of 9 min. each, with 4-min. rest-pauses, Burgerstein 4 periods of 10 min. each, with 5-min. pauses. Typical illustrations of the second procedure are supplied by the investigations of Laser and of Ebbinghaus, who introduced 10-min. computation tests at the beginning of the school day and once an hour thereafter, and also by the studies of Heck and of Robinson. Ebbinghaus is inclined, however, to recommend 5-min. tests as being equally serviceable for the determination of effi- ciency and less likely to develop ennui and carelessness. Offner (35, p. 48) favors short tests for similar reasons and also for the partial avoidance of the practise-error. Treatment op Data. — Computation tests yield two measures of efficiency — speed (or quantity of work) and accuracy (or ^It would seem to the author much preferable to use oral adding, but here, again, it is not possible to predict what effect such an alteration of method might have upon the results. TEST 35 : COMPUTATION [467] 101 quality of Avork). Many investigators, partieiilai-ly wlien exam- ining the effect of practise, fatigiie and similar factors upon per- formance, have found it best to keep the two measures sepa- rate. Some investigators, like Teljatnik, have considered qual- ity of work only; more often, qualitative differences, being relatively small, have been disregarded and performance has been ranked by speed of work onh^ The combining of speed and accuracy into a single score representing net efficiency may be attempted by some of the methods proposed in Test 26. Or, an arbitrary penalty may be contrived for each error and the time consumed may be increased by these penalties. Thus, Simpson, who used material like our Example G, computed the final score of his >S"s by adding to their actual time 10 sec. for each error. An S who added 10 examples in 55 sec. and got seven ansAvers right and three wrong would then be given a final score of 85 sec. Quantity of work is indicated by elapsed time when using the individual method, and by the number of problems solved (sometimes by the number of figures written in the results) in the time-limit method. Quality of work is generally regarded as directl^^ propor- tional to the percentage of correct solutions. Inaccuracy is most often taken in terms of the number of errors committed, less often in terms of the number of errors plus the number of corrections made by S. The simplest, but the least desirable way to compute errors is to score one error for every wrong figure in the result. In the case of certain problems, however, a single error in computation may affect more than one figure in the result.^ For reliable results, these complex errors must be examined and the score adjusted to indicate exactly the num- ber of real errors of computation. For the special experiment patterned after Specht and Bischoff more elaborate treatment of data is called for. (1) The gain in sums added the 6th min. as compared with the 5th min., in its relation to the sums added the 5th min. (/. e., the per cent, of gain) is computed both for all the days with pause 'The problems in Form D (Example H) are inteutionally arranged to reduce tills error. 102 [468] ASSOCIATION^ LEARNING AND MEMORY and for all the days without pause, and the difference between these two relative gains is found. (2) The number of sums added in the first five and in the second five minutes, both on days with and on days without pause, is treated in the same manner. (3) The difference between the sums added on the 2d and the 6th minute on days with pause, taken as a per cent, of gain over the 2d minute, forms the coefficient of practise. (4) To find the coefficient of fatigue let F = the required coefficient of fatigue, P m the coefiicient of practise, A = the sums added the first 5 min. without pause, B = the sums added the second 5 min. without pause, and b = the sums theoretically added the second 5 min. under practise, but not under fatigue. A X (100 + P) 100 X (b — B) Then b = and F = . 100 b Thus, if P — 8.8, A = 1226, B = 1141, then b = 1333.9 and F =: 14.46. (5) The difference between the sums added in the 2d min. and in the 10th min. of days without pause, taken as a ratio to the 2d min., affords another, and in some respects, a better index of fatigue. (6) The total number of sums added the 1st 5 min. of all days gives T (total performance), which affords an approximate notion of >S"s ability to add. T is also made the denominator of a fraction, the numerator of which is the total number of additions made the very first 5 min. The fraction gives some indication of >8"s susceptibility to prac- tise. (7) Another index of susceptibility to practise is se- cured by taking the average of the gains in the first 5 min. from day to day as against the first day and figuring the dif- ference as a ratio to the first 5 min. (1st day). Characteristic results for all these values are given below. Results. — (1) Woodworth and Wells report the average time of college students in the Kraepelin form of adding as 107.2 sec, range 65 to 164 sec. The same authors report for the con- stant increment test (one column of 2.5 numbers) adding 4 : average 33.9, range 24 to 49 sec; subtracting 4, average 41.1, range 25 to 67 sec; adding 17, average 97.4, range 62 to 158 TEST 35 : COMPUTATION [469] 103 sec, with an average of 2.4 errors in the last form of test. These figures are based upon a very limited number of *S"s (7 to 10). In the author's laboratory, tests Avith college stu- dents in adding 50 sections of 10 digits on the Kraepelin form have yielded individual averages per section of from 5.8 to 13.7 sec, while the number of correct sections has ranged from 34 to 46. (2) In all computation tests, and particularh' in those em- bodying mental multiplication, there are marked indwidual differences in speed and accuracy, even among /S"s of the same age and same school grade. Thus, Schulze's best pupil added more than 5 times as fast as the slowest pupil in the same class. (3) Dependence on sex. There is evidently no decided sex difference in computation, since the results of various investi- gators are conflicting. Burt found girls slightly slower in multiplication ; Burt and Moore reckon that 65 per cent, of boys exceed the 50 per cent, record of girls in adding and G3 per cent, exceed the 50 per cent, record of girls in multijjlication. On the other hand, Courtis' New York results show that girls are slightly better than boys in the fundamental operations of arithmetic. Again, Miss Holmes found girls slightl}^ better than boys, and the conclusion of Fox and Thorndike is that the girls in the high school they studied were about 5 per cent, better than boys, though here there may have been a better grade of girls selected by the school. In the solving of arith- metical problems, however, where something more than knowl- edge of the fundamental operations is involved, the work of Courtis and of Thorndike (52) shows a superiority of boys amounting to an excess of some 10 per cent, in the distribution above the median of the girls. "Eoughlj-, boys are about half as far ahead of the girls in the same grade as they are of the boys in the preceding grade." Heck found that boys fell ofO more in quality of work in the afternoon session than did girls (4.25 vs. 1.96 per cent.) — a result possibly due to a greater carelessness on the part of the boys. (4) Dependence on school grade. When sufficiently large groups are compared, there is, of course, a perceptible differ- 104 [470] ASSOCIATION, LEARNING AND MEMORY cnce between the performance of one grade and that of the grades above or below it, but this difference is small in com- parison with the range of variation within any grade, and may, on that account, disappear when small groups are com- pared. Thus, the curves of distribution in adding reported by Chambers for 22 seventh and 22 eighth grade pupils can- not be distinguished, while Courtis generalizes results for his multiplication test by saying that "35 per cent, of any grade membership will exceed the average score of the next higher grade: also, that 35 per cent, of the grade membership will fall below the average of the next lower grade" (12, p. 450). (5) Dependence on practise. All investigators agree that practise produces a considerable improvement in all forms of computation, despite the fact that the associative connections concerned have been long established and often used. Holling- worth, who used the constant-increment test (adding 17 to 50 2-place numbers) found that, even after 35 preliminary trials, one of his groups reduced their average time from 102.7 to 61.2 sec. during 17 further trials, a reduction of some 40 per cent. Similarly, the 19 university students reported by Thorndike, who added daily for a week 48 columns of ten numbers, effected a median reduction in time of about 31 per cent., and in accu- racy- of about 29 per cent., although the total amount of time spent in the work was only about one hour for each S. Not all these *S"s showed such practise effects; for one or two there was no improvement, while one improved as much as 50 per cent. It is worth noting that practise-improvement is shown by those who stand high at the beginning of the work as well as by those who stand low then. The same result has been found also in tests of 29 boys in a New York City 4th grade school, where, according to Donovan and Thorndike, those most efficient at the beginning gained on the average as much or more (in gross gains) as did those least efficient at the be- ginning. Wells' tests of adults (56) lead to a similar con- clusion. On the whole, however, practise in adding tends to reduce somewhat the initial differences between the >S"s, whereas practise in mental multiplication seems not to affect much the relative differences between S's, from w-hich Thorndike con- TEST 35: COMPUTATION [471] 105 eludes that the abilities demanded in ]nental nmltiplication are more de])endent upon original cajjacitj' than are those de- manded in adding. Mrs. Ordahl found that j>ractise in mental multiplication produced a decided gain in speed without much im])rovement in accuracy, and she believes that the improve- ment in this operation resides more in the methods of handling the task than in the facilitation of the numerical associations themselves.** The question of the transfer of practise-effects in computa- tion to other forms of mental activity has been studied by Winch and by Starch. Winch was unable to decide whether special drill in computation produced an increase of skill in solving arithmetical problems; there appeared to be a transfer in some of the classes, but not in others. Starch found that a 14-day drill in mental multiplication developed an improve- ment of from 20 to 40 per cent, in other arithmetical operations, but had little effect upon auditory memory span. (6) Dependence on intelligence. Burt tested English school children, aged 12, to determine the number of additions or mul- tiplications correctly made in 10 min., and found a correlation with intelligence of 0.25 in addition and of 0.41 in multiplica- tion. BroAvn's results for a group of 39 girls, aged 11 to 12, show no correlation between school grades and speed or accu- racy of adding and a correlation of only 0.10 between speed of adding and estimated general intelligence : his results for an- other group of 40 boys of the same age show correlations of 0.28 between speed of adding and school marks, of 0.24 between speed of adding and estimated intelligence, and of 0.11 be- tween accuracy of adding and marks, with absence of correlation Itietween accuracy and estimated intelligence. Simpson used adding in his study of two sharply-contrasted groups of adults; the test separated the groups fairly clearly — only 10 per cent, of the poor group reached the median performance of the good group. Within the good group the results of the adding test correlated by 0.72 Avith estimated intelligence. (7) Reliahilitij. Save for Burt's fig-ures (0.50 for adding and 0.55 for multiplication), the internal correlations for com- "Consult her article for detailed tables and introspective reports. 106 [472] ASSOCIATION^ LEARNING AND MEMORY piitation tests show a good degree of reliability. Krvieger and Spearman, for instance, obtained a reliability of 0.76, Simpson of 0.76 for his good group, 0.90 for his poor group, 0.91 for both together. Brown, who worked with several different groups, reckons the reliability for speed of adding at 0.68 to 0.98 and for accuracy of adding at 0.30 when one application of the test is made, and reliability for speed at 0.81 to 0.99 and for accuracy at 0.36 to 0.74 when the scores represent amalga- mated results of two applications. (8) Correlations. Aikins, Thorndike, and Hubbell compared efficiency in adding with efficiency in the other 'association' tests (misspelled words, cancellation of two letters, and op- posites), and {hy a special method of estimating the index) found the quality of work in adding and quantity of work in associating correlated to a degree of 50 per cent, in 8th-grade, and 20 per cent, in Sth-grade pupils, and net efficiency in adding and net efficiency in associating correlated to a degree of 48 per cent. On the other hand, the percentage of error in adding and in the other association tests exhibited no correlation or one of but slight degree. Thorndike's study of mental resemblances in twins (47) showed a much higher correlation of ability in computation between twins than between siblings ; thus, twins aged 9-11 years revealed a correlation of 0.90 in adding, and 0.91 in mul- tiplication, and twins aged 12-14 years a correlation of 0.54 in adding and 0.69 in multiplication : taken collectively, the index of correlation amounted to 0.75 for the adding, and 0.84 for the multiplication test. Fox and Thorndike found that ability to add correlated to a fairly high degree, 0.75, with ability to multiply, but only to a small degTee, 0.20 to 0.44, with ability to solve fractions or to perform other arithmetical problems. They conclude that "ability in arithmetic is thus but an abstract name for a num- ber of partially independent abilities." These results do not agree well with those reached by Winch in his two studies of the transfer of drill in numerical accu- racy, since he found high correlations (0.68, 0.69 and 0.74) be- tween accuracy in computation and in arithmetical reasoning;-- TEST 35 : COMPUTATION [473] 107 Burris found that school grades in algebra and in geometry, as recorded in 19 rej)resentative high schools, showed, for nearly 1000 pupils, a correlation of 0.45. Simpson publishes the following ''estimated true correla- tions for people in general" Avith efficiency in adding : Ebbing- haus completion test 0.65, hard opposites 0.70, easy opposites 0.56, a-test 0.58, memory for passages 0.42, estimation of lengths 0.00. The extended series of intercorrelations found by Brown, which are in general much lower than those of Simpson, will be found reproduced in Simpson (41, 107f.) as well as in the original text (5, 309-313, 316). Krueger and Spearman found a good degree of correlation between adding and pitch discrimination (raw correlation 0.67, 'corrected' correlation 0.68, 'completed' correlation 0.80) and between adding and the Ebbinghaus test (raw correlation 0.79, 'corrected' correlation 0.68, 'completed' correlation 0.93). Hollingworth (20) has studied the etfect of a long series of trials (over 200) upon the correlations between adding and various tests; the following are tj'pical results: TAPPING CO- ORDINATION DISCRIM. REACTION COLOR NAMING OPPOSITES 1st trial ___ 20oth trial- .45 .57 .21 .16 .23 .15 .26 .76 .23 .76 (9) Relation of speed and accnracy. While it is doubtless true that, for a given individual working under constant condi- tions, an increase of speed tends to produce an increase of errors, it is equally true that under actual working conditions a given individual may show an increase of speed coupled with a decrease in number of errors. When individuals are compared, it is found that the faster *S"s are, on the Avhole, also the more accurate /S"s. In six of his groups Brown found cor- relations between speed and accuracy of adding ranging from 0.13 to 0.43, P.E.'s from .07 to .12. With small groups of col- lege students I have obtained similar positive correlations of 0.19 in the case of adding and as high as 0.86 for mental multi- plication. 108 [474] ASSOCIATION^ LEARNING AND MEMORY (10) Mental defectives. Eeis found that with paralytics and hebephrenics the average performance was less and the varia- bility greater than with normals. Similarly, Specht says that, though there appear decided individual differences in fatigua- bleness as shown by the adding test applied to normal indi- viduals, a still greater amount of fatiguableness appears when it is applied to patients in an insane hospital selected for their tendency toward easy fatiguableness in general. I have combined the data furnished by Bischoft" for 12 normal aS^'s and by Specht for IT normal and 6 insane >S"s when tested by their special form of the Kraepeliu addition test and scored according to the directions given above for that experiment. The results are given in Table 72. The differences between the two gTOups are readilj' obvious in Factors 8 and 9, Avhich reveal tendency to- ward fatigue, and in Factors 10 and 11, which show the total amount of work done. Analogous results have been reported by Katzen-Ellenbogen, who concludes that "the average curve of epileptics is decidedly different from the normal [curve] and characteristic of epilepsy." TABLE 72 Average Scores of 29 Normal and 6 Insane Su'bjects in the Addition Test (After Specht and Bischoff). Normals. Patients. (1) Per cent, gain Gtli over 5th min., with pause. . 14.7 20.6 (2) Per cent, gain 6tli over 5tli miu., no pause — 1.0 ■ — 3.9 (3) Difference between (1) and (2) 15.7 24.5 (4) Per cent, gain 2cl over 1st 5 min., with pause. . 4.4 — 0.3 (5) Per cent, gain 2cl over 1st 5 miu., no pause. . . . ■ — 5.0 • — 11.4 (6) Difference between (4) and (5) 9.9 11.1 (7) Coefficient of practise 9.9 13.3 (8) Coefficient of fatigue —12.6 —21.3 (9) Per cent, gain 10th over 2d min., no pause • — 5.6 — 14.S (10) Additions 1st 5 min. in 1st trial 172.3 87.5 (11) Total additions 1st 5 min. of all trials 3406.8 1157.6 (12) Progress of practise 11.1 2.0 (11) Miscellaneous influences. Hollingworth (19) used add- ing among other tests in his study of the effects of caffein and found that this drug produces pronounced stimulation in the processes of adding. Posture was found by Jones to affect the speed of adding; both children and college students could add somewhat faster TEST 35 : COMPUTATION [475] 109 (approximately 3 to 8 per cent.) witli the body in a Iiorizontal, than with the body in a vertical position. The effect of distraction hy concomiitant activities, e. g., the reciting of a poem, was found by Vogt to reduce very mate- rially (58.5 per cent.) the number of additions made by the continuous (Kraepelin) method, but to have relatively little effect upon the simpler process of adding pairs of digits. (12) Fatigue and other factors of the loork curve. The use of computation tests to determine general mental efficiency at different hours of the day, with special reference to the per- formance of school children under classroom conditions, rep- resents a special form of experiment that oversteps the bound- aries of mental tests in their diagnostic use. In Avhat follows, therefore, 1 have limited the treatment to presenting typical results and to pointing out certain important general prin- ciples that have been established in this field. '^ (a) General analysis of the work curve. In other tests (espe- cially Nos. 10 and 26) we have had occasion to refer to the fact that , attempts to isolate fatigue from other influences affecting a curve of work are rendered difficult because of the presence of other complicating factors. Of these, practise is perhaps the most obvious and influential. Investigators have added, however, numerous other factors, such as recuperation, adaptation, momentum, swing, or fitness for work, warming-up, and spurts of various kinds. Extended accounts of these fac- tors will be found in Meumann (33, II, 8ff. and elsewhere), Schulze (40, 320ff.), and particularly in the writings of Kraepe- lin (26, 27) and his students. However patent and plausible these factors may appear from observation of our daily activi- ties, it seems probable that they have sometimes been invoked in explanation of work curves when actual demonstration of their existence is difficult, if not impossible.^ In work curves obtained from school children it is certain that loss of interest, 'For a comprehensive critique of tlie experimental literature upon worlc and fatigue the reader is referred to Thorndilie (53). In my translation of Offner (35) there will also be found a more general discussion of the whole topic of mental fatigue. *Thorndike is especially severe in his criticism of the numerous lesser factors exploited by the Kraepelinian school. no [47(5] ASSOCIATION, LEARNIN(i AND MEMORY or ennui, and resultant carelessness complicate the performance seriously, and are often mistaken for the effects of true fatigue, in the sense of actual inability to work at a sustained level of efficiency. (h) Individual differences in the work curve. We have noted the i)resence of clear-cut individual differences in the speed and accuracy of computation ; there are also individual differ- ences in the course of the performance. Thus, both Kemsies and Keller conclude that mass results should be subjected to scrutiny to detect individual curves of performance if reliable information is to be secured concerning fatigue and overpres- sure in the schools. The recent work of Miss Martyn (32), similarly, has shown that the introduction of a rest pause may be favorable to some ^'s and unfavorable to others, and also that the effect of fatigue may be met and masked in some S'b by the presence in them of a strong permanent 'set' for accu- rate work. ''We may conclude," she says, "that fatigue cannot be invariably estimated by diminution either in sjieed or in accuracy of work, since habit and method of working bear an important relation to its manifestations" (32, p. 434). Again, the results obtained by Miss Arai and confirmed by numerous investigators make it fairly certain that the most competent workers are the ones least affected by fatigue. These individual differences in susceptibility to fatigue have tempted some investigators to sort /S*'s into certain groups or 'types' of workers. If by 'types' is implied that individuals can be sorted into 'water-tight com})artments,' the hypothesis must be regarded as of doubtful utility. Illustrations are seen in the work of Kemsies, Avho distinguishes between persistent workers who fatigue slowly and profit much by practise and feeble workers who fatigue quickly and do not profit much by practise. Meumann's own investigations lead him (33: vol. 2, pp. 10-11) to posit three types of workers (quantitatively re- garded) : the first type attains maximal efficiency at the start and thence decreases with numy fluctuations: the second at- tains maximal efficiency only after an interval (of a length depending upon the kind of work) ; the third attains maximal efficiency only after a long period, perhaps several hours, of TEST 35: COMMUTATION II work. The iirst typo, llion, is cJiaracterized by ra])iut wlien the same investigator had 1<; »S''s mentally multij)ly a 3-i)lace by a 2-plaee number eontinnously for from H to 8 hours, or (with ])auses for meals) from 4 to 12 hours, only 3 /8"s did as well at the end of their work period as when they had r(?sted; the results showed, as might be expected, a compound of gradually lessening practise and gradually increasing fa- tigue. Miss Aral, who mentally mnltii)lied 4-i)lace numbers for 11 or 12 hours at a stretch after practise-etlet-ts had been largely eliminated, found that the time needed to w'ork such examples was practically doubled at the end of eleven honrs.'-^ 2. Work with interrui)tions. ^Vhen repeated computation tests are made williin an hour, the usual result is a progressive increase in the quantity, but a progressive decrease in the (]ual- ity of the work. Burgerstein's figures (Table 74) furnish a typical exam])le of the results for four 1(1 niin. periods with 5 mill, resl -intervals between })eriods. TABLE 74 Eflicicnci/ in Addition cnid MuUiplicdtioii irithiii an Hour (Burgcrstcin) PERIOO --rHEsV.Vr ^^ \ -MBKR 0. KK.OHS PERCENTAGE OF ERROR I. 28,267 32,477 35,443 39,450 851 1292 2011 2360 3 01 11 III._ 3.98 5 67 IV. _ 5.98 Miss Holmes' results are similar, though, on account of com- puting errors of a different plan ('seriaF errors counting but "111 a tost oonductod iiiuler my direction and as yet unimblished, Mr. rainier, after ])reliniinary 1 rials to remove most of the etfeet of practise, worked at dillimlr mental multii)lit'atioii, l)(\t,'imiinf;- late in the evening after a day's university work and continninij; until the task became impos- sible. The cessation of work was not gradual (with ability, for examiile. 10 multiply 2-i)la('e numbers when 4-i)lace were impossible), but appeared as a sudden eollajise sueh that mental work of any sort was quite im- possible. TEST .J5 : COMPUTATION [479] 11.: as OIK? ci-ioi-j, I)('L- percentage of erroi- averaged }>ut l.'i, as against liurgcrstein's 3.'" The fOMuiion interpretation of results like liurgersteiii's lias been thai jjiaetise increases the speed of the work, while fatigue increases its inaccuracy'. But Ebbingliaus CiA, pp. 400f.j de- nies that })ractise could produce such marked increase of speed, and ascribes both the increase of speed and the decrease of accuracy primarily to increased haste and carelessness. (d) Effect of rent-pauses. ^Vhen, either from ennui or fatigue, efficiency tends to decline, a period of rest generally ('exerts a favorable effect. With school children, as would be expected, such a pause is favorable even after relatively short work, as is illustrated by the data of Table 7.5, which are de- lived by Burgerstein from Hchulze. The effect of rest upon efficiency in mental multiplication after some two hours' work lias already been mentioned with reference to experiments with college students. The tests made by Friedrich upon 10-year-old pui>ils and by Kraepelin upon adults (20, pp. 10-17) furnish similar evidence of the effect of rest-pauses. TABLE 75 Additions per rupil, ii:itlt and without a ReHt-Pauae (Burgcrstein-Hchu'ize) FIRST 25 MINUTES HEST-PAUSE SECOND 25 MINUTES First test 1067 1146 5 min. None 1088 Second test 1042 (e) Efficiency at different periods of the day. Typical in- stances of the use of coini>utation as a test for the fatigue- effects of the regular school program are afforded by the expe- riments of Friedrich, of Laser, and of Ebbinghaus. This method has been adopted in jiart to avoid the entrance of ennui and carelessness pre\iously mentioned. "Miss Holmes' analysis of tlie errors showed that tlieir increase dnring the hour was due primarily to increased inaccuracy in associative pi'oc- esses, rather than to increased fre mkmouv LascM-'s (csts, at hourly iuU>r\als. of ■_'•_'() pujiils (ai;iMi !) i:> vcars) in a lvouij>'sboi\u- liiir(/crfr tho -"tth ptM-iod, tho out- oonio is tho sanio as that of tho tosts I'or an luuir's tinio by Hurg'orstoin, viz.: a j>roiii'ossivo incroaso in sfXH'tl and dori'oaso in aoonrarv of i'onii>ntation. Efflciviwy ill Comiutldlii'ii iritliin a School Session {Laser) TEST AKTKK SCHOOL PKUIOU TOTAL Nl'MBKK 01'' KIOUKKS AOOKll TOTAL Nl'MHKK OK KKUOKS rKKOKNTAO.E OP KKIIUR I • 34.900 40.1U11 43.1-24 43.999 45.890 1147 1400 1713 179G IGGS 3.28 ir. _ . 3.59 III 3.79 IV 4.08 V 3.G3 libbiniihaus, Avho sonii'ht (o dolorniino tho desirability or un- desirability of a Tvlionr t'onlinnons sehool session in a (///;/(- nasiinii and hioher girls" srluntl at Ureslan, obtaincnl results ideutieal with those of leaser so far as the qualitative aspects are eom-erntHl, but dilVerin_i»' somewhat as regards (he af(er almost constant or (o fall otf sliuhlly (oward the clos(> of (he session. Friedvich's results lead liiui lo advise lighter work in the afternoon session, iiellei found that boys and girls aged TJ solved problems in division more slowly and less accurately in thi> af(ernt>on (ban in the morning. Marsh tested bul a few individuals, so (ha( i( is [U'obably unsafe to make induclions from his da(a. which vseemed to indicate a greater etllciency in adding a( noon ilian later in the day. and in multii>licatiou at betwcvn l.,'>() and »> [>. m. than at (! or at lO.oO p. m. JMiss Martin had (I N's add for IT) min. at 10. 1- and 4 o'clock, with the result that slightly more sums wtn-e com- pleted at 12 and somewhat fewer at 4 than at 1(1; the ditl'er- ences are, however, inside the probable^ error. The work of the first 5 min. was relatively poorer in the afternoon, due, she 'I'KST .'55: co.Mi'i'rA'noN I'^-'^M ' ' •"* III inks, (() ii Inter ciil i-juicc of Aiircf/unf/ ;il tliiil lime 'I'lu- tiiosl; snilliorilJil ivc hihoniloiy sliidy of cflicicrK y in <-jiIcijIhI ion at (lirfcTcnf |)Ci'io' his cxiicriiuciils iipfjii (h(i I'tt'dviH of ciilVciii io wiilcli Ihc (liiil.v curvoH of >S"s whoHe work WJiH done iiiidci- cxccpl ioiKilly {^ooiJ coiidifions as r('ji;ardH fdirni- iialioii of {)i'aclise cr-i-or. In I he iis(; of (lie coristMnl-iricrernerit lest (iuUUu^ 17 Io no 2-\>\n((^ iiiiinhcrsj :il S, 10, 12, .'i and 5.30 o'clock ihei-(; iipix'aicd |)i-o}^f(*Hsiv(; fali«4uc Jinioiinl inj^ to about a 2 per cent. Icnj^lhcnin^- of Ihe time ;il each li-iul, with a total lengthening' of 7.50 pei- c(;nt. in the case of 5 women and 10.5 |;ei- cent, in Ihe cas(; of 5 irien. In fni-thei- iis<' of the same test in ii more inlensivc; exijei-iment (15 li'ials between 10.30 A. M. jind 10.30 1*. M.) there appeared, again, a lenglliening of abont 10 i)er cent, toward tlie end of the day. Jleck tested 1153 New York school children ( 18j and 573 Ijynchbnrg, Va., children (17j \villi;i modilical ion of the Courtis tesfs for llie fnndamental a I'i linnet icai opei-at ions. The New York tests lasted 10 min., those at Lynchburg 25 niin., and they were disti-ibnted ovei- various periods of the s<-liool session, particu- larly at 0, 11, 1 and 2.:{0 (j'clock. The general result was an inci'<;as(! in (juanlity and a deci-ease in quality toward the close of the day; at New York, for instance, quantity increased by 1.57, 1.04 and 2.30 jjer cent, in Ihe 2d, 3d and 4th periods, whik (pialily deci'eas<;d by 1.5], 1.41 and 2.2M jjer cent, in the corre- sponding periods. These differences are so slight as to be peda- gogically negligible, in 1he oi)inion of Heck. The inferior qual- ity of with either form of material make a first trial with the left hand; follow Avith a series of 5 to 50 trials with the right hand,' then return to the left hand for a final test. Note how much practise elfect has heen 'transferred' from the one hand to the other. Plot a graph to show the effect of practise, both upon the time and upon the corrective movements. Treatment op Data. — In the standard form of test J^ has available 6 records. Several possibilities appear: aS"s may be compared with respect to (1) their 1st trial, (2) their 6th trial, (3) all 6 trials taken collectively (sum or average), or with re- spect to their rate of improvement, by computing the per cent, of gain either (4) in the Gth, compared with the 1st trial, or (5) in the average of the last three, compared with the average of the first three trials. The third method Avas found by Burt to yield the best correlation with intelligence. On the other hand, the 2d method would seem to have some merit, since Yoakum and Calfee conclude that "the time consumed in the first trial is an individual variation; that of the last [Gth] more nearly represents the individual's place in the group." Until we have more investigations on this point it would be better for E to try more than one method of ranking iS"s and to select the one which gave the most favorable results. Eesults. — (1) The best norms for the patterns are supplied by the results of Yoakum and Calfee, embodied in Table 78; results from a more limited number of college students with the star test are shown in Table 77, and for other groups in Tables 79 and 80. Effect ol Practise TABLE 77 on Hpccd ill Mirror- (Whipple) Drawing. Cotlegc St a dents NUMBER 1st LEFT 1st RIGHT 2d RIGHT 3d RIGHT 4th RIGHT 5th RIGHT 2d LEFT Men Women __ 11 23 169 149 127 127 108 87 96 76 80 67 67 67 88 74 ^If desired, the C patterns may be used, turned to bring the other edges at the baclv, in order to provide drill without direct repetition of the same diagram. 124 [490] ASSOCIATION^ LEARNING AND MEMORY (2) Individual differences in performance are striking; thus, in the star test the time consnmed in making the first tracing ranged, in the author's tests of 34 students, from about 50 sec. to more than 8 min. In the larger group of students examined TABLE 78 I'imcs, ill, fSec, for Mirror-Drawing {Yoakum and Calfec) GROUP TRIAL I II III IV V VI AVER. !-___ Median 243.0 121.0 93.0 82.0 68.0 50.0 110.33 M.V. 94.9 45.5 28.1 34.7 24.7 17.1 36.57 Slowest 517.0 245.0 205.0 180.0 158.0 113.0 210.00 Fastest 69.0 51.0 41.0 43.0 40.0 32.0 53.66 II_ Median 92.0 65.0 48.0 41.0 35.0 28.0 54.70 M.V. 64.1 33.9 26.6 19.3 21.9 14.2 27.40 Slowest 700.5 337.5 303.5 153.5 201.8 171.0 242.37 Fastest 31.5 23.5 19.3 18.3 17.8 17.0 23.95 III____ Median 167.5 105.0 80.0 68.0 56.0 48.0 97.83 M.V. 104.2 39.3 30.3 19.7 19.9 13.5 33.38 Slowest 752.0 277.0 270.0 175.0 121.0 105.0 193.33 Fastest 72.0 49.0 40.0 34.0 33.0 23.0 46.87 Group I comprised 30 elementary school boys. Group II, 52 women, and Group III, 51 men in the freshman class of the University of Texas. The averages for each group in each trial are not here reproduced. at Texas differences range from 31.5 to 752 sec, while the fastest college girl tested hj Miss Weidensall had a record of 18 sec, as compared with 2072 sec. for the slowest reformatory woman (Table 79). These differences, as inspection of the tables will show, are greatly reduced after a little practise. (3) Dependence on sex. That girls decidedly surpass boys and that women decidedly surpass men is shown in all the pub- lished results in mirror-drawing, with the exception of two groups reported b^' Burt and Moore, and in them certain diver- gencies in method and in other test conditions offer a sufficient explanation of the apparent exception. Miss Calfee's averages for six trials give for the freshmen women 64.4 sec, P.E. 22,3, for the freshmen men 101 sec, P.E. 28.5. She finds that only G per cent, of the men reach the women's median, while 90.4 TEST 3G: MIRROR-DRAWING [491] 125 per cent, of the women reach the men's median. It is not only possible, but probable, that this sex-difference is in some part due to greater familiarity of women with the use of the mirror. Burt believes that there is also an innate sex difference at work. (4) Dependence on practise, [a) General practise-effects. The tables given above show that even a single trial produces a decided reduction in time: the median time for elementary school boys, for example, is cut in halves in the pattern test, while that for men and women is reduced one-third by the first trial. (See Fig. 65.) The long practise experiment con- ducted by Starch with the star test shows (Fig. 66) that the reduction is rapid at first, then slower, and that maximal speed is not attained for a long time, apparently not until some 90 trials (Starch's curve represents a series of 100 trials, one per day). (6) Individual differences in practise-effects. Practise curves compounded of the performances of a group of ^'s show a smooth drop (see Fig. 65), but the curves of individual >S"s are not necessarily^ of this form : on the contrary, it is possible, as- Yoakum and Calfee have shown (22, p. 290), to separate >S's into groups that show the 2d trial slower than the 1st, or the 3d slower than the 2d, etc. These investigators summarize these facts by saying : "Some ^"s gain control of the situation by a fairl}^ regular procedure; others temporarily lose control at some point in the series. The majority of the latter lose control at the fourth or fifth trial in a series of six tests." It follows that the rank-order of ^'s in any one trial does not correlate perfectly with their rank-order in any other trial ; actual corre- lations computed by Yoakum and Calfee between the first and subsequent trials are 0.79, 0.76, 0.74, 0.64 and 0.59 for the 2d, 3d, 4th, 5th and 6th trial, resjjectively. The correlation be- tween the first trial and the avferage of all 6 trials is given by them as 0.93. (c) Cross-education. A considerable amount of practise gained with the one hand is transferred to the other (unprac- tised) hand. Thus, Starch's 100-day practise with the right hand effected an improvement in it of 92 per cent, in accuracy and of 84 per cent, in speed. A single left-hand record, made 12G [492] ASSOCIATION, LEARNING AND MEMORY E \ •tMW T O ,, ^ no ^ \ V \ \ V '^ \ -^ 1 3 N ^ ^ FIG. G5. THE EFFECT OF PRACTISE DURING SIX TRIALS UPON THE AVERAGE SPEED OF MIRROR-DBA WING. (Froui Yoakuui aiid Calfee) The abscissas represent the six trials, running from left to right. The orclinates represent time in sec. 'E' is the curve for the 30 elementary school boys. 'B' for the 51 freshmen men, 'G' for tlie 52 freshmen women, 'T' for the 103 freshmen collectively. TEST 36: MIUROU -DUAWIXG [493] 127 at the expiration of this period, showed, in comparison with a single left-hand record made before practise began, an improve- ment of 81 per cent, in accuracy and of 85 per cent, in speed. There is, however, nothing surprising in this so-called 'cross- 66. THE EFFECT' OF LONG PRACTISE UPON MIEROE-DRAWING. (From Starch) education,' since the tracing of the star in the mirror depends primarily upon co-ordinations established in the central ner- vous system : in other words, the transfer is only an outwardly apparent transfer ; in reality, the same factors are at work in the control of either hand. (d) Persistence of practise. The effect of even a short period of practise in mirror-drawing is very persistent. Thus, Burt administered 6 tests in succession, during which the average speed fell from 103 to 39.5 sec. Twelve weeks later, two tests were given in succession ; the average speed developed was 34.5 sec. in the first, and 27.4 sec. in the second : in other words, the 7th test surpassed the 6th, made 12 weeks previously — a condi- tion found in the records of 16 out of 26 boys. The extent to which this persistence of practise-efifect was shared by Burt's 128 [494] ASSOCIATION^, LEARNING AND MEMORY ^'s is further indicated by tlie correlation of 0.52 between their standing before, and their standing after the 12-week interval. Hill's work (10b) shows that the skill developed by one trial a day, continued for 50 days, is so persistent that after an in- terruption of three years the first trial in relearning is as fast as the 32d and more accurate than the 50th trial of the original series, and that in four retrials a speed and accuracy has been regained that is equal to the final records of the original series. Mirror-drawing seems, therefore, to resemble neuro-muscular habits, like skating, typewriting, etc., in the manner in which skill once developed is retained with little loss over long periods, rather than the associative connections of ideational life with their relatively lesser persistence. (5) Dependence on intelligence. Burt reports a correlation between speed and estimated intelligence of 0.67, P.E. .07, for elementary school boys, and of 0.54, P.E. .14, for preparatory school boys. In another group of English school children a correlation of 0.60 was found, according to Burt and Moore. Miss Calfee, however, found no such relations in her group of elementary school children chosen to duplicate Burt's condi- tions : here the correlation with school grades was virtually zero (0.07) ; similarly, in the college students the correlation with grades was — .07 in the case of the men and 0.19 in the case of the women. The author was able to discern no constant differences between the work of five dull and five bright boys. (6) Delinquents. Comparative stud}- of the star-test (5 suc- cessive trials) with college girls, maids in college dormitories and girls at Bedford Hills, N. Y., Reformatory, conducted by Miss Weidensall, reveals a number of interesting results. From advance sheets of her manuscript,- for which I am indebted to Dr. Weidensall, I have selected data referring primarily to the time records only* (Tables 79 and 80) . The first of these tables shows that, both in the first and in the last trial, and whether 'Dr. Weidensall expects to publish also data for the number of errors (corrective movements) and for the degree of 'precision' with which the line is followed. Precision has been measured by ascertaining the total number of cm. in the contour of the star in which, for distances of 3 or more consecutive mm., the tracing remained within 2 mm. either way from the red line. TEST 3G: MIRROri-nUAWlNG [495] 129 maximal, minimal, median, average or ui)per or lower quartile is considered, the three groups are invariably arranged in the same order — students best, Bedford women last and the maids intermediate. The second of these tables shows that there TABLE 79 Times, in »S'cc., Used in the ^tar Test Jjy 36 College Girls, 16 College Maids and 69 Bedford Reformatory Women (Weidensall) COLLEGE GIRLS COLLEGE MAIDS REFORMATORY WOMEN 1st Star 5th Star 1st Star 5th Star 1st star 5th Star Fastest _ 18. 7. 36. 21. 59. 36. Uppor Q. 41. 17.5 54. 29.5 203.6 80. Median _ 66. 28.7 127.5 44.5 420. 117.2 Average 82.6 31.3 133.6 48.6 473.1 124. Lower Q. no. 39. 161. 69. 627. 148. Slowest- 252. 76. 409. 85. 2072. 436.* *Witli two failures in addition. exists a good correspondence between both the time and errors for the star test and the classification made by the institution into three groups depending on outlook for reformation : the differences are more striking in the first than in the fifth tracing. TABLE 80 Scores in the Star Test for Three Groups of Bedford Reformatory Women {Weidensall) FIRST STAR FIFTH STAR Time Errors Time Errors Most capable and promising__. Women with illegitimate chil- dren under 2 yrs. of age Backward and mentally feeble. Unpromising 320.9 562.9 610.5 117.7 211.3 264.4 105.4 123.1 127'.2 36.0 45.6 55.1 In addition to these quantitative results, the star test has proved to possess a value in a perhaps unexpected direction. 130 [496] ASSOCIATION, leauning and :MioiorvY viz.: as a device for sorting out ;S's of the unstable and less tractable type. On this point Dr. Weidensall writes : "This tost isolates better tlian any we liave tried at Bedford tliose wlio are incapable of snstained effort under difficulties. It isolated, of course, the low-grade feeble-minded, for, no matter how hard they try, they do not succeed in tracing a precise star. The epilei)tics have a characteristically bad time and their stars are all "knotted up' \Aith 'blind spots' where they weie caught and held indefinitely. Chietly, however, is the test of interest in the case of those who are bright enough to trace the star well, but too unstable to do so. These are invariably the girls who are dillicult to manage in the institu- tion. The tracing goes well enough until suddenly the pencil at some hard point starts off in the wrong direction. The subject then tugs and pulls, grows more and more irritated, disturbed and excited, makes big black circles and finally throws down the pencil and gives up. When calmed, praised and urged to try again, she will continue and usiially in the end draw a fairly good tifth star. This behavior in tracing the star is typical of their behavior in the institution when the pressure of discipline or re- sponsibility becomes the least bit too exacting." (7) Relation of speed and accuraci/. The curves reproduced from Starch show that practise produces a reduction in the number of corrective movements that parallels fairly closely the reduction in time. Correlations between time and errors obtained by Miss Weidensall are for the students 0.G3, for the maids 0.87, for the reformatory women 0.61. My own work with college students has given a correlation of 0.S6, P.E. .01. (S) ReUahilitij. Burt and Moore give this coeflicient as 0.52. The method used at Texas is evidently superior, since the coeffi- cient of relation between the first and second test, as above stated, amounts to 0.79, and thus assures satisfactory relia- bility. (9) Various correlations. Miss Calfee's tests of Texas fresh- men included three tests previously used by Burt, viz. : card dealing, card sorting and alphabet sorting. Correlations found by Burt between mirror-drawing and these three tests when applied to school children were 0.10, 0.31 and 0.29, respectively ; those found by Miss Calfee for school children were only 0.11, 0.20 and 0.06, for freshmen men 0.19, 0.11 and 0.22, and for freshmen women 0.37, 0.20 and 0.29, respectively. Save, then, for the last mentioned correlation, her figures are invariably lower than tliose of Burt. Other 'corrected' correlations re 'I'ES'I' .')0 : AIlltROK-DKAWIXG [497] 13L ported by Burt for mirror-drawing (average correlations for various groups) are: tapping 0.74, dotting apparatus 0.02, spot- pattern test 0.75, immediate memory 0.38, discrimination of pitch 0.G6, comparison of line lengths 0.55, esthesiometer 0.38, discrimination of lifted Aveights 0.30. (10) Qualitative aspects. Efficiency in mirror-drawing may result from the actual formation of new visual-motor co-ordi- nations (indeed, some aS"s after executing a number of drawings, find that, for a short time immediately thereafter, these new co-ordinations interfere with normal drawing or writing) ; but efficiency may also result, at least in the star test, from the voluntary inhibition of visual control in favor of kinesthetic control, i. e., by thinking the drawing of a star in motor terms, as if working with the eyes shut. Or, the hand-movements may be started in this manner and then carried out by visual con- trol from the mirror. Finally, adults occasionally control the drawing ideational ly, i. e., by applj'ing inferred projjerties of reflection by mirrors. It is evident that the existence of these qualitative differences may affect the test in such a way that the quantitative data for different ^"s may 'measure' different mental processes. Very slow ^*'s get 'caught' at certain difficult points of the drawing, where they make a long series of futile attempts to start in the right direction. Here the normal visual-motor control is too persistent to be readily broken or ignored. Notes. — A further study of the associative connections in- volved in mirror-drawing ma}' be made by the use of dot-tapping through a prism or of the various forms of mirror-writing.-^ For the first test, let *S' shut his left eye, and strike repeatedly with his right forefinger at a mark on the Avail or table-top, ''On mirror-writing, consult Abt, Allen, Downey, Laprade, Locbte, Or