Digitized by the Internet Archive in 2019 with funding from Princeton Theological Seminary Library / https://archive.org/details/influenceofmechaOOkoch VOL, XXXII I7 JAN IS 1924 NO. 5 PSYCHOLOGICAL REVIEW PUBLICATIONS WHOLE NO. I47 1923 Psychological Monographs EDITED BY JAMES ROWLAND ANGELL, Yale University HOWARD C. WARREN, Princeton University ( Review ) JOHN B. WATSON, New York (/. of Exp. Psychol.) SHEPHERD I. FRANZ, Govt. Hosp. for Insane ( Bulletin ) and MADISON BENTLEY, University of Illinois (Index) STUDIES FROM THE PSYCHOLOGICAL LABORATORY OF THE UNIVERSITY OF CHICAGO The Influ ence ofMechanical Guidance Maze Learning BV HELEN LOIS KOCH, Ph.D. Instructor in Educational Psychology in the University of Texas PSYCHOLOGICAL REVIEW COMPANY PRINCETON, N.vJ. Agents: G. E. STECHERT & CO., London (2 Star Yard, Carey St., W. C.) Paris (16 rue de Conde) ACKNOWLEDGMENTS The writer wishes to express her deep gratitude to Professor Harvey A. Carr for his untiring and critical supervision of this research, as well as for the innumerable other kindnesses shown her. To President James Rowland Angell, Professor C. Judson Herrick and Professor J. R. Kantor she is indebted for much inspiration and guidance. CONTENTS PAGE I. Introduction . i II. The Relative Efficacy of Control Introduced at Various Positions in the Learning Process . n A. Results Based upon the Records of Animal Subjects 13 1. Influence of Two Directed Trials Introduced at Various Positions in the Learning Process. 13 2. Influence of Four Directed Trials Introduced at Various Positions in the Learning Process. 22 3. Influence of Six Directed Trials Introduced at Various Positions in the Learning Process. 28 4. Influence of Eight Directed Trials Introduced at Various Positions in the Learning Process. 34 B. Results Based upon the Records of Human Subjects 38 1. Influence of Two Directed Trials Introduced at Various Positions in the Learning Process. 38 2. Influence of Four Directed Trials Introduced at Various Positions in the Learning Process. 46 3. Influence of Six Directed Trials Introduced at Various Positions in the Learning Process. 52 4. Influence of Eight Directed Trials Introduced at Various Positions in the Learning Process. 57 III. The Relative Efficacy of Various Amounts of Control . 67 A. Results Based upon the Records of Animal Subjects 67 1. Influence of Various Amounts of Initial Guid¬ ance . 67 2. Influence of Various Amounts of Guidance Introduced upon the Fifth Trial . 72 B. Results Based upon the Records of Human Subjects 75 1. Influence of Various Amounts of Initial Guid¬ ance . 75 v VI CONTENTS PAGE 2. Influence of Various Amounts of Guidance Introduced upon the Fifth Trial . 80 3. Influence of Various Amounts of Guidance Introduced upon the Seventh Trial . 83 4. Influence of Various Amounts of Guidance Introduced upon the Ninth Trial . 86 IV. The Influence of Control upon Retention . 93 V. The Influence of Guided Learning upon the Adapt¬ ability of the Learned Reaction . 98 VI. General Summary . 106 I INTRODUCTION The general purport of this study is an investigation of the in¬ fluence upon maze learning of the mechanical prevention of error. The research was carried on in the years 1919-1920 at the Psychological Laboratory of the University of Chicago and was a direct outgrowth of an investigation conducted jointly by Dr. H. A. Carr and the author,1 in which an attempt was made to train some albino rats to do a task by guiding them for a time through the critical phase of the situation. The guidance was mechanical in its nature, and, while conducting the animal through the proper response, it still permitted him to initiate all of his activities. The problem to be mastered was the alternate choice of the right and left paths of a T-shaped problem box in each day’s test of ten successive runs. The results revealed that the group of animals which was guided forty out of every fifty trials did not react to the situation so effectively as did the unguided group. The control was not, however, without marked positive effect upon some of the animals. The maze problem was selected as the medium for the present investigation for the following reasons : ( 1 ) In so far as the maze differs from the problem box just described, its employment will cast some light upon the question of the relation between the na¬ ture of the problem and the effectiveness of control. (2) The maze problem is an eliminative type of problem; that is, its mastery in¬ volves primarily the elimination of error, rather than the acquisi¬ tion of any new movement. It is reasonable to expect that control which consists of the prevention of errors will have a greater in¬ fluence upon the learning of such a problem than upon the learn¬ ing of those in whose mastery sheer elimination of errors plays a less dominant role. (3) The character of the problem is such, furthermore, that it permits a more detailed analysis of the learn¬ ing process and of the nature of the effect of the control than 1 “The Influence of Extraneous Controls in the Learning Process,” Psych . Rev., vol. 26 (1919), p. 287 ff. 1 2 HELEN LOIS KOCH does the problem box hitherto employed. (4) The maze presents a problem which is well adapted to the capacities of both human subjects and rats, and thus makes possible a comparative study of the influence of control upon these widely divergent animal groups. Our general method of procedure consisted in preventing cul- de-sac errors in certain periods of the learning, by blocking the entrances to the blind alleys. The guiding device was mechanical and, hence, subject to little variation. The subject, moreover, was permitted to initiate all of his movements. In this respect our method of control diverges from that method frequently employed, in which the subject, who is supposedly passive, is guided by the experimenter through the movements of the act to be learned. The influence of this form of control has been investigated by Thorn¬ dike,2 Cole,3 Yerkes,4 Hunter5 and Ludgate.6 The problems upon which the present study will attempt to cast light are the following: (1) Does the mechanical prevention of errors during part of the learning process have any influence upon learning? (2) Does the efficacy of the control vary with the period of the learning at which it is administered? A comparison, for instance, of the relative effectiveness of guidance given in the initial four trials, as opposed to control given in the trials from the ninth to the twelfth inclusive, will furnish significant data in regard to this problem. (3) Is the efficacy of guidance a function of the amount given? Our general method of investigating this problem is to consider, for example, the relative effectiveness of two, four or six, etc., directed trials, interpolated in the same general position in the learning process. (4) Does guidance have a similar influence upon human and animal subjects? (5) Does the fact of guidance in the learning period have any influence upon 2 “Animal Intelligence,” Psych. Rev. Mon. Suppl., vol. 2 (1898). 3 “Concerning the Intelligence of Raccoons,” Jour, of Comp. Neur. and Psych., vol. 1 7 (1907), p. 21 1 ff. 4 “The Dancing Mouse,” (1907), p. 201 ff. 5 “A Note on the Behaviour of the White Rat,” Jour, of Animal Behav., vol. 2 (1912), p. 137 ff. 6 “The Effect of Manual Guidance upon Maze Learning,” Psych. Rev. Mon. Suppl., vol. 33 (1923). THE INFLUENCE OF MECHANICAL GUIDANCE 3 the retention of the habit? To state the problem more simply: Does a subject who has been guided during the learning of a problem retain the habit as well as the subject who has not been so guided? (6) Does the fact of guidance in the learning period affect the stability of the habit? In other words, is an individual who has mastered the problem with the aid of guidance as likely to be confused when circumstances are slightly altered as is the individual who has learned the problem without extraneous con¬ trol? The maze used in the experimentation upon the animals was constructed of oak boards, YY' in thickness. It was supported by a wooden frame 18" high and was 4/x3'8"x6// in size. The pattern of the maze is indicated in Fig. 1. The runways and the cul-de-sacs were 4" wide. The partitions within the maze, which were made of thin sheets of galvanized iron, were held in place by brass sup¬ ports. The wood and metal parts of the apparatus were painted dull black. A glass cover permitted observation of all of the maneuverings of the animal while he was in the runways. The doors that opened into and out of the uncovered food-box were sliding doors which, with a little care, could be closed without distracting the animal to any considerable degree. The blind alleys were closed off, when desired, by pieces of 4 HELEN LOIS KOCH window glass. Attached to the upper edge of each plate of glass was a blackened brass clamp which, when sunk into the sockets in the walls of the alleys, held the glass partition securely in place. The glass controls were not set in flush with the walls of the main pathway, but rather at a distance of three centimeters from the ends of the cul-de-sacs. This arrangement was adopted in order to avoid, as far as possible, the distraction from the alteration of tactual cues which one would expect to attend the introduction and removal of the controls. Glass, moreover, rather than metal partitions, were employed, in order to reduce to a minimum the visual distraction, which the introduction and abstraction of the control might produce. After some preliminary experimentation, the results of which indicated that directed trials occurring subsequent to the twelfth have a marked deleterious effect, it was decided to concentrate within the first twelve trials whatever guidance was given the animals. Series of two, four, six or eight guided trials were, with one exception, inserted in various positions within the first twelve trials involved in the learning of the maze. A series of two suc¬ cessive guided trials was introduced, in the case of one group of animals, upon the first run; in the case of another, upon the sixth run; and for a third group, upon the eleventh run. Four groups of animals were granted a period of control, extending over four successive trials. The period of guidance began with the first, fifth, ninth or thirteenth trial, respectively. Series of six directed trials were employed, in the case of three groups of animals, the introduction of the directed series occurring upon the first, fourth or seventh trial, respectively. Successions of eight guided trials, commencing upon the first, third or fifth run of the learning pe¬ riod, respectively, were likewise employed. One group of animals was given twelve successive directed trials. The normal group, of course, learned the maze without assistance. It is evident from the schema just described that each group of animals received either two, four, six or eight guided runs at either the beginning, middle or end of the first twelve trials. Such a grouping enables one to compare the relative effect of various amounts of guidance, as THE INFLUENCE OF MECHANICAL GUIDANCE 5 well as the effect of the position of the guided trials upon learn¬ ing. Sixteen groups of ten animals each were employed in the course of the experiment, each group consisting, because of the scarcity of females, of six males and four females. The rats were from seven to twelve weeks old. Any marked biasing of the results, through differences in the strain of the animals used, was avoided by drawing each group from no less than three different litters. The rats were carefully tamed and fed in the food-box of the maze for one week before training was begun. Whatever physical mutilation of the animal was necessary for purposes of identifica¬ tion was effected at least three days before training commenced. The position of the maze was never changed during the learning period of any group of animals. Lighting conditions were con¬ trolled as far as possible. The shades were drawn and the electric light switched on before work was begun. The experiment was conducted in the late morning and early afternoon hours. The cleaning of the cages and replenishing of the water supply, etc., was cared for after the day’s trials. Thus, for the accomplishment of any necessary adjustments to the slight alterations in the en¬ vironment, a period of twenty-four hours was allowed. The same general position of the living cage of any group of animals was maintained throughout the entire experiment. The animals were kept normally hungry. They were allowed to feed seven minutes per day on bread and milk, were fed on an average six sunflower seeds per individual per day, and once a week were favored with a small piece of lettuce or apple. One bite of food before each run served to stimulate the proper incentive. The animals of each group were fed together at the end of the daily experimentation. During the first four days, each rat was granted only one trial per day. Thereafter, however, it was given two trials per day until it succeeded in making four perfect runs out of five suc¬ cessive attempts. This criterion of mastery has previously been successfully employed.7 No record of the distance traversed was 7 Cf. Webb : “Transfer of Training and Retroaction,” Psych. Rev. Mon. Suppl., vol. 24 (1917). 6 HELEN LOIS KOCH kept, as Miss Hicks8 suggests; but the method of counting errors obviated to a large degree many of the difficulties she discusses. Any return over a whole or a part of a unit of the true pathway was counted as a return error. A unit of the true pathway, to be explicit, consists of a section of the runway between two suc¬ cessive turns of the true path, irrespective of the length of the section. Every entrance into a cul-de-sac, as well as every re¬ tracing toward the end of the cul-de-sac, after the animal had headed toward the entrance, was considered a cul-de-sac error. Time, measured by a stop watch, was recorded from the moment the rat left the entrance until its whole body, with the exception of its tail, was brought within the food-box. The maze employed in the experiment upon human subjects was a small stylus maze of the same general design as the maze used for the rats. The cul-de-sacs and the true pathway, which were milled out of a solid aluminum casting, were y 4" wide and deep. The partitions between these wedges were, likewise, J4" thick. The outside dimensions of the maze were 5^4" x 5%". Small brass blocks, from the bottom of which projected pegs that could be inserted into small holes drilled into the floor of the maze, were used to block off the blind alleys in the directed trials. The blocks were inserted at the same relative distance from the end of the cul-de-sacs as were the glass controls in the maze used in the experimentation upon animals. The absolute distance was 3 / // / 16 • The maze was concealed under a heavy black curtain that cov¬ ered the top and three sides of a supporting wooden frame. The frame was ft. wide, 1 ft. deep, and 1 ft. high on the side toward the subject, whereas it was ij4 ft. high on the side ex¬ posed to the experimenter. From that side of the frame which was uncovered, the experimenter could observe the progress of the subject. In order to prevent any distracting movement, the maze was held tightly in place by a small wooden frame which was nailed to the table. The black cloth was hung loosely enough on the side toward the subject so that his arms could be easily in- 8 “The Relative Value of the Different Curves of Learning,” Jour, of An¬ imal Bchav., vol. I, (1911), p. 138 ff. THE INFLUENCE OF MECHANICAL GUIDANCE 7 serted under it. Such an apparatus eliminates the disconcerting ef¬ fect of blindfolding and reduces the problem largely to a tactual kinaesthetic-motor level. The subjects were required to trace the maze with a hard rubber stylus, the lower end of which was 3/i6" in diameter and could be guided easily through the runways. A small rubber shield 5Ae" from the end of the stylus prevented the fingers of the subject from coming in contact with the maze. For both human and animal subjects the same criterion of mastery was employed. The subject was seated at a table and the following written directions were given to him : “Please put your right hand under the cover. Grasp the stylus and hold it as erect as possible. Be sure that neither your fingers nor your hand touches the base of the apparatus. Keep the stylus in the groove and explore the assigned area until you are told to stop. Use any method you desire. The aim of the experiment is to find the shortest possible route through the maze.” When the subject placed his hand under the cover, the experimenter put the stylus in it and placed the stylus in the groove at the entrance to the maze. Then the experimenter said to the subject, “Now you are in the groove. Explore the maze and I shall tell you when you reach the goal.” When the subject reached the goal for the first time, the experimenter said to him, “Now you have reached the goal. The aim of the experiment is to learn to reach this goal without taking any unnecessary steps.” The object of giving instructions to the subject thus piecemeal is to avoid confusing him with a large number of directions, the meaning of which can only become apparent when he has had some experience with the situation. The repetition in various forms of the aim of the experiment is necessary in order that each subject may thor¬ oughly comprehend what is desired of him. Such a procedure as Webb’s,9 in which the subject was forced to deduce the aim of the experiment, is wholly undesirable. Webb asserts that because the human subject is blindfolded and is obliged to fathom the desires of the experimenter, his incentive is similar to that of the rat who is placed in the maze and allowed to seek his own deliverance. Upon a close scrutiny of the situation and the behavior of the sub- 9 Op. cit., p. 14. 8 HELEN LOIS KOCH jects, the similarity is not so evident. The human subject does, from the first, set for himself a problem. The problem, if the sub¬ ject is not instructed, differs with each individual. Some subjects believe that the problem is to traverse every inch of the maze, rather than to find the shortest route. The rat, we assume, moves through the maze at first through sheer curiosity and inability to be quiet unless fatigued, ill, asleep or in danger. After the first few trials, food and the maze situation become associated. Each rat, after the initial trials, has the same incentive. But the human subject has no such incentive and no criterion of success unless he is told what he is to do. While Webb’s procedure does a priori keep the conditions for human and animal subjects strikingly similar, nevertheless, the action of the subjects indicates that the human with a knowledge of the problem has an incentive more nearly comparable to that of the rat than the subject without such information. A comparison of the records for a group of ten sub¬ jects who labored through the maze without a knowledge of what was desired of them, with an instructed group reveals that this information is effective in producing a drop in the trial score from 70.6 to 44.3; in the total error score from 476.1 to 325.1 ; and in the total time score from 2250.82" to 1465.59". Such re¬ sults refute Webb’s10 assertion that knowledge of the aim in the maze problem has little effect on learning. Because of the great difficulty of inducing a large number of rather disinterested subjects to serve regularly for any length of time, the same distribution of trials was not employed as in the case of the investigation with animals. The experiment was so ar¬ ranged that the problem could usually be mastered at a single sit¬ ting. No subject was detained, however, for more than 1*4 hours at a time. If the problem was not mastered within this period, the subject was requested to return every 48 hours until a mastery was effected. If a subject became fatigued, he was dismissed after one hour. Between the first and second trials, as well as between the second and third, a rest of one minute was granted. If the first trial required more than fifteen minutes, the subject was granted all the time he needed to recover from fatigue. After the second 10 Op. cit., p. 15. THE INFLUENCE OF MECHANICAL GUIDANCE 9 trial a rest of one minute was given between groups of two trials. This distribution of effort in a large measure obviated fatigue, as well as the confusion which results from too steady application. The subject was at no time informed that he was being guided. The insertion and extraction of the blocks were accomplished noiselessly. The scheme for the introduction of the controlled trials was slightly varied from the one employed in the experimentation up¬ on the animals, in order to bring out more clearly the effect upon learning of various amounts of guidance. Six groups of subjects were controlled for a period extending over two successive trials. The control, in the case of these groups, was introduced upon the first, third, fifth, seventh, ninth and eleventh trial, respectively. Three groups of subjects were each granted a period of guidance, four trials in length. The series of four controlled runs began, respectively, upon the first, fifth and ninth trial. Two groups were guided through a series of six successive trials, one of the groups being guided through the first six trials of the learning period; the other, from the seventh to twelfth trial, inclusive. Two groups of subjects, likewise, were given eight directed trials; one group being guided from the first to the eighth trial, inclusive ; the other, from the ninth to the sixteenth. One group was guided through the initial twelve trials. The normal group learned the maze without assistance. Each of the fifteen groups consisted of ten individuals — eight women and two men. All were students or instructors at the Uni¬ versity of Chicago. Most of the subjects were naive, so far as the maze situation was concerned. The few students who were famil¬ iar with the maze problem were equally distributed through all of the groups, in order to prevent a weighting of the results by this factor. The retention and distraction tests were carried on 48 hours after the mastery of the problem had been effected. Retention was tested by a single tracing of the maze. Each subject traced the maze under nine different distracting conditions. The distractions employed were a shifting of the posi¬ tion of the maze through 90°, 180° and 270°, respectively; the 10 HELEN LOIS KOCH silent recitation of the first stanza of “Mary had a little lamb”; reading aloud; drawing triangles with the left hand while the right traversed the maze, and vice versa; tracing the maze with the left hand; and traversing it from the so-called goal back to the en¬ trance. When the maze was shifted 90° from its original position, as well as 180° and 270°, the subject was not informed of the nature of the change, nor was he told that he was being distracted in any way. This procedure of keeping the subject in ignorance of the altered environment until he discovered it for himself, was adopted in order that the conditions might more closely resemble those of general life in which changes in a situation are seldom discreetly labelled, but are rather left to the individual to decipher. II THE RELATIVE EFFICACY OF CONTROL INTRODUCED AT VARIOUS POSITIONS IN THE LEARNING PROCESS It is of great practical value to know what is the most oppor¬ tune time in the learning process for guiding or instructing an in¬ dividual and to be cognizant of the effects of guidance at the various stages in the mastery of a problem. Toward the solution of this question the present study has taken only one small step. Our problem is a description and analysis of the relative efficacy of mechanical guidance introduced at various stages in the learn¬ ing of a maze. For purposes of analysis it is desirable to consider the influ¬ ence of control from three points of view : namely, from the point of view of the result immediately manifested, the result subse¬ quently exhibited, and the total result. The immediate influence of the control is that which is evident during the period in which guidance is being administered. The subsequent influence is that manifested in the undirected trials following the controlled series. The total effect is a composite of the two influences just described and will be represented by such measures as total time, total errors and total trials. Before a discussion of the results is attempted, a comment on the method of deriving the scores is necessary. In order to gain the increased validity resulting from an increase in the number of cases, all available measures which bore on any point in question were employed. There were ten groups of rats, for example, which were unguided in the first two trials. Hence, since there were ten animals in each group, the normal score for these two runs is an average of one hundred measures. The scores for the eleventh and twelfth trials in the normal series, on the other hand, are based upon only twenty measures : i.e., upon the records of the animals in the group which received no guidance and those ii 12 HELEN LOIS KOCH in the group guided from the thirteenth to the seventeenth trial. The other eight groups whose records were used in deriving the scores for the first and second trials had received some guidance before the eleventh run. Since a normal score for a given trial is an average of the scores for an unguided run, the total series of trials preceding which are also unguided, the records of these eight groups could not be employed in computing the scores for the eleventh and twelfth trials in the normal series. The same principles of selection and elimination were utilized in deriving the scores for the directed trials. There were five groups which were guided during the first and second trials. Hence, the scores for the first two- directed runs are based upon fifty measures. The scores for the third and fourth trials, when these runs are themselves controlled and preceded by directed trials only, are computed on the basis of forty measures. The record of that group receiving guidance in only two of the initial trials could not be employed in computing the scores for these trials, as it had been in deriving the scores of the first two directed runs. Hence, the reason for the reduction from fifty to forty measures is evident. The number of cases used in deriving any average is indicated in Table i. THE INFLUENCE OF MECHANICAL GUIDANCE 13 TABLE 1. NUMBER OF CASES USED IN COMPUTING A GIVEN SCORE— ANIMAL SUBJECTS1 Serial Num¬ ber of the Trials for which an Average Score is Sought Nature of the Trials Number of Uncontrolled Trials Preceding Trials for which an Average Score is Sought Number of Controlled Trials Preceding Trials for which an Average Score is Sought Number of Cases upon which Average is Based Controlled Uncontrolled 1-2 X 100 3-4 X 2 80 i-4 X 80 5-6 X 4 40 1-6 X 40 7-8 X 6 30 1-8 X 30 9-10 X 8 20 I-IO X 20 11-12 X 10 20 1-12 X 20 1-2 X 50 3-4 X 2 40 i-4 X 40 5-6 X 4 30 1-6 X 30 7-8 X 6 20 1-8 X 20 5-6 X 4 20 7-8 X 4 2 20 5-8 X 4 20 5-8 X 4 30 5-12 X 4 20 3-io X 2 30 4-9 X 3 30 7-12 X 6 20 9-12 X 8 20 1 All scores not indicated in the table are based upon 10 cases. A. Results Based upon the Records of Animal Subjects 1. The Influence of Two Directed Trials Introduced at Various Positions in the Learning Process Let it be recalled that there were three groups of animals which were guided for only two trials. One of these groups was con¬ trolled in the first and second trials; another, in the sixth and seventh trials; the third group, in the eleventh and twelfth trials. For convenience, the groups will be designated by the numbers 14 HELEN LOIS KOCH 2 (1-2), 2 (6-7), 2 (11-12), respectively. In the tables and throughout the discussion the first digit in the configuration will be used to indicate the amount of guidance, while the figures in brackets will indicate the trials in which guidance was given. TABLE 2. INFLUENCE OF TWO DIRECTED TRIALS UPON TRIALS— ANIMAL SUBJECTS Group Trials Absolute Saving Relative Saving (Per Cent) 2 (1-2) 25-5 8-3 24.67 2 (6-7) 28.8 5-0 14.70 2 (n-12) 3i-9 1.9 5-6o Influence upon trials : It is apparent from Table 2 that the con¬ trol in the three positions acts to reduce the number of trials re¬ quired for learning the maze; and its efficiency is greatest when it is introduced in the initial trials. This beneficial effect decreases as the series of guided runs is removed farther from the begin¬ ning. It should be noted that the probable errors are not given in the tables. Because of the ambiguity of the meaning of this measure when based upon so few cases, its omission was deemed advisable. The reliability of the results may be judged by the consistency of the general tendencies exhibited. The symbol of negativity employed in some of the subsequent tables also requires explanation. It is used when the score of the guided group is greater than that of the normal group. A nega¬ tive saving, then, of an absolute or relative sort, means that the score of the controlled group exceeds absolutely or relatively that of the uncontrolled group by the amount indicated. THE INFLUENCE OF MECHANICAL GUIDANCE 15 TABLE 3. INFLUENCE OF TWO DIRECTED TRIALS UPON TOTAL ERRORS— ANIMAL SUBJECTS Group Total Errors Absolute Saving Relative Saving (Per Cent) 2 (1-2) 143.22 16.93 10.57 2 (6-7) 120.16 39-99 24.97 2 (11-12) 14570 14-45 9.02 Influence upon errors (Table 3) : Guidance reduces the total number of errors made in the entire learning process. Group 2 (6-7) is the most benefited; groups 2 (1-2) and 2 (11-12) ex¬ hibit about the same amount of saving. By a detailed examination of the error scores we may ascertain some of the factors and mechanisms which combined to produce the totals just described. The immediate effect of control in its most obvious form is the prevention of cul-de-sac errors. The number of cul-de-sac errors prevented during the control period is a function of the length of the directed series, as well as its position. In the later trials when, through the very nature of the learning process, fewer errors are being made than early in the learning, the influence of the guiding device as a means of pre¬ venting errors must necessarily decrease. For the same reason, as the series in any given position increases in length, the saving in cul-de-sac errors per trial decreases. But in this sheer physical prevention of cul-de-sac errors we have little interest. The truly significant immediate effect of the control will be revealed in the return error scores. i6 HELEN LOIS KOCH TABLE 4. IMMEDIATE INFLUENCE OF TWO DIRECTED TRIALS UPON ERRORS— ANIMAL SUBJECTS A. Group Av. No. of Cul-de- sac Errors per Directed Trial Absolute Saving Relative Saving (Per Cent) 2 (1-2) 0 6.06 2 (6-7) 0 3-89 2 (11-12) 0 3-05 B. Group Av. No. of Return Errors per Directed Trial Absolute Saving Relative Saving (Per Cent) 2 (1-2) 22.11 — 1. 00 —474 2 (6-7) •55 2.41 81.42 2 (11-12) i-95 —.92 —89.32 Table 4 indicates that the control has little, perhaps even a slightly deleterious, immediate effect, upon the return errors when it is introduced in the initial position. This unfavorable influence on return errors in the initial trials is, doubtless, a result of the fact that in the early stages of the learning, entrances into cul-de- sacs do operate to prevent complete returns to the beginning of the maze. When the cul-de-sacs are blocked, the animal which has once headed back toward the starting point has little to prevent a complete retracing of the pathway. Hence, when returns are the only possible errors, more of them are made than when cul- de-sacs may also waylay the victim. Entrances into cul-de-sacs in the intermediate trials, on the other hand, instead of operating to prevent complete retracings, as they do in the initial trials, tend to increase the opportunity for retracing, for the return path in the first few sections of the maze, at least, has by the sixth trial, as a rule, been well mastered. Hence, when cul-de-sacs are blocked, the animal that is still in a some¬ what exploratory stage of the learning and not greatly dependent on fixed cues, is guided easily forward to the goal. The difficulties and confusion caused by the cul-de-sacs is lacking. Confusion re¬ sulting from the novelty of the situation has worn off. This com- THE INFLUENCE OF MECHANICAL GUIDANCE 17 plex of conditions is probably the explanation of the large abso¬ lute, as well as relative, saving in return errors in the sixth and seventh trials. In the case of that group of animals for whom the guided series was most distantly removed from the beginning, however, a marked increase in return errors is exhibited during the period of control. The animal, presumably, by the eleventh trial has be¬ come greatly dependent on certain cues. Blocking the entrances to the cul-de-sacs modifies some of the cues, causes confusion, and, since return errors are the only ones possible, more of them are made than normally. In our tabular analysis of the subsequent effect of control we shall consider only the average number of cul-de-sacs and the average number of return errors amassed per trial in the post¬ control period. For a more minute analysis of the subsequent in¬ fluence of guidance, i.e., the determination of the period of great¬ est influence, the persistence of the effect, the conflict of the dis¬ tracting and beneficial forces, we may rely for our data on the learning curves. TABLE 5. SUBSEQUENT EFFECT OF TWO DIRECTED TRIALS UPON ERRORS— ANIMAL SUBJECTS A. Group Av. No. of Cul-de-sac Errors per Trial for Trials Subsequent to Control Absolute Saving Relative Saving (Per Cent) 2 (1-2) 2.12 —.14 —7.07 2 (6-7) •93 .84 4746 2 (11-12) .92 •37 28.68 B. 1 i| 'I - Group Av. No. of Return Errors per Trial for Trials Subsequent to Control Absolute Saving Relative Saving (Per Cent) 2 (1-2) 2.00 —.88 —85-98 2 (6-7) .89 — .08 — 9-87 2 (n-12) •65 —.14 —2745 i8 HELEN LOIS KOCH Group 2 (1-2) manifests an increase above the normal in the average number of cul-de-sac errors amassed per trial for the period subsequent to the guided interval; groups 2 (6-7) and 2 (11-12) exhibit, on the other hand, a positive saving (see Table 5). Each of the groups, however, was unfavorably influenced, so far as the average number of return errors accumulated per trial in the post-control period is concerned, the relative loss being greatest for group 2 (1-2), least for group 2 (6-7). It is signifi¬ cant that the deleterious influence of the control is confined largely to the realm of the return errors and that in each group the return errors fare worse than the cul-de-sacs. Benefit, then, when it oc¬ curs, is in terms of a saving in those errors which, during the period of the control, had been prevented by physical force. In order to make a more complete analysis of the influence of control upon errors, we have superimposed the curves of the di¬ rected groups upon that of the normal group. The discussion which follows is based upon a consideration of the relations be¬ tween the scores of the normal and guided groups revealed by this device. Immediately following the period of control the curve of group 2 (1-2) 2 exhibits the large steeples which raise it alternately above and below the normal. After four trials characterized by this wavering, the curve of the directed group follows the normal fairly closely, but tends, on the whole, to fall slightly more rapidly and irregularly. The steeples following the period of control indicate confusion. It is significant that this wavering is short¬ lived. Reactions developed in the first two trials have little oppor¬ tunity to become fixed, and, hence, the period of their effect is limited. The curve for the group guided on the sixth and seventh trials (Figure 2) shows somewhat different characteristics. It remains below the normal curve following the period of direction, although exhibiting many steeples and a tendency to a slight rise in the eight trials subsequent to the control. After the period of the steeples, the curve follows that of the normal group with little significant 2 Only a few typical curves are given. The reader may know that a curve has been omitted unless it is referred to by number in the text. f/vvrs- i i i ss\ 30 ZS I I I I I I I i i i i i •I il :l :l :| :l ;i :| i 5 •1 l il 4 il :l •I ! GRO«p 4>-