A STUDY IN OVERLEARNING WITH HUMAN SUBJECTS James H. Popham Submitted to the Faculty of Graduate Studies and Research in partial fulfillment of the requirements for the degree of Master of Science MoGill University Montreal 1951 AOKNOOŒDGMENTS I am indebted to Dr. W. N. Kellogg :for his lecture course at Indiana University in which he suggested the possibility that overlearning might account :for same apparently contradictory resulta in learning exper~ents. The present exper~ent grew out of an elaboration of his suggestions. The apparatus used in this experiment was built to my specifications by Mr. Jospeh Labrecque of the Canadian Broadcasting Corporation. INDEX. ACKNOWLEDGMENTS ••••••••••••••••••••••••••••• Page 2 THE NATURE OF THE PROBLEM •••••••••••••••••••• 4 EXPERIMENTAL :PROCEDURE •••••••••••••••••••••• 9 ................................. Procedure ................................ Apparatus 9 r·-, 9 ..................................... 13 ................... 15 SID.[LWY •••••••••••••••••••••••••• • • • •• • • • • • • 19 ................................. 21 RESULTS DISCUSSION A11D CONCLUSION BIBLIOGRAPHY "[.,'! -4- THE NATURE OF THE PROBLEM The present experiment is an attempt to show how conditions that lead to a fixated response may account for soma cantradictory resulta in learning experimenta. More specifically. it will be shown that overlearning fixates a response in terms of its kinestheticmotor pattern so that a fixated response will not readily give way to more appropriate behavior under experimental conditions that require a change in the kinesthetic-motor response. In general, maze learning experimenta do not uphold a proprioceptive theory of learning. The·majority of experimenta favor a cognitive theory although the evidence is by no means unequivocal. Dashiell's (4) open-alley multiple choice maze in which the rat follows no fixed path to the goal contraindicates a movement hypothesis in learning as does the field orientation exper±ment with rats of Hebb (6), the cortical destruction experimenta of Lashley and Ball (10), the water versus land maze of Macfarlane (11), and the tracing experimenta with humans of McGeoch (12). The supposedly crucial experiment of lengthening or shortening maze segments after learning has given contradictory resulta. Carr and Watson (2) reported interference with both equal and differentiai lengthening and shortening, Hunter (8) found no effect with either, Honzik found that equal shortening of the segments disturbed behavior but that equal lengthening did not, Kohler 1 maintains equal lengthening or shortening should not affect performance as the maze retains its "Gestà.lt" properties but that 1 Reported by Hunter (8) as persona! communication. - 5 - differentia! lengthening or shortening should interfere with performance as the "Gestalt" properties of the maze are lost. Carr_ in a later experiment (1)~ argues that the introduction of new stimuli or a greater range of stimuli disrupts behavior but that the removal of stimuli does not. Although in his experiment the lengths of the segments is not one of the variables, it is evident that lengthening the segments increases the amount of stimulation and therefore performance level should drop but shortening the segments should not affect performance as the amount of stimulation is decreased. In the experimenta where a change in the required movement did not result in a drop in performance leval~ or where the drop per- sisted for the first few trials only,.the resulta are taken to support a cognitive theory of learning_ but where performance is significantly interfered with by a variable requiring a different motor response the resulta are taken to support a motor theory of learning. degree of rapprochement might be effected by the use of criptive concepts "docile" ,and "fixated" ( 13). response Some To~an's des- Thus where the learned gives way rapidly to other behavior, should the latter be more appropriate, the learned response is said to be "docile". Similarly "fixated" is used for a learned response that does not give way rapidly to another more appropriate response when the experimental conditions are changed. An experimental situation where the response is docile would be taken to support a cognitive theory of learning, and one where the response is fixated would be cited in support of a proprioceptive theory of learning. The experimental conditions that give rise to either docile - 6- or fixated behavior have received little experimental investigation. An understanding of these conditions would, however, reconcile many of the contradictory resulta already mentioned. They would be reconciled much in the manner that the continuity-noncontinuity controversy has been partially resolved, at least in its experimental aspects, by recourse to the experimental conditions, whether the discrimination is difficult or easy, whether the discriminanda are central or high in the field of vision, whether the trials are massed or distributed, whether the motivation is strong or weak, and whether the subjects are naive or sophisticated. "Fixated11 and 11 docile 11 are comparative terms, and, rigorous- ly, an experiment should be performed under both conditions with our present knowledge of their effects in order to adjudge their relative contribution to the resulta. A response that appears fixated may, under changed experimental conditions, be comparatively docile. Similarly, a fixated respamse may be given a spurious appearanc~ of dooility by virtue of the experimental conditions. One of the sufficient conditions to establish a fixated responae is excess frequency. Krechevsky and Honzik (9) trained rats to varying degrees of excess frequency on the longer path of a maze in which the shorter was blocked. On removal of the block the rapidity of shift to the shorter path was approximately inversely proportional to the degree of overlearning. It is important to note that Krechevsky and Honzik only claim that excess frequency is one sufficient condition for fixation. Hebb (6) shows that excess frequency does not necessarily result in fixation. In the experimenta so far ci ted, any change in the experimental situation has required a new response, and behavior termed fixated or docile depending on the rapidity with which the new respQnse supplanta the old. Under conditions where a new response is not required, but the same response remains adaptive to the changed experimental conditions, we know little about the affects of fixation or docility of a response. From Krechevsky and Honzik it can be inferred that what is fixated is a pattern of movements, a kinesthetic-motor pattern, and so predicted that a change in the experimental situation which does not. require a change in the kinesthetic-motor response pattern should not interfere with performance when the response is fixated. Under such conditions a fixated response would have a false appearance of docility. It can be further inferred from Krechevsky and Honzik that when a new response is required interference with performance will be greater after the old response is fixated than before. Accepting their evidence that excess frequency is a sufficient condition for fixation, the following theory is proposed. Learning that requires a motor response is docile in its early stages. Performance involves the whole situation and any change in that situation w211 interfere with the response. After overlearning theresponseis fixated in terms of the kinesthetic-motor pattern and only a change in the experimental situation that requires a change in that pattern will interfere with performance. Where a change in motor response is required a fixated response will be more greatly interfered with than a docile response. In the following experiment, based on this theory, human subjects were set the task of learning a checkerboard maze where the correct path was tapped out with a stylus, a light indicating a correct - 8 - square and a buzzer a wrong square. After the maze was learned to a predeter.mined criterion of learning various modifications were introduced, some requiring a change of motor response~ some not. subjects then overlearned the unmodified pattern, after which the same modifications were re-introduced • • The - 9- EXPERIMENTAL PROCEDURE Apparatus A checkerboard maze of ten by ten rows of li inch square brass plates placed i inch apart on a 21 by 21 by 1 inch board painted black was used. The brass squares were glued to the board and a lead soldered to their backs which lead through a drilled hole to the back of the board. A metal unit containing a flashlight bulb and a buzzer was attached centrally at the top of the board. The pattern was made by connecting in series with flexible copper wire the leads of the appropriate squares. the wire then lead through the light to a six volt battery. nected in series through the battery. All other squares were similarly con~uzzer to the same terminal of the A metal stylus on a three foot cord·'.soldered to the other terminal of the battery completed the circuit. The stylus was about the size of a pencil 1 it had a metal tip and an insulated grip. The board was inclined at an angle of 30 degrees trom perpendioular by means of two legs which were invisible from the front. See Figure 1. Procedure After a short preliminary test on two subjects. a thirtysquare pattern was judged too diffioult for the purposes of the experiment and a pattern of twenty-one squares within a checkerboard of eight by eight squares was used. The outside square of plates was masked wi th brown paper. Only one of the brass plates in the bottom row was attaohed - 10 - Figure 1 Schamatic Diagram of Apparatus (Basic Unmodified Fattern Shown) ·-·-·!-l=l=----·-·-·- i r··-~=::l---:·-·-·-· ci-:=:-r~:-! ~ ---+--' : : Z:=; . " - 11 - to the light circuit, this constituted the starting point. one square in the top row was the goal. Similarly, The nineteen intermediate and consecutive connections formed a pattern of straight !ines and right angles. There were no diagonals. connected to the buzzer circuit. All other brass plates were The basic unmodified pattern is shawn in Figure 1. The five subjects used were collage students between the ages of nineteen and twenty-seven. They were given the following instruc- tions. "In front of you is what is called a checlœrboard maze. correct path thro~gh The the maze will be shown by the white light. Every time you touch the right square with the stylus the white light will go on and you will know that you are on the right path. Every time you touch a wrong square you will hear a buzzer and know that you are on the wrong path and must try another square. You can move only one square at a time and only in a horizontal or vertical direction, not diagonally (demonstrated). Find the correct square in the bottan row, there is only one, and then the series of intermediate squares that is the correct path as indicated by the light to the last square of the pattern which is in the top raw. Do you under- stand?" Following are the details of the experimental procedure. Time per trial was recorded with a stop watch and pne error counted for each sound of the buzzer. Part 1. The maze was learned to a criterion of ten·consecutively correct trials. The following modifications were then - 12 introduced one at a t~e for ten trials each except where noted with five trials on the unmodified pattern between eaoh. a. The light indioating the correct path was disoonnected for ten trials. .. The subjeots were told that there would be no light indicating the correct path. All ten trials are included in the results. b. The subjects were instruoted to follew the maze fran "goal" to "start". All ten trials are inoluded in the resulta. o. The sub jects performed the maze wearing a thick winter glove. All trials are inoluded in the results. d. The whole pattern was shifted one place to the right so that the form of the pattern remained intact but its relation to the rest of the checkerboard was changed. The subjects were told the pattern had been changed and one unrecorded trial was given before the ten test trials. e. An entirely new pattern of twenty-one squares was learned to a criterion of ten conseoutively correct trials. Part 2. Two hundred overlearning trials were given on the unmodified pattern. Part 3. T~e was recorded for the last ten trials only. Following the overlearning trials Part 1 a. to e. inclusive was repeated exoept that a third new pattern of twenty-one squares was substituted for 1 e. and learned to a criterion of ten consecutively correct trials. - 13 - RESULTS In all cases the error score dropped quickly to zero therefore time in seconds will be the only measure of performan ce reported. The results are summarized in Tables 1 1 2 1 3 1 4, and 5. Table 1 Averagè Time in Seconds per Last Ten Trials and Grand Mean for Five Subjects Before and After Overlearn ing the Maze. 1 2 3 4 5 Mean Before Overlearning 6.0 4.2 4.8 5.4 5.0 5.1 Atter Overlearning 4.4 3.7 4.0 4.2 3.8 4.0 Subject ,. Table 2 Average Time in Seconds for Ten Trials and Grand Mean for Five Subjects Under Four Condition s Before Overlearn ing Subject Condition 1 2 3 4 5 No Light 6.8 5.4 5.6 6.2 6.0 6.0 Reverse 6.0 6.1 5.6 6.0 5.4 5.8 Glove 5.6 4.6 s.o 5.4 5.2 5.2 Shi tt 7.5 6.6 6.8 7.0 6.6 6.9 Mean - 14Table 3 Average Time in Seconds for Ten Trials and Grand Mean for Five Subjects Under Four Conditions After Overlearning Subject Condition - ~- -~ ~--- 1 -- 2 3 4 5 Mean 1--- No Light 4.4 3.7 4.0 4.2 3.8 4.0 Reverse 6.7 7.8 6.5 6.8 5.8 6.7 Glove 5.3 5.8 5.4 5.8 5.4 5.6 Shift 5.8 5.2 5.0 5.2 4.8 5.2 Table 4 Numerioa1 and per cent Increase of the Grand Means Before and After Overlearning Under the Four Conditions Condition Before Overlearning (0 = 5.1) Per Cent Inorease Increase After Over1earning (0 = 4.0) Per -rren~ Increase Increase No Light 0.9 17.6 o.o o.o Reverse 0.7 13.7 2.7 67.5 Glove o. 1.9 1.6 40.0 Shift 1.8 35.3 1.2 30.0 Table 5 Average Time in Seconds for First Ten Trials and Grand Mean for Five Subjects to Learn a New Pattern Before and After Overlearning the Basic Pattern Subject 1 2 Before Overlearning 20.3 lUter Over1earning 15.9 3 4 5 Mean 10.8 15.6 18.4 14.4 15.9 7.5 11.4 11.8 8.2 10.6 1 - 15 - DISCUSSION AND CONCLUSIONS On the baeis of the theory it would be predicted that before overlearning all four conditions (Tables 2 and 4) would result in a drop in performance level from the preceding ten (criterion) trials (Table 1). Af'ter overlearning however (Tables 3 and 4) 1 only those conditions that require a change in the kinesthetic-motor response pattern should show a drop in performance leval from the last ten overlearning trials (Table 1), that is, the "reverse" and nglove" trials. Further, there should be a greater interference with performance after overlearning than before in the "reverse", "glove", and in the "new pattern" (Table 5) trials. With one exception the resulta in Tables 2, 3, and 4, are in conformity with the theoretical predictions. The variant results concerna performance when the pattern is shifted so that its relationship in the stimulus situation is changed but not the form of the pattern. The resulta in Table 5 are directly contradictory to the theoretioal predictions. After overlearning (Tables 3 and 4) the eue value of the light dropped to zero, there being no interference with performance, whereas before overlearning (Tables 2 and 4) removal of the light from the stimulus situation resulted in a drop in performance leve!. This is in conformity with the theoretical prediction that the pre-overlearning response is docile and affected by any change in the stimulus situation but that excess frequency fixates a kinesthetic-motor response which is not disturbed by changes in the stimulus situation that do not require an adaptive motor response. A false appearance - 16- ·of docility is given to the overlearned response and of fixation to the pre-overlearning response if the results are viewed without reference to the experimental conditions. In the "reverse" trials, where the stimulus situation remains unchanged but the required response is different, interference is greater after overlearning than before (Tables 2, 3, and 4). It is not immediately clear that this result supports the contention that what is fixated is a kinesthetic-motor response, but it does damonstrate in a different situation than Krechevsky and Honzik's (9) that excess frequency will fixate a response. In the Krechevsky and Honzik exper- iment fixation was demonstrated by a change in the stimulus situation, the stimulus situation in the present experiment is unchanged in the "reverse" trials. It might be expected that practice affects operate against these resulta, but even so, fixation by excess frequency is dominant. The change of kinesthetic eues in the "glove" trials resulted in only a slight and probably insignificant interference before overlearning (Tables 2 and 4), but after overlearning (Tables 3 and 4) interference is marked and supports the contention that excess frequency fixates a kinesthetic-motor response. The glove permitted the subjects to retain the s ame relative positioning of their fingers as without it, the affects of kinesthetic eues might be better shawn, particularly before overlearning, by a change in the stylus which would change both kinesthetic eues and finger position. Where the pattern is shifted it was predicted that there would be no interference with performance after overlearning as the - 17- kinesthetic-motor response is unchanged by the change in relationship. There is. however, considerable interference with performance after overlearning (Tables 3 and 4). There are two possible explanations for this. If it can be argued that a change in relationship ipso facto involves a change in response, then it is incorrect to state that the response is unchanged when the pattern is shifted. In which case it would be expected that performance after overlearning would be disrupted; however, it would also be expected that such disruption would be of a larger magnitude after excess frequency and consequent fixation of the response than before. As this is not the case. the more likely explana- tion is that the kinesthetic-motor response is fixated more specifically to the stimulus situation than has b een assumed 1 and a change of relationship in the stimulus situation is sufficient to interfere with a fixated response even though the motor response is unchanged. The possible affects of practice make these resulta and their interpretation equivocal. It is possible that practice acted to lower the magnitude of interference after overlearning, in which case the first explanation is the more likely. Had the pattern been shifted to the left after. oveF- learning instead of to the right. the resulta would be less ambiguous. It is to be concluded l'rom Table 5 that, contrary to the predicted result 1 excess frequency and consequent fixation on one pattern does not interfere with the learning of another. in fact learning is initially aided. In all cases the average for the first ten trials of the new pattern is lower after overlearning than before. The best explanation of this probably lies in the verbal report of the subjects that as the experimental sessions progressed they got the "feel" of what to do and how to do it. The phenomenon is likely similar ta - 18 Harlaw's monkeys (5) "learning-to-learn", and suggest to the writer that naive subjects are not necessarily desirable. In view of these results the procedure would probably be tmproved by the inclusion of several pre-experimental patterns to acquaint the subjects more thoroughly with the type of situation. It must be concluded from the present resulta however, that for naive subjects overlearning on one task leads to more rapid initial learning on another task campared with initial learning on a similar task before overlearning. It is not, in any instance, assumed that the response is identical in its kinesthetic~otor components from trial to trial, only that there is a gross similarity between responses that become more fixated with repetition. It is to be doubted that there is ever an exact repetition from one trial to another of the same movements (Cook, 3), a possible and special exception to this being muscle action potentials in the foreperiod of reaction time experiments 2 • Observa- tion of the subjects in the present experiment allowed the presence of oniy gross similarities to be concluded. 2 • Unpublished resulta repcrted by Dr. R. course at Indiana University • c. Davis during a lecture - 19 SUMMARY We have assumed that an overlearned response is a fixated response - a response unadaptable to changed exper~ental conditions, and attempted to show that this fixation is of the kinesthetic-motor response pattern. The results of the "light", the "reverse" trials support the theory. 11 glove", and possibly The "shirt" and "new pattern" variables are contradictory to the theory. The results of the "new pattern" trials are likely due to the experimental procedure and should be quite easily controlled by the suggested pre-exper~ental trials. Even so, the fixated response need not be of a kinesthetic-motor pattern to obtain the predicted results, it need be only a response 'unadaptable to changed experimental conditions' without further specification. The 11 shift" results are not readily explained and more work is required to clarify the affects of a change in relationship after fixation of the response. The uncontrolled effects of practice make any alternative explanation doubtful. In both the "light" and 11 glove 11 trials the pattern itself remained unchanged, in the former the removal of the visual eue was effective to interfere with the docile response but not the fixated, in the latter the changed kinesthetic eues interfered with the fixated response to a marked extent but not with the docile response. The "reverse" trials do not necessarily support the requirement that a motor response is fixated, only that a response is fixated, but they do give further support to Krechevsky and Ronzik (9) in a different situation. - 20 - In spite of the ambiguity of some of the results it is obvious that performance will be radically different depending on whether the response is docile or fixated. In this experiment it has been assum.ed that the response was fixated after overlearning and by overlearning, and that it was docile before overlearning. It is probable that there are other conditions that will fixate a response earlier in the learning process, these supposedly extraneous factors will make the interpretation of the resulta difficult and contradictory in a way analogous to the continuity-noncontinuity results. - 21 - BIBLIOGRAI'HY J. anim. Behav., 1917 1. Carr, H. A. 2. Carr, H. A., and Watson, J. B. Orientation in the White Rat• 1._· comp. Neurol. and Psychol., 1908, 18, 27-44. 3. Cook, T. W. Repetition and Learning. I. Stimulus and Response. Psychol. Rev., 1944, 51, 25-36. 4. Dashiell, J. F. Direction Orientation in Maze Running by the White Rat. Comp. Psycho1. Monogr., 1930, 7, No.32. 5. Harlow, H. F. The Formation of Learning Sets. 1949, 56, 51-65. 6. Hebb, D. 7. Honzik, C. H. The Role of Stimulation in Maze Learning: Specificity và. Patterning. J. comp. Psychol., 1939, 27, 149-164. - 8. Hunter, W. S. A Further Consideration of the Sensory Control of the Maze Habit in the White Rat. J. genet. Psychol., 1930, 38, 3-19. 9. Kreohevsky, I. and Honzik, C. H. Fixation in the Rat• Univ. Calif. Publ. Psycho1., 1932, 6, 13-26.-- 10. Maze Studies with White Rats. 7, 259-275. o. Lashley, K. Studies of the Organization of Behavior. 1. Behavior of the Rat in a Field Orientation. !· canp. Psychol., 1938, 25, 333-354. s., and Ball, J. Spinal Conduction and Kinesthetio Sensitivity in the Maze Habit. 1929, 9, 71-106. 11. · 12. 13. Psycho1. Rev., J. comp. Psyohol., Macfarlane, D. A. The Role of Kinesthesie in Maze Learning. Univ. Calif. ~· Psychol., 1930, 4, 277-305. McGeoch, J. A. The Psycho1ogy of Human Learning. New York: Longmans, Green and Co., 1946. Pp.xvii - 633. Tolm.an, E. C. Purposive Behavior in Animals and Men. New Yorkt The Century Company, 1932. Pp. xiv - 463.
© Copyright 2026 Paperzz