370.15 T51 v.2 (3) qhorndike __. he Psychology of earnin ‘Ace. No. ace " mo ee) Keep Y our Card in This Pocket Books will be issued only on presontation of proper: library 8. Waless labeled othorwise, bookn may tw retained for two weeks. Borrowers finding books marked, de faced or mutilated are expactod to roport sume at library desk; otherwise tho last borrowor will Le: held responsible for all imperfections discovoru. The card holdor is responsible for all books draw on this card. Penalty for over-duo books 2c a day plus cust of Lost cards and change of residence munt be re ported promptly. Public Library Kansas City, Mo.anna se 2 Me rl LitanyTHE FUNDAMENTALS OF LEARNINGTHE FUNDAMENTALS OF LEARNING By EDWARD L. THORNDIKE AND THE STAFF OF THE DIVISION OF PSYCHOLOGY OF THE INSTITUTE OF EDUCATIONAL RESEARCH OF TEACHERS COLLEGE, COLUMBIA UNIVERSITY THE INVESTIGATIONS AND RESULTS REPORTED IN THIS VOLUME WERE MADE POSNIBLE BY A GUANT FROM THE CARNEGIE CORPORATION Bureau or Puniications Teachers College, Columbia University NEW YORK 1982Copyright, 1932, by ‘Teaxcumes Contnce, Covers UntvenstrrPREFACE Tins volume reports investigations covering a period of three years, dealing with the fundamental facts and forces in learning. Except as noted in the text, the writer is re- sponsible for the planning of the experiments, for the in- ferences drawn from their results, and for the methods of presenting these. But the execution of the work would have been impossible without the expert assistance of Dr. Ka Woodyard and Dr. Irving Lorge. Much of the ex- porimentation reported in Chapter X was done by Miss Mabel Wileox and Miss Parrish Little (Mrs. Jette). It was our original purpose to work at certain basic prob- lems until satisfactory solutions were reached; and in general we have done so. But the results suggested certain new problems of such great importance that we have thought it wise to study these also, even without expecta- tion of attaining a final settlement of them. As an indirect consequence of this, one important problem in our original set, that of the relation of the formation of conditional re- flexes to associative shifting of the ordinary sort, has been treated only very cursorily, FKowanp L. Tronnprxe Sune, 19.91CONTENTS OHAPTER PAGE I. Inrropucrion Se ee eee wee 1 If. Tus Invnvence ov tim Rererition ov a Srrvation. . . 2. 6 §1. The Purpose of the Experiments... . 6 §2. Experiments in Responding to a Length dy a an ‘pati mate of Its Magnitude... 6. ee sae 9 §3. Experiments in Responding to a Signal by a Movement 2... 0. 2 ee ee ee nh §4. The Nature and Measurement of Strength of Con- nection. 5. 2 ee ee be eee 18 §5. Experiments in Connecting Numbers with Words, Spollings with Sounds, and Words with Parts of Words 2. 6 6 ee ee 26 §6. Summary of Results Concerning the Effect of Repeti- tion of a Situation upon the Connections Leading from That Situation. 2. 6. 6 ee ee 62 UT. Tae Ixrovence ov tux Reretrrion oy a Connzorion WITHOUT BELONGING 66 ee ee 64 IV. Tse Ixvocever ov tie Repetirion or 4 Connecrion wintt BELONGING © 6 6 ee 78 si. The Problem... 6 6 0 ee ee ee 78 . A Typical Experiment . . . wee. 80 6 Further Experiments on the Repetition of ‘Connections 90 §4. The Amount of Strengthening Caused by the Repeti- tion of a Sequence with Belongingness, but No Satisfyingness 6... ee 108 $5. The Limitations of the ‘Action of Connections by the Mind’s Set or Adjustment... ........ 1 §6. The Relation between the Number of Repetitions ofa Connection and Increases in Its Strength... ©. 115 §7. Interference, Independence, and Reinforcement from Additional Connections with the Same First Term 121 §8. Factors Productive of Variations in Gain in Strength from the Same Number of Occurrences. . . . . - 126 viiviii CONTENTS CHAPTER Pade YV. Tae IvFivence or THe Iacpressivenrss oF tte Finsr Treat upon THE Gan IN StrenatH vrom A Given Numer of Ocourrencus ow THE CONNECTION © 6-1. ee eee 1k VI. Tae Invivence oF THE Distriurion or tan Ocourr af or A CONNECTION UPON Its Surenatir. 2 2. ee ee 148 VII. Tae Povarrry or Menrau Connwcrions . . 2. 2 ss Ihe VIII. Tae Inrevencu or Rererrrion ov A SERS UPON THE Ostisston ov Irs Inrermupiare Terms 2. 6. 2 1 ee way IX. Tae Inrivenop or tim Avrer-Kerects ov A CONNECTION 170 §1. The Inadequacy of Frequeney as a Cause of Learning 170 §2. Experiments on the Influence of the Aftor-lMffects of a Connection... . See ee eee 176 §3. The Law of Effect and Other Explanations of Tearing #181 §4. Further Wxporiments 2... 6. ee ee INE X. Tam Ivruvence or cite Avrer-Evrecrs or 4 Con nworion Waren Orrorrunrties ro Repeat or Revive Pr or ANY Equivavenr or Susstrrire vor Ir Arn Exenrpmp . . 207 XI. Tae Inrivence or Rewarps AND PuNIsHMENTS . . 6). 278 XII. Tue Paystonoaican Basis or vires Streneritentna or Cone wxotions by Trem Arrer-Merwors . . . Stet §1. Experiments on the Influence of the Time Following. the Action of a Connection before Its Situation by Responded to Again 2... ee ee ee OF §2. Experiments on the Influence of Disturhaneo of the Underlying Physiological Provess . 6 6. 006 6 6) BRE XIII. Reapnvess, Ipentirraumery, any AvAWaniuity . 6... 388 §l. Readiness 2... Se eee eee §2, Identifiability ©... ee §3, Availability ©... ee ee MG XIV. Tae Inrovenon or Menvan Systmms. BE §1. The Complexity of Mental Connections... . . 303 §2. Mental Systems Apparently Unexplained by Ordi- mary Connections . 6. ee ee ee ee OH §3. Sensory Systems . 2... ee ee ee ee BET §4. Instinct Systems 2... ee ee 888 §5. Customary Systems... 1... ee ee ee BO §6. Transcendent Systems . . . ‘ se eee BEG §7. Repetition and Reward versus | Pranseendent Systema 371CONTENTS OKAPTER 7 XV. Desmes, Purposes, Inrerests, ano Motives 27... # XVI. Associative Sinmeresa anp Hm Conpirional Ruruex “7 XVIT. Minor Exvsrixents, Coumeyrs, anp Succestions For Forvitmr INVESTIGATION ©. ee §l. Gradual Strengthening versus the “All or None” Prineiplo . §2. Failures of Satistiers to Strengthen Conn §3. Miscellancous Comments and Suggestions . 2... XVII. Apversn Eyinence ann ARGUMENTS ©... 2 2 ee §1. Alleged Evidence of the Poteney of the Repetition of a Situation 2 2. 2 $2. Alleged Evidence Against the Power of the Repetition of a Connection to Strengthen It... 2. . §3. Alleged Evidence Against the Power of Satstying After-Kffeets to Strengthen a Connection . . . §4. General Criticisms of the Law of Effect . 2... §5. Substitutes for the Law of Effect 2... §6. Doctrines of Indirect Action. a Mid of Representa- tions §7. Summary of See tions 3 to 6 op esl ot APPENDIX I, Nxreriments IN Resronping to A Leworit by AN Estmcare: ov Ips MAGNITIDH 6 ee Tl, Exrsiiaents In Responpina to A Sianal py Making a Movement. Expurents 67 22.0.0. 02 -0-- TLL. Exrreatents tv Connectina Nummers wirit Worps . . . LV. Matertans any Resuvrs por Experutent 38, wrrit vm Angel Worn Numuer Seams... ee . V. Exprninents IN Learning Cope Sunstrrution . 2... V1. MATERIALS Wok EXPERDMENTS IN THR Ponarity of Connec- TIONS cw ee ee ee VI. Taw Ineuuence or Parstacy anp ov Reomwor . . . . 1 + VILL Ksrosares op ti: Txenuence of Rewarps AND PuNIsH- ments in Various Experacents wir Animau LeARNING ix _ PAGE 392 401 413 413 415 421 430 430 431 430 461 44 478 480 489 497 526 551 571x CONTENTS APPENDIX PAGE IX. Tue Use or Frew Assoctations in Stipes ov Mev'ran DYNAMIGG 6 we ee ee ee wee OME ¢ X. Experiments wit Associative Siripring aNp Con- DITIONAL REFLEXES © 6 6 ee ee ee ee ee ee OO Lrst or REFERENCES... . . . see 620 INDEX ee ee ee O85LIST OF TABLES TABLE PAGE 1. Sample Tabulation of Responses for Ono Length... 2.2... 10 2, Distribution of Responses of T to Longths 5" to 11", in Successive 30's 12 2. Distribution of Responses of Br to Lengths 5” to 11", by Successive go | 486 4. Distribution af Responses of Sp to Lines 5” to 11", by Successive 30’s 488 5. Distribution of Responses of Wo to Lengths 5” to 11", by Succes- HIVE 80'S 491 G. Sumunary of Pirst 30 and Last 20 of 180 Reaponses by Br... . 403 7. Summary of First 30 and Last 30 of 210 Responses by Sp... . 494 & Summary of First 30 and Last 80 of 240 Responses by Wo... . 494 9. Summary of tho Firat and the Last Quarter of the Responses by Hf 495 10, Summary of the Firat ond tho Last Quarter of the Responses by To 495 11, Summary of First 80 and Last 30 of 70 Responses by D. . . . 496 12, Summary of the Main Changes in the Responses from the First to the Last Halves of Experiments 1 to 4... 0 ee ee 496 13, ‘The Responses of T in Drawing 950 Sets of Lines, Each Set Com- prising Four Responses, AIL Boing Without Knowledge of Results 16 14, Experiment G: Drawing 2” Lines with Byes Closed. Distribution of tho Responses at Each Sitting: Subject Ts... 2 ee 498 15. Experiment 6: Distribution of the Responses in Drawing 2" Lines: Subject T 16, Experiment 7: Distribution of the Responses at Fach Sitting. Draw- ing Linen to Equal 4 with yes Closed. Subject Ts... 500 17. Experiment 8: Drawing 4” Lines with Byes Closed. Subject TI: Dintribution of tho Responses at Eneh Sitting... 0. ee 18. Exporimont 9: Drawing 4” Lines with Eyes Qlosed, Subject ILE Distribution of the Responses at Hach Sitting... + 503 19, Experiment, 10: Drawing 6” Lines with Eyes Closed, subject “un: Distribution of Responses at Buch Sitting ©... + BOL 20, Kxperiment 11: Drawing 2" Lines with Hyos Closed, Subject R: Distribution of Responses at Each Sitting 6... - 505 21, Experiment 122 Drawing 4" Lines with Fyes Closed. Subject R: Dis- tribution of Roxponses at Each Sitting 506 periment 12: Drawing 6” Lines with Fyes Closed. Subject R: Dixtribution of Responses at Each Sitting. 6... 2. ee 507 23. Experiment 14: Drawing 2" Lines with Kyes Closed. Subject W: Distribution of Responses at Successive Pairs of Sittings . . . 508 24. Experiment 15: Drawing 4” Lines with Byes Closed, Subject W: Distribution of Responses at Ench Sitting... 0. 2 ee es 509 25, Experiment 16: Drawing 6” Lines with Byes Closed. Subject W: Distribution of Responses at Bach Sitting... - .. + > «510 xixii LIST OF TABLES TABLE PAGE 26. Experiment 17: Drawing 2” Lincs with Eyes Closed. Subject Wo: Distribution of Responses at Buch Sitting. . 2... 1. ee au 27, Bxporimont 18: Drawing 4” Lines with Hyes Closed. Subject Wo: Distribution of Responses at Wach Sitting... 2... eee nia 28. Experiment 19: Drawing 6” Lines with Hyes Closod. Subject: Wo: Distribution of Responsos at ach Sitting « «ss 2. = = = «+ 513 29, Summary of Experiments 5 to 19. Tho Influeneo of Hepetito Drawing Lines with Hyos Closed. 2... LSM 30. Summary of Distributions of tho Responses im the Mit and Last ‘Halves of Experimenty 8 to 19... . ONS 31. The Influence of Repetition: Drawing Linon with Hyer Claved. Mothod—Board 20. ee Ao 32, Summarized Distributions of Responses in tho Wrst ‘Thre snub Last Threo Sittings of Exporimont 20... 5... cee es BIT 33, Results of Cheek Experiment in Drawing 4" Tiney . 2... nao 34, Experiment 21: Drawing 20° Angles with Byes Closed. Subjovt ‘Ts Distribution of Responses in Groups of Two Nittinggs, in Groups of Four Sittings, and in Groups of Wight Sittings... 6... . S22 85. Experiment 21: Drawing 45° Angles with Byes Closed, Subject Ts Distribution of Responsos in Groups of Two Sittings, in Grom of Four Sittings, and in Groups of Night Sittings... 2. 6... 28 36. Experiment 21: Drawing 60° Angles with Kyex Closed, Subject ‘Ts Distribution of Responses in Groups of Two Sittings, in Groups of Four Sittings, and in Groups of Wight Sittings... 6... ith 37, Lxporimont 22: Drawing 45° Anglos with yos Cloxwd, Sul Distribution of Rosponses by Ritts, and by Groups of ‘Phres Sittings ee ee 38. Tho Responses of Subject 1 to and, are, in ix, the, be, nf, and hat Arranged by Succossive 1’... ee tt el le no 39, Word-Numbor Fxporiment: Tho History of Rexpomies Aerardingt to Their Froquoncy in tho First 10 Occurrences... . mM “40. Modian and Average Froquoney of Oceurroneo in Tater Tent af Re sponses Occurring Once, Twice, Throo Times, Four ‘Timm, Five Times, and Six Times, Respectively, in tho Pirwt Ten. . a6 41, Tho Average Numbor of Occurrencos, in Each Ten After the Mirst Twonty, of Responses Which Occurred 5 Timox, 6 Times, 7 ‘Tins, and s0 on in tho First Twonty ... te OB 42, Same as Table 41, but for Responses with a Frequency of 13 or More in the First 20, and Without tho Roport for the Responses ta AIL Bix Words Combined . . . 4 43. Froquenclos of Responsos to admire, bread, able, anewer, ute in Hue: cessive Tons. SubjerbO. 2... ; . . her 44, The Average Numbor of Occurrences in Hach Ten, Aftor thi Pirat Twonty, of Responses Which Occurred. Four Times, Five Tita, Six Times, ete., in the First Twonty .. . we 888 48. The Reduetion in Variubility in Rxporiment 23: Disteibutions of the Frequencies of tho Frequoncios 1 to 10. 0 ee ee 88TABLE, AG, 47. 2 61. 62. 6 ot, 15, 6, OT, 6R. 69, . Record of Subject And’s Responses to the Sound A 9, The Frequencies of the Response to ou and aw Which Were the = Sample Record (for Subject O), in Experiment 27 7. The Reaponses of Individual 2 in Experiment 23 . The Frequencies in the Second and Sixth 160’s of the Responses . Nonsensd Syllable and Number Pairs . LIST OF TABLES Tho Reduction in Variability in Experiment 24: Distributions of the Frequencies of tho Froquencies 1 to 10 in Successive Tons for admire and breed ‘The Reduction in Variability in Exporiment 24: Distributions of the Frequencies of the Frequencies 1 to 10 in Successive Tons for able, answer, ote. Most Froquent in the First Ton . The Intluones of Repetition upon the Connections Between Sounds and Spelling Tho Influence of Repetition in Completing Words... Completing Words: Tho Average Frequency, in the Last 8 Repetitions, of a Responso According to Its Length and Its Prequeney in the Firtt 8 ee eee Tho Reduction in Variability in Fsperiment 25: Distributions of tho Frequencies of the Frequencies 1 to 10 Tho Roduetion in Variability in Experiment 26: Distributions of tho Frequencies of tho Frequencies . The Reduetion in Variability in Experiment 27: Word Completions: Distributions of the Frequencies of the Frequencies: Which Wore tho ‘Throo Most Proquent and the Throe Least Pre- quent in the Firat 160 of Experiment 23... 0. 00.002 e The Frequencies in Various 160’s of the Responses Which Were the ‘Threo Most Frequent and tho Three Least Frequent in tho First 100 of Hxperiment A. ee Voreentages of Correct Responses for 100 Paits in the adopt Sories Comparison of the Number of Correct Responses for Neutral Con- nections with Seattored Connections with: (A) Neutral Connections with Sequential Occurrences; (B) Connections with Pleasant First Mombora; and (C) Connections with Unpleasant First Members . . Reaults of Experiment 34, with Serics Number Number 3586, Read ‘Twice to 8 Colloge Studests .. ee ee Uenulta of Experimenta 35, with Borie Number Number 3986... Sories Number Number G44. Frequencies of Each Two-Figuro Number vee eee ee eee Rewults of Experiment 36 with Series Number Number 644... Bacon Tost: Frequency of Orcurrenee of Ench Number from 10 to 99 Numbor of Correct Responses for 8 College Students and 81 Summor Behoot Students 2. 6 ee The Relation Retwoen Frequeney of Occurrence of @ Number in the Bacon Series and Number of Times tho Number Occurred (1) as a Wrong Response and (2) 28 a Correet Response, in 57 Individuals xiii PAGE 531 45 aT 49 50 53 53 54 54 55 55 57 84 90 93 95 97 99 103 « 104 ixiv LIST OF TABLES TABLE Pay, 70. The Relation Between Number of Occurrences of a Connection and the por mille of Correet Responses, Less an Allowaneo for (he Effect of Chance . 18 71. Correct Responses por Thousand, for Qeeurrenees of 0, in the bacon Series 2... ee 123 72, Pairs Whose Two Members Were Strong! " Equal Nuntbors of Repetitions .. .. . » 128 TOA. Phe Percents Correct for Pairs in the adopt Series Chon ta ave Words of Pleasant and Unpleasant Connotation, or Meaning Assoviations 2... sony 73. Number of Right Responses in eats After the force revered Series Le! 74, Scores in Completing Quotations When Words Are Omitted at the Bnd and at the Beginning... 0... 1. 75. Seores in Completing Words When Letiors Are Omitted at the ¥ and at the Begining... . 0... 0s ee UN 76. Lotters Written in 16 Successive Trials to Complete buat ss os VFL 77, Records of the Suecossivo Mstimates of the Aread af Certain Shapes 174 78, Frequencies of tho Right and tho Moxt Mroquent Weary Response by the Monkeys Skirrl and Solko in the First Series with the Yerkes Multiple-Choieo Apparatus 2... es 79. Te Influence of Hiffeet: Tho Accuracy of Katimaten wf Lewsthia 2 em, 27 em, a 10 em. Line Being Shown... fo ae 80, The Tuftenee of Moro Repetition: ‘Thy Avcurucy wf Kxtimater at Lengths 3 am, to 27 em, a 10 em, Line Boing Shown. 0... y Ist 81. The Tnflueneo of Meet: Tho Accuracy of Extimates wf Lenyrtha Bem. to 27 em. a 10 em. Lino Reing Shown . 5 Lo TsO 82, Tho Influence of Hifect upon Drawing Tinos Whee Blindfolded: ‘Che Percent of Right Responses in Barly sud Late Texts and Duving tho Training Itself... TN 83, Tho Influence of Hifect upon Drawing Linon When Diindfolled . 5 147 84. Tho Tufluenco of Repotition Along upon Drawing Lines When Hind foldol ... ee 1s 85, Tho Influcnco of Tifect im Six Subjects from the Group Which Made Zoro Tmprovemont hy Repetition Alone... . . . Th8 86, Drawing 3”, 4”, 5, aud 6” Lines, with No Sequences Correct Response in Wach Tost of 20 Trials, ren) . Drawing 3”, 4”, 5%, and 6” Lines, with No Sequence.; Suna of the Median Mrror (rogardlesa of signs) from 0”, 4", 5" uml 0", tn tenths of minch 2... re] 88, Numbor of Correct Judgmonty out of [10 in Succewive Pain at Training Series of Linos 3”, 314", 344", 2"... to I in length 196 89. Number of Correct Responses (out of KO) in Hatimatingg Lantte thy, 3 the Number of gs 3”, 3%", 4"... to 12”, in Succaswive Trinly sw ws . wT 90. Tossing Balls at a Horizontal Targut: Scorew Attained in ‘Tats wal Practices, 320 Practices... . a be MD 91, Tossing Balls at a LUorizontal Targot: Seoron Attuined in Toaty and Practicos, 240 Practices oeLIST OF TABLES xv TARLE PAGE 92, Number of Squeezes Correct (i.c. within certain limits of error) out of 36, in Suecessivo Trials with Dynamomoter: 14 Adult Subjects . 204 93. Records of Subjects Who Had 10 or More Corroet in the Initial Tost and 10 or More Correct in the First Half of tho Practice Period . 205 Number of Words Underseored in Each Position Before and Aftor Training ee ee 21g The Influence of Frequency of Underlining in Certain Positions in A I and A UL upon tho Tendeney to Paver Those Positions in BI and 1 LT, Frequencies for AU Positions in Training and Tests... . 214 46, The Changes in Underlining for the Unknown Words in @ I and C IL (Spanish Words) from tho Initial Test to the Trial up to Whieh the Frequency for Positions 1 smd 2 Wqualled or Exeoaded the the Froqneney for Positions 4and 5.0... ee 220 + ‘The Changes in Underlining for Unknown Words in @ I and 0 If from tho Tnitial Tosts to the Trial Up to Which the Frequeney for Positions 1 and 2 Equalled or Bxeeoded the Frequeney for Posi- tions 4 and 5, Supplement to Tablo 96... ee 221 The ufluonce of Repeatedly Completing 160 Words by the Addition of v after a, (after b, ote. upon tho Tondeney to Use v after a, 1 Ot. a 95, after 6, ote, in Completing Other Words... 230 AM, ‘The Relation Between Frequeney in tho Dictated Completions and Gain in Frequeney from the Warly to the Late Tost... . 231 100, The Enfluence of Froqueney on the Connections Between Features of Cards Each Showing Four 4” Lines and the Choieo of a Dine in a Cortain Position as the Longest... ee 235 101, The Intluence of Satisfying and Annoying After-Hffects on the Con- nections Between Peatures of Cards Each Showing Four 4" Lines and the Choivo of a Ling in a Cortain Position us the Longest. . 236 Samo ax Table 101, but for Tndividuals 42 to 47 - Same as Table 101, but for Individuals 48 to56. 2. 239 + The Influence upon Two Groupa of Satisfying After-Tffocts Com. pred with the Lifluencn of Frequency in the Case of Conn Between Cortain Peatures of Cards and Judgmonts Phat a Line ina Certain Position Ts the Longest oo ee 240 105, ‘The Frequency, Prequoney of Roward, and Percent Rewarded of the Kteaponses of ‘Trials 1 to 12, Uxeluding Lengths Under 4”... 247 106. Chanyes in the Strength of Connections of Groups RR, R, and L, Operating with Equi Frequoneies but Differing Amounts of Rowards oe ee ele Lk 248 107, Tho Frequency, Frequoney of Reward, and Pereont Rewarded of tho Keaxponses to Triaks t to 8, from 624" to WB"... ee 252 108. Clingex in tho Strength of Connections of Groups RR, R, and I, Operating with Bqual Frequencies, but Differing Amounts of, Reward, Wyn Lines oe ee ane 109, Sample of Data and Computations Used in Obtaining the Measuro- ments of Table 110, Individual Po. ee 255 110, The Tendoney to Use Increasingly, in Estimating Areas, Numbers Which Have Buen Rewarded... ses 256svi LIST OF TABLES TABLE PAGE 111. Frequencies Among Sixty Individuals (bacon Series) of Various Magnitudes of the A/B ratio... 2... 5. 2, Frequencies of Various Magnitudes of the A B ratio in the apes ment with the adopt Series... ss 113. The A/B Ratio in Bach of Hight Individuals in Hach of Fi ive ‘Rape ti monts 2. ee ee ee . 114. The Influence of Announcements of Right After uel Leh Right Response and of Wrong After Hach Wrong Response . . 7 279 115. Tho Influence of Announcements of Right After veh Right Reaponae and of Wrong After Wach Wrong Rexpouse . 2 6 6 ee ENT 116. The Influence of Announcements of Right After Bach Right Reaponse and of Wrong After Hach Wrong Response... Lote 117. The Influence of Announcements of Hight After Macl ch Riya Reapons its and of Wrong After Hach Wrong Response... ss ‘ eno 118. The Influence of One Announcement of Light Compared with “hat of One Announcoment of Wrong... . eee 2 UST 119. The Influence of Two Announcements of Right Compared with) Two Announcements of Wrong . . . wee . SHR 120. The Influence of Reward and Punishment in the Cue of ‘aman Adult Learning of Mazes... 1.04 - at 121. Responses of 13 Rats in Suecessive ‘Trials with Kuo’ 8 “Mattie Choice Apparatis. 2... . . noo . Tho Number of Rights (d), Wrongs w, and § Minute De ny tar) per Rut on Fach Day... 2. ee ee . a0 123, The Number of Rights (d), Wrongs (1), and ! site: Delay “can per Rat in Days 1 to 7 (first 35 trials) and in Days | ta 10 i OStrids). ee ee _ 07 14, The Bifect of Interpolating, After Bach Unit of Completions, 1 Unit of Chocking Numbers, Contrasted with the Hitect at Lnterpolating, After Hach 20 Units of Complotions, 20 Unita of Checking Numbers 325 125. Distribution of the Strength of the Tendency to Reapond to Cortain Words by Obviously Habitual Verbal Soquents . 2 6. nh 0H 126. Poreontages of Responses by Opponites. .. . : 127. Froquoncies (por mille) of Various Sorts of Connoction Ac Our Classification of tho Responses in the Frow Amiciation ‘Pest. 128. Words Given as Responses to Words from tho Kent Rotini List and to Cortain Others, Including Certain Prefixes, Suftixes, nud Numerals 129, Words and Numbers Used in angel Tout L with Number of Qeeurrencrt 180. Words and Numbers Used in angel Test UL with Number of Oe curronces. 2. tele et dalalale le let 181, Frequoney of Numbors in the ange? Hxperinent 182, Exporiment 88: The Number of Correct, Rexpannen for Words in thie angel Series Followod by a Singlo Numbur .. 2... 188, Experiment 38: Tho Numbor of Correct Responnvs in the angel Neri, in the Case of Words Followed by More ‘Thun One Number... 134, Experiment 90: Responses in Coding Letters of the Alphabet 4 Theew Equated Groups... 2. deol lela sling te . OaTABLE 1385. 136, 137. 138. 139, 140, 141. 142. 143, Ld. 145, 146, 147, 148. 149. 150, 11, LIST OF TABLES xvii PAGE Tho Relation of Strength of Connection to Number of Occurrences . 544 ‘Tho Influence of Primacy in tho Case of Drawing Tinos... . . 556 ‘Tho Influence of Primacy in Completing Words... 2... 560 The Infucneo of Roward and Punishment of One or Two of tho First ‘Three Responses upon the Sequent Connection: Crows and Canaries 573 ‘Tho First 24 Responses of Monkeys Sk and Sob and Pigs ¥ and M in the Yerkes Multiple-Choiee Apparatus... 0.0 ee 515 ‘The First 24 Responses of Monkeys Sk and Sob and Pigs F and M in the Yerkes Multiple-Choieo Apparatus... 00. eee 576 ‘The First 24 (or 56) Rosponses of Monkeys Sk and Sob and Pigs F and M in tho Yerkes Multiple-Choiew Apparatus... 0... SIT ‘Tho First 56 Responses of Monkeys Sk and Sob and Pigs F and M in tho Yerkes Multiplo-Ghoico Apparatus... 00... 578 ‘The First 24 Responses of Sk, Sob, Ora, F, and M in Problem 1 of the Yerkes Multiple-Choiee Apparatus... 2 0... ee 580 ‘Tho Lutluence of Roward and Punishmont of On or Two of the First ‘Three Responses upon the Sequent Connection, Monkeys and Pigs . 581 The Relative Inthuenco of Reward by Food and Punishment by Con- fnement in Experiments with the Yerkes Multiplo-Choies Apparatus 590 Sum of Scores for Hach Individual When the Responses Clearly Due to Habitual Sequences are Each Scored 1, and Those Clearly Not Due to Habitual Sequences are Bach Seored 5, Doubtful Responses Boing Seored, 2, 3,0r4 00000 00. ee ee es 600 ‘Tho Analysis of the Last Fifty Responses of the Subjects Who Answered the Questions After Experiment 101 thus: (1) No; (2) 5 BNO A) ae ee ee 608 Tho Responses of tho Subjects Who Answered the Four Questions after Experiment. 101 with Auswers Other Than: (1) No; (2) —} No; (4) —} With Resulting Action Rogarding Record... . 611 Tho Analysis of the Last Fifty Responses of the Subjects Who Anawerod the Pour Quostions After Exporimont 101 thus: (1) No; (2) 5 (8) Yess (4) Qualification of (3) to Practical No... . 614 The Analysin of the Last Fitty Responses of the Subjects Who Annwored thy Four Questions After Experiment 101 thus: (1) Yes; (2) Qualification of (1) to Practical No; (8) No; (4) —. . . . 614 ‘he Annlysin of tho Tast Pitty Responses of tho Subjects Who Auawered the Four Questions After Experiment 101 thus: (1) Yoo; (3) You; (2) and (4) Qualifications of (1) and (3) to Practical No 615 . The Rexponses of the Subjects Who Answered as Shown the Four Queations After Experiment 104, With Resulting Action Re- qurding the Record oe ee 618 . Types of Roxponses in Bach Suecessive Twenty-five Records for Buch Individual 2... ee + +» 620CHAPTER I Inrropucrion We ane concerned in this volume with the fundamental facts of learning whereby a situation which first evokes response A later evokes response B, different from A. The main questions which we have asked and tried to answer by adequate experiments are the following: (1) What happens when the same situation or stimulus or state of affairs acts repeatedly upon an organism which is, except for the action specified and for accidental varia- tions, the same? What does the mere repetition of a situa- tion, in and of itself, do to the mind? More particularly, does the mere frequeney of an experience cause useful modifications in the one experiencing it? And do the more frequent among the varying responses to this one situation gain in strength and increase their probability of oceur- rence in the future, while the less frequent ones weaken? (2) What happens when the same connection occurs re- peatedly in a mind which is, exeept for the connection speci- fied, the same? What does the mere repetition of a connection, in and of itself, do? In particular, does the vecurrence of two mental events in temporal sequence strengthen the tendency of the first of them to evoke the second? Our experiments show that mere sequence, in and of itself, is nearly, or quite, powerless, and that the second term of the sequence must in some sense ‘belong?’ to the first, be a sequent of if or a response to it, if learning is to restit, So we ask (2a) What does the repetition of a se- quence whose two terms ‘belong’? together do? What happens when we repeat again and again a connection in the sense of a situation and some one same response to it? (3) What effect has the after-effect or consequence of a 12 THE FUNDAMENTALS OF LEARNING connection upon it? What elementary and general facts explain the influence which rewards and punishments have upon the learning of animals and man? In particular, docs a satisfying state of affairs which follows or accompanics and “belongs to’? a mental connection really strengthen it, or does it only seem to; or is the difference a mistaken oh- servation of what is really a produet of frequency or re ceney or congruity, or freedom from inhibition, or consummatory quality, or something else? Our experi- ments prove that satisfyingness attached to a connection does strengthen it, and so they lead to a fourth question, (4) How does the satisfying food or approval or prog ress toward a desired end add strength to the connection to which it is attached? Jn particular, is its action direet, or does it operate by causing the animal to review or re. hearse the connection? Does its potency take effect then and there when it follows the connection, or later when the animal, on being confronted by the situation again, recalls or imagines that such and such a response to that situation produced such and such a satisfying consequence? Is the action universal, so that any satisfier will strengthen any connection to which it is attached; or may it be specialized, so that certain satisfiers benefit only certain sorts of cou rections? (5) How does the unfavorable effect of punishment come about? Is it the opposite or the negative of reward, an noying consequences weakening connections in the same fashion that satisfying consequences strengthen them, or does punishment operate in a different fashion of its own? Less important problems in the psychology of learning concern (6) the influence of primacy and (7) that of re. cency, (8) the so-called ‘‘law of the resolution of physio. logical states,” that an oft-repeated serios tends to lose ite intermediate terms, and (9) the polarity or unidireetioual quality of mental connections. The facts found have visanaia simportance for learning of two principles: the(identifiabilityiof the situa-INTRODUCTION 3 \ tion and the availability “of the response; and Chapter XIII is devoted to demonstrating and illustrating these two principles. Chapter XIII includes also a discussion of the principle of readiness, though we have little that is new to contribute in this case. Two other chapters are devoted to mental systems, and to associative shifting with special reference to the phenomena of the conditional reflex. Among the new techniques used in our experiments on the influence of the repetition of a connection is the presen- tation of series of two hundred or so pairs occurring from one to forty-eight times, making a total series of one to four thousand pairs. The pairs may be of any sort, and the subject may be set to attend to them in various ways, such as to copy them, or to listen to them, or to try to learn which seeond member goes with which first member of a pair, or otherwise. This technique offers a new and profita- ble approach to certain problems concerning the distribu- tion of practice, the influence of intelligibility on memory, the impressiveness of certain stimuli, and other topics. Sample investigations to show its applicability are reported in Chapters V and VI. One fact observed was of such general importance that we made special experiments to confirm it and others to ex- plain it. This is that the impressiveness of a first member in a pair (say a word-number pair) not only causes that word to he remembered, but also strengthens the conneec- tion leading from it. Suppose, for example, that the series contains kiss 38, devil 47, vomit 19, derive 26, hasten 53, and seldom 81, each pair occurring » times. Not only will the subjeets remember that the series contained kiss, devil, and vomit much better than they remember that it con- tained derive, hasten, and seldom; they will also remem- ber that kiss was followed by 38, devil by 47, and vomit by 19 much better than that derive, hasten, and seldom were followed, respectively, by 26, 53, and 81. The experiments made to confirm and explain this fact lead to hypotheses which, if further work verifies them, will be of great im-4 THE FUNDAMENTALS OF LEARNING portance for the theory of learning and the practice of teaching. The searching and critical analysis to which the concepts of learning, habit, and the association of ideas have re- cently been subjected, and the positive doctrines which h: been advocated, are reviewed in the light of our experi- mental results. We have not sought to reproduce in full the views of Watson, Carr, Peterson, Woodworth, Me- Dougall, Tolman, Cason, Hollingworth, Hobhouse, Kéhler, Koffka, Ogden, Lewin, Gengerelli, Van der Veldt, Warden, and others who have dealt with these problems. But we have tried to include the important facts and criticisms, and to work out and present a reasonable solution so far as is possible from existing knowledge. This solution takes the form of a new associationism, or better, since it differs deeply and widely from that older British assoviationism, of a new connectionism. As a test of the adequacy of connectionism, we have searched for the varieties or features of mental life which seem least amenable to explanation by it, and have tried to discover in what respects it is really inconsistent with the facts. One of the hardest facts to reconcile with a pay chology that limits itself to connections and readinesses, original and acquired by repetition or reward, is the tend- ency for a word to evoke its opposite. We have studied this case of apparent transcendent ac tivity as part of a general study of systems whieh organize and direct trains of thought. The results, presented in Chapter XIV, justify the connectionist’s faith. They also set the phenomena of the frec-association experiment ina new light, and should result in substantial improvements in the use of the experiment as an instrument of diagnosis or detection. Our presentations of all these matters are perhaps un necessarily burdened with descriptions of the experiments and tables reporting the results. Some of this evidential material has been isolated in appendices; and perhaps moreINTRODUCTION 5 ‘ of it should have been. The critical student will, however, desire to have all of this available, at least for reference; and in general he will prefer to have it along with the general argument and conclusions, rather than in appen- dices. The reader who cares for conclusions’ rather than evidence can attain his end by first reading the first ten lectures of Human Learning [Thorndike ’31], a popular account of some of this work and kindred matters, and then reading the present volume with the following omissions: Chapter IT. Section 5 IV. Sections 2, 3, and 4 VII. IX. From p. 190 to the end XI. XIE. Section T XIV. Table 128 XVIII. Sections 3, 4, 5, and 6 Appendices I, 11, UI, IV, V, VI, VIII, and IXCHAPTER IT Tue Inruvence or tim Rereririon or a Srruatiwn § 1. Trp purrose oF THE EXPERIMENTS As 1s well known, two laws or principles of connection forming have been proposed. The first, which we may call the law of exercise or use or frequency, asserts that, other things being equal, the oftener a situation conneets with or evokes or leads to or is followed by a certain response, the stronger becomes the tendency for it to do xo in the future. If, for example, by some means Rz is somehow made to follow closely upon S: a hundred times, the tend ency for S: to evoke Rz will become stronger than it was and may become stronger than some other tendency, SiR, which was originally stronger than it. This law has been generally accepted as a part of orthodox pay. chology, but it has recently been challenged. The second, which we may call the law of effect, asserts that what hap. pens as an effect or consequence or accompaniment or close sequel to a situation-response, works back upon the eonnve tion to strengthen or weaken it. This Jaw has not been so generally accepted; and among those who do aceept it in the form stated here, there would be a rathor wide variety of opinions concerning just what it is in tho effect or sequel of a connection that works back to modify that connection, and what conditions are necossary to enable it to do so. We shall subject both of these laws to new experimental tests, designed to discover more fully and oxactly what uso or repetition of a mental connection or bond does to that connection, and what certain sequels of a mental connee- tion do to it. The present chapter does not, however, deal directly with either of these laws, but with the still more 6REPETITION OF A SITUATION 4 Ton AR tal question of the influence of repetition of a situation. If the same situation is presented to a person again and again without any effort to alter or control thc connection or connections leading therefrom to a responsc or responses, what happens? In particular, if a given situation evokes two or more responses with varying fre- quencies, will repetition of the situation strengthen the stronger tendencies at the expense of the weaker, or will repetition merely maintain the status quo? Given a situa- tion or state of affairs S: to which (presumably becausc of minor accompanying circumstances in the animal) therc will be, in the first twenty occurrences, varied responses Ri, Re, Ra, ete., with frequencies of 16, 3, and 1, will the next tweuty occurrences tend to show Ri oftener than 1€ times and Rz and Rs less often than 3 and 1, or will the relative frequencies remain unchanged? Psychologists, in so far as they have faced this question at all, have usually taken it for granted that the first answer is correct, but nc proof of its eorreetness has ever been given. The theoretical importance of the question lies in its bearing upon the laws of use and effect and upon certain theories of facilitation and inhibition in the nervous sys- tem. If the second answer is eorrect, the law of use has a very restricted influence, and some law of effect must be true in eases where one response is selected for survival from many responses to the same situation. If the first answer ix correct, there must be some physiological selec- lion by repetition, It may be a change at the synapses whereby the repeated conduction of a stimulus over thc same path actively enriches the tendency so to conduct ai the expense of tendencies to conduct elsewhere. Or it may bo some sort of selective integration whereby the repeatec action of more or less of the associative system as a whol in a certain pattern depresses the tendency for it to ac in other patterns. Or it may be something else. But if ar animal which responds to S: by R: 90 times and by Ra 1( times, by this mere fact of relative frequency becomes ar8 THE FUNDAMENTALS OF LEARNING animal which responds to Si by Ri 91 times and WR: 9 times, and so on, until it is an animal which responds to Si by Ri 100 times and by Re zero times, there has been some active influence of the physiological basis of Si ~Ri upon the physiological basis of Sr~ Ra, inereasing the proba- bility of the former occurrence at the expense of the proba- pility of the latter. If the first answer is false, or if the quantitative amount of influence of relative frequency upon future relative fre. quency in favor of the more frequent. is small, we must. be suspicious of ‘drainage’? theories and the like, in general. The case of repetition strengthening the strong at the ex pense of the weak is a favored case for them. The importance of the question and answer for the prac tical control of learning will be only briefly mentioned here. In proportion as repotition per se increases the strength of the more frequent connection at the expense of the strength of ihe less frequent, mere practice may be trusted to produce learning as soon as the desired connec tion is put at a level of strength above any other one cou nection, provided that other things (such as intensity, re cency, or the effects of the connections) do not act against it in sufficient strength. In proportion as repetition per se simply confirms the status quo, mere practice will often be fruitless and attention should be given to avoiding practice in error from the outset, and to making the effeets of the connections such as will strengthen the desired one and weaken the others. Moreover, great practical importance attaches to this subsidiary question: “If repetition in and of itself does strengthen the stronger connections at the expense of the weaker, what is the amount of its infla- ence?’ Our experiments are such that if the first answer is shown to be true they will provide measures of the force of repetition.* Tt should be noted that wo aro not in this chapter studying the influence of the froquoncy of occurrences of a connection, excopt in go far as it iv a consequence of the frequeney of occurrence of a situation. We are hereREPETITION OF A SITUATION 9 In experiments to identify and measure the influence of repetition of a situation upon learning, the two desiderata are: (1) that the function or ability used shall be modifia- ble, and (2) that the consequences of the connections shall be prevented from having any influence, or be so arranged as to have an equal influence upon all the connections. The second is very hard to attain, and few, perhaps not any, of our experiments do attain it fully. Some of them were, in fact, designed to show the competing action of the repetition and the consequences of connections. The limi- tations, merits, and defects of each experiment may be left for later discussion. § 2. EXPERIMENTS IN RESPONDING TO A LENGTH BY AN ESTIMATE OF ITS MAGNITUDE Experiment 1 A set of strips of paper was made, alike in all respects save length, and containing 10 strips 5” long, 10 strips 514” long, 10 strips 514” long, and so on, up to and includ- ing strips 11” long. Hach strip had an identification num- ber. These were presented to subject T (the writer) in a random order. 'T estimated the length of each and entered his estimate after the identification number on a record sheet containing these, saying to himself always as he did so, Number is inches.’’* From 100 to 150 judg- controlling only the situation and letting it evoke such connections, and consequently responses, as ith will In later chapters wo shall report oxperi- ments upon the influence of the repetition of a connection, in which a situa tion and somo given response to it are presented ropeatedly in temporal soquones, with or without a certain relevance or unity or mutual belonging. * The purpose of this was to equalize the time and attention given to the connection between the identification numbor and oach judgment of ita length. However, tha identification numbers played approximately a zero part in the case of T in this experiment, so far as he was aware. In only four cases, all lute in the series, did tho identification number offer any suggestion as to the probable length of the strip. Numbers 105, 205, and 230 came to be thought of as 5” strips, as thoy in fact wore, Number 16 came to be thought of a» probably an 11” strip. Except for ten or fifteen Into judgmonts of these four strips, all the 2500 judgments scemed to represent connections10 THE FUNDAMENTALS OF LEARNING ments were made at a sitting. After the entire lot of 250 strips had been so judged, they were shuffled, the record sheet was filed, a new record sheet was taken, and the procedure was repeated. This was done in all ten times, the schedule of dates including twenty sittings scattered irregularly over a period of five days (Duc, 10-14, 1925). At the beginning of the period T knew that the series had 10 strips of each length and began at. 5” and progressed by 14", but he was not sure whether it ended at 11” or 12”. Not until the second round was well under way did he realize that since there were 250 entries on the record sheet, the series must end at 11’! During the entire period T never measured any of these strips or any other object, and acquired no experiences concerning lengths save that given by these 2500 uncorrected judgments themselves and by the fact that he knew the constitution of the series as just stated. Any given length, say 744”, was thus judged 100 times, so that we may tabulate the frequency of the connections as shown in Table 1. Tanis 1 SAMPLE TABULATION OF RESPONSES FOR ONH LED Situation Response Frequency in Succ 12 3 4 5 9 10 734" strip seen Saying 61* 1 1 734" strip seen Saying 62 L 74" strip seen Saying 63 734" strip seen Saying 7 1 5 1 734" strip seen Saying 71 22 1 1 3 4 4 I ot 744" strip seen Saying 72 1 6 5 14 5 5 6 7 & 734" strip seen Saying 73 5 3 4 2 8 B26 734" strip seen Saying 8 1 * The verbal and written form 61 was used for 634, 62 wax used for 024, 63 was used for 634, ete. Such tabulations have been made for vach length and studied in many ways to discover to what extent the re- sponse which has the greater relative frequency at the be- from the length of the strip to tho estimate. ‘This waa true also of nll the other experiments in judging lengths, to be reported lator.REPETITION OF A SITUATION i ginning, increases its frequency at the expense of responses of Jess frequency. Whatever the way taken to measure the waxing of the more frequent and the waning of the less frequent, the result is a negative. For example, con- sider Table 2 which gives the frequencies for successive thirties, omitting the first ten. Table 2 also includes the results for thirty repetitions at a date two weeks later (Dee. 31, 1925, and Jan. 2, 1926). The response which is the most frequent in trials 11 to 40 gains in trials 71 to 100 in seven cases, remains the same in three, and loses in fif- teen. On the other hand, the connections involved in esti- mating these lengths do not seem to remain in status quo, but. rather to change notably. Nor are these changes hap- hazard. For example, there is an inerease from trials 11- 40 to trials 71-100 in the use of 5 and 11 as estimates from 94 to 120 times, and there is a tendency to estimate the lines as longer in trials 71-100 than in trials 11-40. In the series of trials 101-130, done two wecks later, 5 and 11 are used 119 times; the lines are estimated very much shorter than they were in trials 71-100, and shorter than they were in trials 11-40 or in trials 41-70. Kxperiments 2, 3, and 4, reported in Appendix J, were of the same general plan as Experiment 1. Their results show (1) that the tendencies to respond to a length to be judged constitute a susceptible, variable condition, (2) that with continued repetition of the situation this condition becomes less variable, tending toward stereotypism, but (3) that the more frequent tendencies do not gain at the expense of the less frequent. § 3. Experiments IN RESPONDING TO A SIGNAL BY MAKING A MOVEMENT Experiment 5 Subject T (the writer), with eyes closed, drew a line to be as nearly as possible 2” long, then one to be 4” long, then one to be 6” long, then one to be 8” long. This series of12 Tasty 2 THE FUNDAMENTALS OF LEARNING DisTRIBUTION OF RESPONSES OF T TO LENGTHS 5" To 11”, IN succussivE 30’s Sit. Res. 51 5 52 5 53 5 61 51 62 52 Frequencies in 41 71 101 to to to 70 100 130 29 29 30 11 0 23 26 28 7 4 2 2 16 17:10 «14 11 16 4 4 5 1 16 15 17 10 10 11 1 4 2 1 22 22 4 6 13 3 2 1 3 6 19 19 13 6 5 15 1 2 2 1 10 4 13 19 6 5 6 19 14 i to 70 56 4 40 19 1 4 31 23 2 17 1 7 to 130 59 1 Sit. Res. 63 71 ve] iL to 40 2 4 13 6 4 H eeoNeNn ohaa Frequencies in 41 71 to to 70 100 1 5 17° 9 7 14 7 8 3 16 10 5 10 6 1 1 5 4 8 8 8 3 il 3 1 1 1 ll 2 14 16 3 OL 4 16 6 10 17 7 101 to 130 7 17 6 eee don BY ectRarw Baan otSit. Res. WR HOW robo - HSe su we HOR ADOM me wenn eGa He REPETITION OF A SITUATION Taste 2 (Continued) Frequencies in Al to 70 wantoon ERD CONT OO EEE 71 to 100 ll 18 me nwo neSoorw HRI OMOM 101 to 130 13 12 7 me ” row Boo Ho Smt enraSar i to 70 1 1 il rose SSGnw 7 to 130 14 18 1 WwW 15 rere co Ko Kien eo me erhar a em So Sit. Res. 92 93 10 101 Frequencies in ll 41 71 to to to 40 70 100 VRONE an a8 Heng Ber Eine wokuranw Wena Howawe oo Bes we # oye 101 to 130 no OO BD ODEO LD 14 © me waBovr eRBa noSantor 13 won Sa hee ome14 THE FUNDAMENTALS OF LEARNING Taste 2 (Continued) Sit. Res. Frequencies in Sit. Res. Frequencies in W141 71 «10l 1 72 W 41 71 10L k 7L to to to to to to to to to to to to 40 70 100 130 70 130 40 70 100 130 70 130 102 101 1 1 2 ll 101 1 1 1022 5 7 1 14 12 15 102 103 15 19 18 14 34 32 103 1 4 2 5 2 iL 8 3 ll 2 il 13 ll 25 26 30 28 51 58 lll 1 1 ut 112 1 1 103 102 2 2 2 2 113 2 2 103 11 15 7 9 26 16 11 19 13 23 19 32 42 four acts he repeated 950 times. He saw no line of 3800 until all had been drawn, nor did he have any experience in drawing any other lines during the experiment. Hach line was drawn quickly as one dash, and no correction was made even when T thought that the result was far too short or too long. T tried throughout to make cach line as a single unitary movement due to a single impulsion connected with the idea of Draw 2 inches or Draw 4 inches or Draw 6 inches or Draw 8 inches, as the case might be. As the experiment progressed the movements seemed to him to be somewhat less spasmodic and less determined by an original jerk of the hand and more determined by the sensation from the pencil on the paper. He was tempted to make the movement more slowly so as to be able to stop it when it seemed to be going too far and to extend it when it seemed to be short; but withstood this temptation com- pletely so far as he is aware. He changed somewhat from a tendency to push his hand with a certain initial force to a tendency to push the pencil to hit a spot 2” (or 4” or 6” or 8” as the case might be) to the right. Tho second type of movement was, to the best of his knowledge, no slower on the whole. Thirty or 60 lines of each length were drawn at a sitting; from one to four sittings were held daily. The technical arrangements were very crude. T drow the lines always with the same sort of pencil sharpened toREPETITION OF A SITUATION 15 about the same bluntness on the same kind of paper (ordi- nary letter size) held on a large pad, with the longer edge parallel to the direction in which the lines were drawn. He drew 3 sets of 4 lines each on,a sheet; then with eyes still closed, put it face down on a pile, opened his eyes, placed another sheet on the pad, shut his eyes and repeated. After he drew the four lines, he pulled the paper up on the pad so that the position of the arm,and hand would be approxi- mately the same for the next four lines. In spite of their crudeness, these arrangements were fairly satisfactory, perhaps more so than a rigorous but artificial control would have been. The variations in conditions may have existed in respect of single drawings, but for one hundred compared with another hundred these disturbances are probably equalized. The important desiderata were to have conditions on the average alike in successive hundreds, and to have the sub- ject deprived of any knowledge of how long the lines which he drew were. The first seems to have been obtained, at least approximately. The second surely was. T had no awareness of any changes either progressive or abrupt save those noted above. He had no idea during or at the close of the experiment that his lines were too short, that they had become longer as the experiment progressed, or that they varied as much as they did. There was a very great increase in the length of the lines in the later fifties over the length for the first fifty in the case of all four lengths, but he was utterly un- aware of this. There was a notable decrease in the length of the last hundred of each length as compared with the previous hundred, but he was utterly unaware of this also. Nor was he any more aware of the more gradual changes elsewhere. The lines were measured to .1 inch (in a straight line from start to finish). 20 means from 2.0 inches up to 2.1 inches, 21 means from 2.1 inches up to 2.2 inches, ete. The responses to the situations, “Draw a 2” line,”16 THE FUNDAMENTALS OF LEARNING “Draw a 4" line,” “Draw a 6" line,” and “Draw an 8” line’? were tabulated by successive fifties and studied. They were then combined into nine hundreds and one fifty. In this form they are presented here, as Table 13. The response which was most frequent in the first hun- dred does not gain at the expense of the less frequent re- sponses. This is true by any reasonable method of choosing the most frequent response. The method which we have used is to take that sequence of lengths (1) which has a total frequency greater than any other sequence and (2) which has a sum of frequencics nearest to 25, and (3) which includes, or is close to, the median length. If Tasre 13 THE RESPONSES OF T IN DRAWING 950 SETS OF LINES, BACH SUT COMPRISING FOUR RESPONSES, ALL BEING WITHOUT KNOWLEDGE OF RESULTS 2 inch lines 4 inch lines Re Re- sponse Frequency sponse Frequency 2888 2222 88 28882228 88 gss2s3s38 38 Sess s388 88 SERB SSSR EB ERE SSE RE 9 2 16061 10°05 “7 11° 10 18 #2 1 12 9 1 to. 1.9 11 1 13 10 8 4 1 2 1 20 6 1 1 14°69 62 2 1 3 1 21 9 3 3 1 15°52 5 61651 121 22 17 10 5 2 1 16 19 1 Ml 1 8 7 8 2 6 23 68 4 1 2 301 1.7 6 17 2 138 1215 12 6 4 9 24 112 6 56 & 1 4 9 9 2 19 12 2% 4 11 8 2.5 6 18 15 10 6 2 8 2 2 1.9 4 9 4 18 10 14 13 6 17 17 26 9 61112 9 38 7 22 20 3 5 8 8 18 13 12 6 16 18 27 4 81415 8 8 18 17 21 3 6 4 8 312 6 9 19 18 28 8 714 9 8 15 12 6 218 22 13 3 9 512 9 7 Ll 10 29 5 7 8 it W 6 8 4 32 23 1 3°75 26 8 3.0 Siw Ww 9 0 8 61 24 1 1121 41 31 2 1 4 8 1 8 6 4 4 13 25 1 1612 41 3.2 43 8 7122 7 4 2 7 2.6 2 330 1 7 4.7 5 6 17 10 2.7 1 1 34 1 1 25 4 9 16 10 3.5 1 1 312 7 7 9 8 3.6 53 3 4 7. 3.7 29 81 7 3 3.8 301 1 2 3.9 3 21 2 4.0 1 4.1 1 * * gxxregezagkh as B@egenreh ae Basses esta ad Saaa aad a oo o 888285888988 SRSSRsRagag * Determination made from 50 responses,17 REPETITION OF A SITUATION Tasiz 13 (Continued) 8 inch lines 6 inch lines Frequency sponse Frequency 096 09 198 mw wwegeywcoouen wae woo 098 oF 124 sone Hoanwaanemmmeda om mg as 094.94 102 AMMO MANOS een om eg OF ‘002 © 109 SARAOH OORT OMEN OD HOCH wo oF 009 © 109 MHA onweR NON wawnaanamamn 16:9 Le (009 0% 10% HH wn wmcongqowmewwnan A we ow or % 108 HArdanwoa nn hongenace ae 08 9% 102 Hannhegwannewewaamnes one 002 09 Tor Anse AAKRanwRBERwoOn mt we Wg, AAMAN CMON RN Qwawn Aeon we os 086 01 108 a eanaggguono wna ao ow 098 192 7 Mewonngqooehewa 4 wr 094% 02 re wine we mt aero 002. 109 atwonwengoowwwoooenae ote 009 0% 109 Hnanatnogengaswawaana eee 009 © 105 Honenwoman gun wwacn awh 68 00F & tog HaARaAnonoenwogqemadn ees WE 19g «= HM ANORIeDQugaana we ow 002 64 Yor OR Weorrenessesn mann we Worn MPC Qoegranwoonawwr ~ re wp * Soo footnote on page 16. the first and third of these criteria conflict, we make a rea- For 2” lines all three agree on 1.5 and 1.6. For 4” lines we use 2.2 and sonable choice, usually favoring the third. 2.3. For 6” lines all three agree on 3.2, 3.3, and 3.4. For 8” lines we use 4.4, 4.5, and 4.6. The status quo is not maintained. On the contrary, there18 THE FUNDAMENTALS OF LEARNING are shifts of mean tendency and changes in variability. The shifts of the mean tendency are in general toward longer lengths and toward the correct lengths. The varia- ble error is in general reduced. The results of Experiments 6 to 22, which are all of the same general pattern as Experiment 5, are reported in Appendix II. They show that the situation represented by a signal or a command makes widely varying connections, with notable shifts of the central tendency and changes in the variation around it, but that the shifts are not by waxing of the initially most frequent connection at the expense of the initially less frequent. They do not give any support to the doctrine that the response which an animal makes oftenest to a maze alley or a puzzle box or a multiple-choice apparatus will thereby gain one iota in probability of future response. § 4. THE NATURE AND MEASUREMENT OF STRENGTH OF CONNECTION So far, we have used the terms situation, response, and connection without dependence upon any rigorous defini- tions of them. What we have found and stated will be equally true and useful by any reasonable definitions. And the terms situation and response may continue as crude but convenient names for some state of affairs without or within the organism and some sequent act or condition of the organism, which is, or seems to be, determined in part by said state of affairs.* The meaning which we attach *Hlement of a situation and feature of a response would be moro exact names for what we mean by situation and response. Any situation, such as we use in our experiments, or teach children to respond to in school, or deal with in any of the mental sciences, is, of course, always a part of a total situation, and is always responded to by a person whoso mind has a set characteristic of that moment and perhaps of no other. Any responso that science deals with is similarly always a part of a total behavior of the animal. We abstract from some total state of nature when we say that a child responds to the situation ‘‘How much is 7444 297? by the rosponse «Rs, Si RBs, .... Si R,), is obviously closely related to the strength of the connection S:—*R:, by any reasonable definition of ‘‘strength of a connection’’; and the facts which we have so far presented concern the strength of connections and the influence of repetition of a situation upon the relative strength of the connections leading from it by any reasonable definition of’ strength, It seems best, however, to adopt a provisional definition of ‘‘strength of a connection’? so that we may, in the experiments next to be described, consider the question of the strengthening of the stronger connections leading from Si at the expense of the weak connections leading from Si, as well as the question of the increase of the more frequent responses to S: at the expense of the less frequent. The most generally accepted and acceptable meaning for the strength of a connection with S:, say of S:— Ri, is the probability that R: will occur if S: occurs. This proba- bility, of course, ranges from 0 to 1. Among the connec- tions with a probability of 1 at any given time and for the ordinary conditions of the person in question and for the ordinary occurrence of S:—that is, among those connec- tions which are sure to occur whenever S: occurs to-day to the person awake and attentive—there are, obviously, vary- ing degrees of strength. Some of these connections would20 THE FUNDAMENTALS OF LEARNING still operate and some would not if S: occurred when the person was very sleepy, or plus distracting accompani- ments, or after an interval of a year with no practice. We can differentiate’ among these connections with a proba- bility of 1 at the present under ordinary conditions, by measuring their probabilities (a) at some future time, or (b) after disuse, or (c) with stronger competition. It is not certain that probability of occurrence now against ordinary competition is the same as (that is, is perfectly correlated with) probability of occurrence now when the subject is asleep, or of occurrence after one hun- dred days of disuse, or of occurrence in excitement. In- deed, it seems almost certain that some of these correlations are not perfect. In general, caution is desirable in arguing beyond the time being, or to extraordinary conditions of the organism. We can differentiate among these by the time required for relearning to some defined degree of strength. Con- nections which do not act to produce responses of any appreciable frequency, that is, which show 0 occurrences, may be of widely different strength. The reader may, supposing him to have studied Latin in the past, not re- spond by the correct English equivalents of any of these Latin words: dedi, stare, utor, exegi, frumentum. The fro- quency may be 0 for every one of them; but the connections involved may be relearned in vastly different amounts of training. In the present series of experiments and arguments we shall not need to argue concerning anything beyond the time span of a few days or different from ordinary mental conditions. Nor shall we have to differentiate among prob- abilities of 1.00 or of only infinitesimal amount. For all the probabilities with which we shall deal will be less than 1 and clearly above zero. We are concerned with cases where SiR, Sr Re, Si, Ra, ete., all have proba- bilities with values above the infinitesimal and below 1. Our difficulties are factual and statistical within this sim-REPETITION OF A SITUATION 21 ple scheme. The factual difficulty is that we cannot deter- mine the initial strength of the connections leading from S: at all exactly without repeating S: many times. What, for example, shall we set as the initial strength of the con- nections with the situation Draw an angle of 45° under the general conditions of Experiment 22, leading to the re- sponses of producing an angle of 0° to .99°, 1° to 1.99°, 2° to 2.99°, and so on, in Br?* If we use the first 333 occur- rences of Si, responses producing angles of 40° or over have 0 strength, and responses of 18° to 23.99° have a strength of = But a few days later, responses producing 237, angles of 40° or over have a strength of >,~ while re- 280 sponses from 18° to 23.99° have 0 strength. In the case of many of Experiments 1 to 22, we cannot argue in a straightforward way from the observed fre- quency of occurrences to the strength of connections, be- cause of the instability of the connections. To obtain such an inventory of T’s tendencies in drawing 3” lines or 4” lines that the future would, except for changes wrought by repetition, repeat the past, might well require ten thousand occurrences of Si! If our initial sampling represents a random picking from a stable status quo, we still havea statistical difficulty. Suppose, for example, that the true status is as follows: SiR, P= .0300 Re .0700 Rs -1000 Ry -1400 Rs -1600 Re -1600 Ry -1400 Re -1000 Ry .0700 Ro .0300 *Seo Appendix II for the details of this experiment.22 THE FUNDAMENTALS OF LEARNING If we draw at random 100 occurrences of S: we shall not have exactly 3, 7, 10, 14, 16, 16, 14, 10, 7, and 3 occurrences of Ri, Re, Rs, ete. The frequencies of Rs and Re will not always be equal, nor will cither always be greater than the frequency of Rs. The connection observed to be of great- est frequency may not be the connection of greatest real strength. In some of the experiments soon to be described, it is desirable to work with total initial frequencies of only 8 or 10in 100. The effect of the chance error of sampling may then be large in comparison with small differences in real strength, such as the differences between .300, .200, -100, and .050. There are means of allowing for its influ- ence which we shall be careful to take. Our definition of strength seems at first sight to accord ill with one very common case of strength, that where a person makes a certain response, say 70 times out of 100 and ‘‘does nothing’’ 30 times out of 100. In such a case there seem to be no connections with a strength of .30 to fill out the probability. And in a sense there are not. Any one of so many thousands of things may occur, no one of which has any appreciable connection with Si, that we may reasonably think of S: in these 30 cases of a hundred as leading on to nothing, the future mental action being determined by other factors than S:. In a stricter sense, however, there are connections. The person really does something in response to S: cach time of the hundred that it occurs, but what he does in the 30 cases is negative in its practical consequences (as when he thinks ‘I don’t know that’’), or is something which might fit any other situation as well as it fits S: (as when he turns back to the course of life which S: had inter- rupted), or is both negative and not specific to Si. When the 30 percent of connections leading from S: are such that we do not care what they are beyond the fact that they are not R: or Re or Rs, ete., it is common to think of them as blanks, failures of response. Our definition and measurement of strength will fit such a case convenientlyREPETITION OF A SITUATION 23 enough, if we think of the .70 as the strength of Si:—>Ri, and of the .30 as the strength of Si— not Ri, and of ‘not Rx’’ as made up of a large number of responses each of infinitesimal strength. The response ‘I don’t know’’ or “T can’t’’ is, by this view, taken to mean that if I were forced to give an answer, I would be as likely to give any one of ten thousand or more answers as any other of them. The response of inattention or turning away, is, by this view, interpreted as a turning to any one of thousands of other things, no one of which has more than an infinitesimal probability of being turned to from S:. We are now in a position to distinguish clearly between the alleged potency of great use to draw strength away from the little used connections, for which we have found little or no evidence and against which we shall find evi- dence later, and the genuine, authenticated potency of use. Repetition undoubtedly does strengthen connections. In- deed, common observation teaches us that repetition strengthens even those connections whose effects are not satisfying, as when a given name or place or face evokes ideas which irritate and pain us. Case I. Consider the following typical case. A situa- tion, 8:, for example, ‘‘What does the word levirate mean?” has a connection with the correct meaning (call this Ri), of infinitesimal strength and thousands of con- nections with ‘‘not R:,’’ each of infinitestimal strength. ‘We say ‘‘infinitesimal’’ rather than ‘‘zero”’ because any situation to which an animal is sensitive does probably have some infinitesimal tendency to evoke almost every response of which the animal is capable. We shall use 0+ as the symbol for the amount of strength of such a connection. The connections from §: are then, at the beginning: SR; strength of 0+ S,:>R, strength of 0+ SR; strength of 0+ SR. strength of 0+24 THE FUNDAMENTALS OF LEARNING The subject reads, or is told, ‘‘levirate means the custom of requiring a man to marry his brother’s widow.’’ This experience (S:—R:) changes the status to: S:>R. strength of 1.0 S:>R, strength of O+ SOR. strength of 0+ SiR: has been given enormous strength relative to Si: Ry, ete., for the time being. Repetitions of the ex- perience, or recalls of the SiR: from within, add to its strength for the future or for action against stronger com- petition.* We may think of this Case I as the case of addition of strength to a connection of 0 + strength, against the 0+ competition of many other connections, also of 0+ strength, Case II. The addition of strength to 0+ against ap- preciable competition. Suppose that the original status is: S,>R,, with a strength of 1.0 due to 10 repetitions, one a day for 10 days. S>R,, with a strength of 0+ SOR, with a strength of 0+ SR. with a strength of 0+ Suppose that the person experiences Sr Re (with an acceptability equal to that of the SiR: experiences) once a day for 20 days. Just what will then happen when S: occurs the thirty-first time, n days later, cannot be pre- dicted surely. It will depend on the exact natures of Ri and R:z and the total sets in which they occurred. But there will remain, in the long run, other things being equal, an * It should be noted that in all, or nearly all, casos oxporiences which #0 change status by increasing the strength of a conncetion are endowed with a certain acceptability. They are not rejected or opposed by tho animal, but are allowed to live and play their parts, What happens when such an S,—>R, is accompanied by repulsion or pain, is a matter for scparate consideration.REPETITION OF A SITUATION 25 appreciable strength of SiR: and an appreciable and greater strength (due to greater frequency and a shorter period of disuse) of Si—>Re. The status will be approxi- mately: S:>R, with a strength of 3+x S:>R,, with a strength of .7+y S:>Rz, with a strength of 0+ S:>R., with a strength of 0+ The subject will be much more likely to respond to Si by R, or Re than by Rs or Ry or Rs, ete. He will relearn either of them more readily than he will learn the connections from S: to Rs or Rs or Rs, ete. He will be more likely to respond by Rez than by R:,* and will relearn it more easily. Case IIT, The addition of strength to one of several com- peting connections, each of appreciable strength. Let the status be: S:>R,, with a strength of .1 S,>R,, with a strength of .2 S:>Rs, with a strength of .7 S:>R,, with a strength of 0+ Si*Rz, with a strength of 0+ S,>Re, with a strength of 0-+ Suppose that the person is led to experience Si—Rz 20 times. Just what the result will be will depend upon sev- eral things, including especially the number of the past experiences which have produced the status just described; but, in the long run, other things being equal, SiR» will gain in strength relative to S:—*R:, SiR, 8S: Ru, ete. There are other instructive cases which might be dis- cussed, but these three are sufficient to introduce and il- “If the influence of primacy were very strong, this might not be true, but we shall show elsewhere that the influence of primacy is zero or very weak.