M J l MIMI/J {H i g m E MU I r W! I ll 120 732 HTHS CONCEPT ATTAINMENT EN CHILDREN WITH TYPE OF PRACTICE ANB MEASURES. 07* 31103535 VARiED Thasis for £419 Dames of M. A. MICHIGAN STATE UNIVERSITY Charlene A. O’Reilfiy 2963 654," ?--c"“ £15 ILA-513 LIBRARY Michigan State University C— ' $39433" CONCEPT ATTAINMENT IN CHILDREN WITH TYPE OF PRACTICE AND MEASURES OF SUCCESS VARIED BY Charlene A. O'Reilly A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1963 ABSTRACT CONCEPT ATTAINMENT IN CHILDREN WITH TYPE OF PRACTICE AND MEASURES OF SUCCESS VARIED by Charlene A. O'Reilly The present experiment tests two basic hypotheses: (1) that subjects given practice defining while learning a con- cept graphically will do significantly better on a terminal defining task than will those who have not had this practice, and (2) subjects learning the concept verbally will do better on the terminal defining task than those who learn graphi— cally, but not as well as those who learn graphically and also have practice defining. The first hypothesis was sup- ported by the data. The test of the second hypothesis showed no significant differences, although the mean differences were in the predicted direction. The combined data was analyzed further by utilizing both of the two available measures of concept mastery (converging operations) and the agreement between the two measures was found to be significantly greater than chance. The differ- ences noted above were no longer present when this method was used, but it is possible that the methodology used in- troduced experimental error that obliterated the differences. by Charlene A. O'Reilly The methodological conclusions to be drawn from this experiment are: (1) Ratings can supply an effective dependent variable when collected under proper conditions. (2) Type of practice as well as amount of practice must be controlled. (3) Ability to define a concept after labeling it cor- rectly may be a function of the type of conceptual materials M /8 W3 used. ACKNOWLEDGEMENTS Sincere appreciation and thanks are extended to Dr. Donald M. Johnson for his guidance, encouragement, and consideration during the course of this investigation. Gratitude is expressed to Drs. John R. Hurley and Alfred G. Dietze for their criticisms of this manuscript. Special thanks are extended to the staff members of the schools involved in this study for their assistance in the acquistion of the subjects. INTRODUCTION RECRUITMENT OF SUBJECTS . . STIMULI . . INSTRUCTIONS PROCEDURE . RATINGS . . RESULTS . . DISCUSSION . SUMMARY . . BIBLIOGRAPHY TABLE SUBJECTS OF CONTENTS ii Page 12 13 16 22 26 28 TABLE OF TABLES Table Page 1 Matching data by school systems . . . . . . . 7 2 Comparison of frequency of correct defining responses in three groups of 20 subjects each . . . . . . . . . . . . . . . . . . . 18 3 Comparison of three groups of 20 subjects each on mean trials to criterion . . . . . 18 4 Agreement between two measures of concept attainment . . . . . . . . . . . . . . . . l9 5 Number of subjects who learned concept as indicated by converging operations and by definition measure alone . . . . . . . . . 21 iii [{[fl l[[l’l INTRODUCTION In previous experiments on concept mastery and concept formation researchers have shown that when subjects have proven their mastery of a concept by sorting or labeling the materials given them, there has been little assurance that they would perform successfully on a defining or other ver- bal test for mastery of the same concept. For example, Smoke (1932, p. 20) noted that "a considerable number of subjects wrote defective definitions after having gone through the test series without error," more generally, that "indi- viduals who have learned concepts may be unable to give an accurate verbal formulation of them." Logic would indicate that we cannot be sure that the sub- ject has actually learned the concept involved if he cannot successfully meet two or more criteria of concept mastery. Only after he can describe, label, classify and pass other tests of concept attainment can we be sure that the concept in its entirety has been assimilated. The necessity for the use of "converging operations" is discussed by Garner, Hake, and Eriksen (1956) in regard to perceptual tasks, and Johnson (1963) discusses its implications for research on concept formation. "Beyond the general requirement of explicitness the principal recommendation of Garner, Hake, and Eriksen, extended to the present topic would assert that, since a concept is a theoretical construct, its vali— dation requires 'converging operations.'" When converging operations have been used, the lack of agreement between measures such as labeling and defining has been an important point of discussion. we cannot expect the difficulties of any two measures to be exactly equal, but in the past, few if any, attempts have been made to equalize the difficulty of the materials and measures used. In the present experiment an attempt is made to equalize the dif- ficulty of the materials for all three groups. This study also investigates the effects of varying types of practice on performance on a terminal definition task. The attempt to equalize the difficulty of the materials consisted mainly of trying to present verbal and graphic materials of equal difficulty. The manipulation of performance on the ter- minal definition task by varying the type and amount of practice brings us to the primary concern of this study. The typical format of previous studies in this area is as follows. The subjects learn a concept such as "DAX" pic- torially or verbally and after reaching a criterion of eight or ten consecutive correct responses they are asked to define the concept. Experimenters have found that the subjects have often failed to give an acceptable definition under these conditions (Smoke, 1932). Why? Our contention is that at least a part of these results are an artifact created by the method used by the experimenters. Our thinking runs as follows. While the subject was learning the concept he was also, in a sense, ”learning to learn" pictorially. He had perhaps fifty or a hundred exposures to a pictorial or ver- bal learning situation and no exposures to a defining task until the end of the experiment. Thus the results may be a reflection of the fact that the subject had practice on one operation and no practice on the other. Our first hypothesis, then, is that subjects given practice defining while learning a concept pictorially will do significantly better on a terminal defining task than will those who have not had this practice. A second hypothesis is that subjects who learn the concept verbally will do better on the terminal defining task than those who learn pictorially, but not as well as those who learn pictorially and also have practice defining. The basis for this hypothesis is that the verbal materials are more closely related to the defining task than are the graphic materials. A brief discussion of the use of converging operations is included. The idea of converging operations implies a relationship between measures. The relationship between the task and the transfer task is examined in terms of amount of agreement in determining whether and to what extent a sub- ject has mastered a concept. RECRUITMENT OF SUBJECTS A letter requesting permission to run this experiment was sent to twenty-five school superintendents in the Lansing, Michigan area. The names of these superintendents were taken from a list of Michigan school systems located within thirty miles of Michigan State University published by the Michigan State University Placement Bureau. Of the twenty—five let— ters, seven brought a favorable response: Grand Ledge Public Schools, St. Johns Public Schools, Laingsburg Community Schools, Dansville Agricultural School, DeWitt Public Schools, Elsie Public Schools and Durand Public Schools. The Dans- ville and Durand school systems were not visited. The four school systems represented in this experiment lie in predom— inantly rual communities, although Grand Ledge and DeWitt might more accurately be termed suburban communities since they are butten miles from the city of Lansing, Michigan. SUBJECTS The subjects were children from four public elementary schools in the Lansing area. There were 60 subjects, 39 boys and 21 girls, who ranged in age from eleven years to twelve years and eleven months. The matching of subjects was done on the basis of sex and on a total percentile score or a grade equivalent score on a standardized achievement test. The test score used varied from one school system to another. A summary of the matching data is shown in Table l. The criterion for matching children was that they be within .5 units of each other on a grade equivalent score or within five percentile points of each other on an average percentile rank on an achievement test. Within each block the subjects were randomly assigned to three matched groups of twenty each. This was accomplished by putting the names of the three children into a box and drawing them out one at a time. The name drawn first was assigned to Group 1. The name drawn second was assigned to Group 2, and the remaining child was assigned to Group 3. STIMULI The stimuli consisted of fifty-four 4 X 5 cards. Twenty— seven of the cards had simple drawings of birds on them and the other twenty-seven cards had four phrases describing the birds printed on them. On each of the fifty—four cards three features were varied, each in three ways. The beaks were either short and pointed, long and pointed, or hooked. The wings were either red, yellow, or blue. The tails were either orange, green, or black. The twenty—seven bird picture cards were made such that no two cards were alike in respect to the three variable properties, but all cards were identical in every other re- spect. Since there were three different possibilities for each of the three significant features, there were twenty- seven possible ways in which they could be combined without duplication. Thus each varying element was combined with each other varying element an equal number of times in the set of twenty-seven cards. In an attempt to equalize the difficulty of the verbal and graphic modes of presentation a matching phrase card was constructed for each of the twenty-seven picture cards. For example, if the bird pictured on picture card #1 had a hooked beak, a red wing, and a black tail, the bird described ver- bally on phrase card #1 also had a hooked beak, a red wing, and a black tail. Because the picture cards contained many irrelevant cues that the phrase cards did not, such as size TABLE 1 Matching data by school systems School . . Group Group Group System Criteria 1 2 3 Grand 5th grade Stanford 53 58 58 Ledge Achievement average 90 85 85 percentile rank, 65 62 60 Form J 20 21 24 x = 57 x = 56 5 x = 56.75 St. 6th grade Metro- 7.9 8.2 7.9 Johns politan Achieve— 6.9 6.9 7.2 ment Test, Form C, 6.6 6.7 6.7 grade equivalent 5.4 5.2 5.2 score 2 = 6.7 x = 6.75 x = 6.75 Laings— 5th grade Stanford 6.8 6.6 7.2 burg Achievement Test, 6.5 6.5 6.6 Form J, grade equ— 6.3 6.1 6.2 ivalent score 6.2 6.3 6.2 6.5 6.5 6.6 5.8 6.1 5.8 4.5 4.4 4.3 5.3 5.3 5.2 4.8 4.7 4.7 5.6 5.3 5.4 4.6 4.6 4.5 x = 5.71 x = 5.67 x = 5.70 DeWitt 5th grade Total 125 122 123 Intelligence score 120 121 121 °n callfornla 2 = 122.5 x = 121.5 x = 122.0 Short-Form Test of Mental Maturity and shape of the birds, it was decided to add a fourth phrase, I'two feet with toes,‘ as a constant to each phrase card in a second effort to equalize the difficulty of the two sets of cards. INSTRUCTIONS Instructions were different for each group and are re- ported below. It will be noted that the instructions to Group 1 and Group 3, whose subjects learned the concept pic- torially, include the statement that "it is either something about the beak, the toes, the wing, or the tail." It was felt that the additional distracting cues that were unavoid- able in the pictorial presentation would make learning by this method more difficult, i.e., it would take the subjects more trials to reach the criterion by this method. In an attempt to equalize the difficulty of the two modes of pres- entation the subjects in groups 1 and 3 were told that the correct answer was something about either "the beak, the toes, the wing, or the tail." Thus the pictorial learning situation became more nearly equivalent to the task in which the subjects were asked to read the four phrases descriptive of the beak, toes, wing, and tail. Group 1 Instructions "I am going to show you several pictures of birds. No two of these birds look exactly alike. Some of the birds III '7' i l I'll-I'll I II III are gunkle birds and some are bunkle birds. I want you to look at each picture carefully and then make a guess as to whether it is a gunkle bird or a bunkle bird. I will tell you whether you are right or wrong. I want you to try to learn how to tell a gunkle bird from a bunkle bird. Only one thing about the birds is important. In other words, the correct answer is a single thing rather than a combination of two or more things. It is either something about the beak, the toes, the wing, or the tail. Do you have any ques- tions as to what you are to do?" After ten consecutive correct responses: Ask "how do you think you tell a gunkle bird from a bunkle bird?" Then: "Now I am going to show you several cards with phrases written on them describing the bird pictures you have just been looking at. Some of these birds are gunkle birds and some of them are bunkle birds, just as was true of the bird pictures. The same thing that made a bird a gunkle bird on the picture cards makes him a gunkle bird on these cards. Are there any questions?” Group 2 Instructions ”I am going to show you several cards with phrases de- scribing birds printed on them. Some of the cards describe lO gunkle birds and some describe bunkle birds. I want you to read each card carefully and then make a guess as to whether the card describes a gunkle bird or a bunkle bird. I will tell you whether you are right or wrong. I want you to try to learn how to tell a gunkle bird from a bunkle bird. Only one thing about the birds is important. In other words, the correct answer is a single thing rather than a combination of two or more things. Do you have any questions as to what you are to do?’I After ten consecutive correct responses: Ask "how do you think you tell a gunkle bird from a bunkle bird?" Then: "NOW I am going to show you several cards with pictures on them of the birds you have just been reading about. Some of these birds are gunkle birds and some of them are bunkle birds, just as was true of the cards with the phrases written on them. The same thing that made a bird a gunkle bird on the written cards makes him a gunkle bird on these picture cards. Are there any questions?" Group 3 Instructions ”I am going to show you several pictures of birds. No two birds look exactly alike. Some of the birds are gunkle birds and some are bunkle birds. I want you to look at each . lll III‘ III I! ‘ li‘i‘ll {lull-l lllltll..ll|ll|u"l. II I 11 picture carefully and then make a guess as to whether it is a gunkle bird or a bunkle bird. I will tell you whether you are right or wrong. I want you to try to learn how to tell a gunkle bird from a bunkle bird. Only one thing about the birds is important. In other words the correct answer is a single thing rather than a combination of two or more things. It is either something about the beak, the toes, the wing, or the tail. After you have guessed five times I will ask you to tell me how you think you can tell a gunkle bird from a bunkle bird. I will ask you the same question after each set of five guesses. Do you have any questions as to what you are to do?” After ten consecutive correct responses: Ask "how do you think you tell a gunkle bird from a bunkle bird?" Then: "Now I am going to show you several cards with phrases written on them describing the bird pictures you have just been looking at. Some of these birds are gunkle birds and some of them are bunkle birds, just as was true of the bird pictures. The same thing that made a bird a gunkle bird on the picture cards makes him a gunkle bird on these cards. Are there any questions?" In the few cases where there were questions following the instructions the instructions were paraphrased by the 12 experimenter but no information not specifically stated in the original instructions was added. PROCEDURE On entering the experimental room the subject was invited to take a seat on the left of the experimenter, who was seated at a table. After a brief conversation in which the experi- menter attempted to put the child at ease, the instructions were read aloud to the subject. When the subject indicated that he understood the instructions, the trials began. The cards were shuffled thoroughly before each subject started his trials and also each time the subject went com— pletely through the set of twenty—seven cards. This was done to eliminate any serial effect that might have been present and to minimize the possibility of rote memorization of the correct responses. A card was held up by the experimenter at a comfortable distance from the subject. The subject was given an upper limit of ten seconds to look at the picture or to read the phrases, although this was not mentioned in the instructions. Only two or three subjects waited the full ten seconds before responding. The response given was the subject's guess as to whether the card in question re— presented a "gunkle bird" or a “bunkle bird." The experi- menter told the subject whether the guess was correct or l3 incorrect by giving a simple "yes” or no, or by saying, "yes (or no), that's a gunkle (or bunkle) bird." The card was then turned face down on the table, and the experimenter recorded a plus (correct) or a minus (incorrect) sign on the record sheet. This procedure constituted one trial and the subject was ready for the next trial. In Group 3 the pro— cedure varied from that described above in that after every set of five trials the subject was asked how he thought he could tell the difference between a gunkle bird and a bunkle bird. He was not told whether his response to this question was correct or incorrect. After the subject reached the criterion of ten consecutive correct responses he was asked to respond to the question “how do you tell a gunkle bird from a bunkle bird?" Then he was given ten trials with the opposite set of cards from those he had been using. Thus those who originally learned from the picture cards were switched to the verbal cards and vice versa. RATINGS Ratings were used to determine whether the childrens' responses were correct or incorrect, more precisely, to what degree the individual responses were correct. Three raters were used. The first rater has a master's degree in college 14 personnel work and is currently a head resident adviser in a womens' residence hall at Michigan State University. The second rater is a female graduate student, six credit hours away from a master's degree in counseling and guidance at Michigan State University. The third rater is a male Ph.D candidate in the anatomy department at Michigan State Univer— sity. Each rater was given the set of sixty responses in a random (shuffled) order and was then given the following set of instructions. Raters' Instructions ”I am going to give you a list of sixty responses that were given in answer to the question, "how do you tell a gunkle bird from a bunkle bird?” The question was asked as a part of an experiment in which several children participated dealing with concept formation. It is now necessary that the responses of these children be scored; that is, we must de- cide whether or not they had learned the appropriate concept at the end of the experiment." "On the next page you will find a scoring scale ranging from O for a completely incorrect response, to 4 for a totally correct response. Categories 1, 2, and 3 are included for those partially correct responses which do not fall into the 15 O or 4 categories of 'totally incorrect' and 'totally correct' respectively. An 'irrelevancy' is defined as any response or portion of a response that does not include or directly refer to the correct response. The correct answer is that gunkle birds have black tails and bunkle birds have orange and green tails, although a limiting response such as 'only gunkle birds have black tails' or 'no bunkle birds have black tails' would be scored completely correct. Thus an irrelevancy would be any response or part of a response that did not per- tain directly to the color black or the word 'tail,‘ i.e., red wing or hooked beak. The response 'black wing' would be scored a '1' since it is a partially correct response (black) with one or more irrelevancies (wing)." Rating Scale When the rater had read the instruction sheet he was given the rating scale given below and a sheet on which to record his ratings. No rater had access to the rating of any other rater before or after completing his ratings. 0. Totally incorrect response Example: "The gunkle birds have hooked beaks and bunkle birds have straight ones." (or) ”I don't know. I was just guessing.” 16 l. Partially correct response with one or more irrelevancy Example: "Bunkles have lighter tails on them than gunkles, and shorter beaks." (or) ”By the color of the wings and tails." 2. Partially correct response with no irrelevancies Example: "It's something about the colors, I think.“ (or) "It's something about the tails, but I can't figure out what." 3. Correct response plus one or more irrelevancy Example: "Gunkle birds have black tails and red wings." (or) "Bunkles have curved beaks and orange tails and gunkles have straight beaks and black tails.” 4. Completely correct response, no irrelevancies Example: "The gunkle birds have black tails and the bunkle birds don't.” (or) ”Gunkles have black tails." RESULTS The reliability of the mean of the three raters' ratings was computed according to the procedure suggested by Guilford 17 (1956, p. 281). The results showed that the reliability of the average of the three ratings was very high in all groups (Group 1 = .946, Group 2 = .924, Group 3 = .933). The three ratings were summed for each subject, leaving scores ranging from zero to twelve. A score of nine was used as the cut- ting point in determining whether a response was correct or incorrect. With a rating of three representing a minimally correct response, a score of nine was needed to insure that the response in question had been given an average rating of three by the combined raters. Thus a response with a summed rating of eight was scored incorrect," while a nine was scored "correct." The data, as summarized in Table 2, support the hypothesis that subjects given practice defining while learning a con- cept pictorially do significanly better on a terminal defin- ing task than those who do not have this practice. Nine sub- jects in Group 1, fourteen subjects in Group 2, and sixteen subjects in Group 3 gave adequate definitions of the concept. The only significant difference between the three groups on performance on the definition task was between groups 1 and 3. Those subjects who did receive practice defining did sig- nificantly better on the definition task than those who did not. TABLE 2 Comparison of frequency of correct defining responses in three groups of 20 subjects each Grou s No. Correct Chi Square p Definitions Group 1 Group 2 l 9 ---------- 2 14 2.55 ----- 3 16 5.22* .53 *Significant at .05 level The three groups were also compared on the basis of trials to criterion. The results are summarized in Table 3. TABLE 3 Comparison of three groups of 20 subjects each on mean trials to criterion t Group Mean SD Group 1 Group 2 1 181.35 80.49 ——————————— 2 87.35 57.12 4.13** ----- 3 140.25 70.18 1.95 2.27* *Significant at .05 level **Significant at .01 level Had Group 3 taken longer than Group 1 to reach the crit- erion of ten consecutive correct responses it might have been 18 l9 hypothesized that the greater success of Group 3 subjects on the defining task was a function of the greater amount of labeling practice they received. Table 3 shows the t-ratios obtained by the matched groups technique (Edwards, p. 169) from which it can be seen that the actual mean difference is in the opposite direction. Although Group 1 subjects took longer to learn the concept, Group 3 subjects did signifi- cantly better on the defining task. The advantage of the use of converging operations in determining whether a given subject has actually mastered a concept was discussed earlier in this paper. Table 4 sum- marizes the agreement between the two measures of concept attainment used——the number of transfer trials correct and the correctness of the definition given. TABLE 4 Agreement between two measures of concept attainment Transfer Definition Task Task Correct Incorrect Correct 29 8 Incorrect 10 13 To see how much agreement there was between the defining task and the transfer task in terms of determining whether or 20 not a given subject had mastered the concept, the standard error of a proportion was computed (.065) according to the formula suggested by Guilford (1956, p. 175). Using an estimate of .5 as the hypothesis of chance agreement, the difference between the obtained proportion of agreement, .7, and .5 was tested. The t obtained was 3.07, which is significant at the .01 level. Thus we can see that the agreement between the two measures is significantly greater than chance. To determine what, if any, differences the use of con— verging operations would produce in the results of this study, the experimenter combined the results of the two tests of concept attainment used--the number of transfer trials cor- rect and the correctness of the definition given. The cor- rectness of the definitions was previously determined by the ratings and was discussed on pages 10-12. Of the ten transfer trials, the subject was judged to have mastered the concept if he responded to at least nine correctly. Using converging measures, then, a subject was scored as having learned the concept if, and only if, he was successful on both tests of concept attainment. Thus if a subject was scored correct on the basis of the definition task, but responded correctly on only seven of the ten transfer trials, he was not scored 21 as having learned the concept. Combining these two measures produced the results summarized in Table 5. The results using only the definition measure are also included, for contrast. The significant differences between groups 1 and 3 dimin- ished to a nonsignificant level when converging operations were used to determine whether a given subject had mastered a concept. It must be remembered, however, that these groups had been previously equated in that all subjects had performed Table 5 Number of subjects who learned concept as indicated by converging operations and by definition measure alone Group Correct Incorrect Conver in Group 1 7 13 O eratgong Group 2 12 8 9 Group 3 10 10 Definition Group 1 9 ll Measure Onl Group 2 l4 6 V Group 3 16 4 successfully to a criterion of ten consecutive correct re- sponses. Comparing the two groups after equating them in 22 respect to the: labeling 'taSk introduced experimental error that may have obliterated real statistical differences. DISCUSSION The procedurevas outlined on pages 9-10 was developed after many lengthy sessions with 11—12 year—olds in the East Lansing area. Originally the question "how do you tell a gunkle bird from a bunkle bird?” was phrased "how do you tell a good bird from a bad bird?" and it was asked after each trial for subjects in Group 3. This was found to be unwise as well as time consuming. Subjects soon became annoyed, highly anxious or bored, and tended to give the same response repeatedly. It was decided to try asking the question only after every three trials, but this too proved to be inef— ficient. Next questioning after every five trials was at— temped, and this turned out to be quite satisfactory. The subjects now appeared to be stimulated rather than bored by the questioning, and it appeared to generate greater interest in the task. Only two subjects appeared annoyed because the experimenter would not indicate whether their responses to the "how" question were correct or incorrect. Finding suitable materials to use was also a lengthy pro- cess. The materials varied in three ways and it was originally 23 decided to use the terms "good bird” and ”bad bird” rather than "gunkle" and "bunkle” birds. The correct response was "red wing." The children in the pilot study consistently ran over two hundred trials using this combination. The ex— perimenter observed that these children most frequently began by looking at the beaks, so it was decided to try "hooked beak" as the concept to be learned. This proved to be too easy a task, for most children associated "bad” bird with a hooked beak, presumably because birds of prey notably have hooked beaks. Next it was decided to try the combination "black tail" and retain the "good bird"-—”bad bird" response. This combination also turned out to be a poor one. The children this time seemed to be associating "bad” with the color black and were learning the concept with little trouble. At this point the experimenter decided that the "good bird"-— ”bad bird" responses were the main problem, and changed the required response alternatives to "gunkle bird" and "bunkle bird," retaining the property ”black tail" as the correct re- sponse. This appeared to set the problem at a suitable level of difficulty for these subjects. The problems involved in using complex but familiar conceptual materials such as the birds used in this experiment 24 appear to be quite different from those involved in using abstract materials such as nonsense syllables or geometrical forms, where difficulty level and previous contact with the material can be relatively easily controlled. Many previous associations have been formed to familiar conceptual material such as ”hooked beak" or the color black—-associations that may aid or hinder the process of learning a specific label for the concept or learning to sort the material in a new way. Although the use of familiar concepts introduces new problems, the experimenter suggests that one of the largest problems-—that of manipulating the difficulty of the mater- ials-—is not an insurmountable one. Modifying the instruc- tions offers the easiest method of control, and the intro- duction or elimination of irrelevant cues is also an effective means of controlling the difficulty of the materials. As mentioned on page 1, Smoke found that his subjects could not define after having gone through the test series without error.‘ Our results show that 39 of the 60 subjects in this experiment defined satisfactorially to a relatively strict criterion. The experimenter suggests that the use of familiar conceptual materials makes the definition task more reasonable than it was in Smoke's experiment where the subjects were asked to define a concept dealing with complex geometrical 25 forms. Further experimentation is needed to determine whether the inability to define that Smoke found to be so prevalent is a real inability or whether it is a function of the type of materials used. The attempt to equalize the difficulty of the graphic and verbal materials used in this experiment was not entirely successful, as is shown by the large mean differences (sig- nificant at the .01 level) between groups 1 and 2 on trials to criterion. It appears that introducing one irrelevant cue on the verbal cards (two feet with toes) was not suf— ficient to make up for the numerous irrelevant cues on the picture cards. Judging from the comments and responses of the subjects, the “two feet with toes" response was elimi- nated from consideration within the first fifteen trials whereas the subjects who received the picture cards examined the toes at some length, taking several trials to count them, noting whether the slant of the legs was different, etc. For future studies, it is suggested that one of two proce— dures be followed. Either more irrelevant cues could be added to the verbal cards or the instructions for the subjects who learned pictorially could be further modified to exclude the toes variable. 26 The high reliability of the average rating in each group indicates that such ratings supply an effective dependent variable when collected under proper conditions. Few exper- imenters have reported their criteria for scoring a given response, and in such cases one may assume that the decision was arbitrary and subject to the bias of the experimenter. By using a carefully prepared rating scale and giving ade- quate instructions the problem of scoring responses can be nearly eliminated. In this experiment the three raters were not previously trained. One rater is in anatomy and has had no rating experience or training in psychology. The other two are in a field related to psychology and have had no pre- vious rating experience, yet the reliability of the three combined sets of ratings is very high. The prime considera— tions appear to be adequate instructions and a scale with clearly'definedcategories rather than previous rating exper- ience or training in psychology. SUMMARY The present experiment tests two basic hypotheses: (1) that subjects given practice defining while learning a con- cept graphically will do significantly better on a terminal defining task than will those who have not had this practice, 27 and (2) subjects learning the concept verbally will do better on the terminal defining task than those who learn graphi— cally, but not as well as those who learn graphically and also have practice defining. The first hypothesis was sup- ported by the data. The test of the second hypothesis showed no significant differences, although the mean differences were in the predicted direction. The combined data was analyzed further by utilizing both of the two available measures of concept mastery (converging operations) and the agreement between the two measures was found to be significantly greater than chance. The differ— ences noted above were no longer present when this method was used, but it is possible that the methodology used introduced experimental error that obliterated the differences. The methodological conclusions to be drawn from this ex- periment are: (1) Ratings can supply an effective dependent variable when collected under proper conditions. (2) Type of practice as well as amount of practice must be controlled. (3) Ability to define a concept after labeling it cor— rectly may be a function of the type of conceptual materials used. BIBLIOGRAPHY BIBLIOGRAPHY EDWARDS, ALLEN L. Experimental design in_psychological re— search (Rev. Ed.). New York: Holt, Rinehart and Winston, 1962. GARNER, W. R., HAKE, H. W. & ERIKSEN, C. S. Operation- ism and the concept of perception. Psychol. Rev., 1956, 63, 149-159. GUILFORD, J. P. Fundamental statistics in psychology and education (3rd ed.), New York: McGraw—Hill, 1956. JOHNSON, D. M. Cognitive structures as hypothetical con- structs. Unpublished paper, 1963. SMOKE, KENNETH L. An objective study of concept formation. Psychol. Monogr., 1932, 4;, No. 4. 29 IIIII (ii; ”1111117111"Mimi