{ 1 I» W i l w ”I l l I; ll llfll W l l HTHS THE EFFECT OF IRRELEVANT FBRM CUES UPQN LEARNENG AND TRAE‘ESFER 0F COLOR CONCEPTS EN ‘t'OUNG NURMAL CEQELDREN Thesis far the Sages of M. A. MECHiGAi‘é STATE UE‘éi‘ngRSE'EY BRUCE LOWELL BACHELDER 1359 At al-..-...*- A 9mfl'fl3ds.{:‘! t" LIBRARY THESIS Michigan State. 3 University I M .r‘ ..2 m1 ur- ABSTRACT THE EFFECT OF IRRELEVANT FORM CUES UPON LEARNING AND TRANSFER OF COLOR CONCEPTS IN YOUNG NORMAL CHILDREN By Bruce Lowell Bachelder The present study stemmed from efforts to improve the efficiency of teaching color concepts to retarded children. The behavior taught was a conditional discrimination in which the E5 learned to choose one of two colors depending upon which of two color names was spoken by E. Thus, when E said "red” a reward was found under the red discriminandum T1 Iffi H and when p said olue a reward was found under the blue discriminandum. A stimulus—response analysis of this learning paradigm suggested that form cues would influence the development of the color concepts because § would be reinforced in the presence of both form and color cues. As a result the form cues should exert some control over §fs reSponding. Three levels of form variability were compared in their effects on learning and transfer of color concepts. Nermal children, 3-h.5 years old served as gs in a withinggs design in which 2% 'observations' under each of three conditions were made on each § (for a total of 72 observations of each g). Each observation was a conditional discrimination problem with six training trials (which used colored geometric forms) and one transfer trial (which used colored silhouettes of familiar objects). The E spoke color names on each trial to provide the conditional cue for correct responding. In the problems of condition I the form cues were the Bruce Lowell Bachelder same within a trial and between trials of each problem. In problems of condition II the form cues were identical within each trial but changed from trial to trial. In problems of condition III, form cues differed within trials and changed from trial to trial. No main effects due to conditions were found. There was slight evidence of interactions between conditions and blocks of practice for training trials and transfer trials. The interaction of conditions with blocks of practice on training trials was such that the curves of the number of correct responses as a function of blocks of practice were positively accelerated in the case of conditions II and III but negatively accelerated in condition I. In addition, in the first block of practice the curves for conditions II and III are higher than the curve for condition I; in blocks 2 and 3, the curve for condition I is very slightly higher than the curves for conditions II and III; in the final block of practice the curves for conditions II and III are again above that of condition I. The interaction of conditions with practice on transfer trials was such that transfer following conditions II and III was relatively con- stant throughout all training sessions while the transfer following condition I was at first below that of conditions II and III then rose above conditions II and III in the final block of practice. It was concluded that for normal gs with NAs about 3-h.5 years, the form cues had little effect upon learning and transfer of color concepts, although there was some evidence of interactions of condi- tions with practice for both training and transfer of the color concepts. This conclusion was not extended to retarded children for whom there was still some reason to feel that the form cues would be important. Bruce Lowell Bachelder A theoretical discussion of the conditional discrimination was pre- sented. The interaction of conditions with practice on training trials was attributed to an error factor Operating only in condition I. The interaction of conditions with practice on transfer trials was attributed to external inhibition of cue responses. A number of suggestions were made for improving the efficiency of teaching color concepts to retarded children. THE EFFECT OF IRRELEVANT FORM CUES UPON EARNING AND ‘IRANSFE OF COLOR CONCEPTS IN YOUNG NOREAL CHILDRE By Bruce Lowell Bachelder A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1969 ACKN OW EDGE-ENE S I wish to thank the members of my thesis committee for patience and flexibility. Each has been especially helpful. Dr. M. Ray Denny advised me as I began to develop the original ideas for research. He continues to influence my thinking via elicitation theory. Dr. William Stellwagen was quick to assume Dr. Denny's place on my committee while Dr. Denny was out of town for a year. He helped me through administrative matters and provided helpful suggestions for this and future research. Dr. Lester Hyman, acting as my chairman, kept me pointed toward the goal. At times he was solely responsible for renewed activity on my part. He helped reveal order I did not believe existed in my data. Dr. Lauren Harris devoted considerable time to criticism of my writing. Again and again, in this and other contexts, he broadened my understanding of psychology and behavior. He gave me problems I had never had and changed the character of problems I had had a long time. ii TABLE or CO INTRODUCTION . . . . . . . . . . . . . I-TETHOD . . . . . . . . . . . . . . . . RESULTS . . . . . . . . . . . . . . . DISCUSSION . . . . . . . . . . . . . . LIST OF REFEREI'JCES . . . . . . . . . . iii LIST OF TABLES The complex forms . . . . . . . . . . . . . . . . . The color test . . . . . . . . . . . . . . . . . . Between 83 balancing of conditions each day . . . . Analysis of variance of the effect of blocks of practice and conditions upon the number of correct responses on training trials . . . . . . . . . . . Analysis of variance of the effect of blocks of practice and conditions on the number of correct responses on test trials . . . . . . . . . . . . . Analysis of variance of the effect of blocks of practice and blocks of trials on the number of correct responses . . . . . . . . . . . . . . . . Analysis of variance of the difference in number of correct responses given on trials 5-6 and trials 7 O O O O O O O I O O O O O O O O O O 0 O 0 iv Fri 0) 18 23 29 33 Mlldr afa o o o a o o o o The portable Wisconsin general test apparatus . . . The complex forms . . . A flow~chart of st progress through the screening and pretraining problems and experimental problems The effect of practice on the performance on training trials of learners and non-learners The effect of practice and conditions on the performance on training trials of learners The effect of practice and conditions on the performance on test trials of learners The effect of practice and trials on the performance of learners Page . lO . 15 0 10-17 . 20 . 3l . 3A - 35 INTRODUCTION The application of the data of learning experiments and of learning theory appears to have improved methods for training the severely and moderately retarded (IQ 20-50) to the extent that these individuals are now learning skills once felt to be well beyond their capabilities. These skills include subtle visual discriminations (Sidman and Stoddard, 1966), reading and learning simple qualitative and relational concepts (Denny, 1966 and Evans, 1968), and diligent classroom behavior (Bijou, Birnbrauer, Kidder, and Tague, 1967). Initially, the present research was directed at studying the learning of color concepts by retarded children. Prior training attempts by the author suggested that severely retarded children may have more difficulty learning to respond apprOpriately to color names than to names of other qualities such as rough-smooth, large-small, and relational words such as up-down, and before-behind. Thus, in part, the present paper represents a stimulus-response analysis of the problem of learning colors. This analysis attempts to understand this difficulty and to suggest more efficient training methods. A pilot experiment on the problem used retarded children as Ss but proved so tedious that it was decided that normal children who did not know their colors would be better Ss, at least during the develOpment of a new experimental paradigm. The choice of normal children also seemed appropriate because a variety of studies (see Denny, 196A) have shown 1 2 that the performance of normal children on learning tasks is compar- l able to the performance of retardates of similar mental age. History of training retarded peeple The approach to the training of the retarded that is taken in the present paper comes out of the learning theory tradition in psychology, as exemplified in the work of Skinner and Hull. As such, it em- phasizes the training of behaviors, or responses. This approach can be contrasted with one that has its roots in British empiricist philos- Ophy, as exemplified in the writings of John Locke. Locke asserted that all knowledge comes through experience, either through the senses or through reflection on sensory data. For the education of the retarded, this theory suggested an emphasis not on responses or behaviors directly but on the 'prior’ senses or perceptions. As l The normal children were not intended to substitute for retarded children. Complete examination of the problem would include one group of retardates and two groups of normal children. One of the two normal groups would have mental ages (MA) comparable to the MAS of the retard- ed Ss, while the other normal group would be comparable to the retarded Ss In chronological ages (CA). The use of two normal groups was suggested by Denny (196A) not only to provide developmental data, but to classify the deficits of retarded children into one of two types: the low IQ deficit or the low IQ-low MA deficit. If the retarded group performs below the equal- MA group, we can conclude that the deficit is a low IQ deficit, since the two groups differ only in IQ. If the retarded group performs below the equal-CA group, then the deficit is a low IQ-low MA deficit because the groups differ on both measures. Both normal control groups are logically necessary to discriminate the low IQ deficit from the low IQ-low MA deficit. Denny considers the low IQ deficit to be of greater import for understanding retardation itself because the low IQ-low MA deficit is readily attributable to the retarded person’s being at an early stage of intellectual develOpment (as defined by IQ test scores). If, how- ever, compared with normal children at an equivalent stage of intel- lectual development, retarded peeple are still deficient on certain types of tasks, then such evidence strongly suggests that these low IQ deficits are fundamental to retardation itself. Analysis of low IQ deficits may result in improved concepts of intelligence and mental age. 3 Kolstoe (1956) has suggested, modern educational practice has attempted to include both sensory and behavior training for the retarded, but the stress plainly has been on the former. arias The first serious attempt to train a retarded person, undertaken by Jean Itard between 1800 and 1805, was based upon Locke’s empiricist philosophy. Itard felt that mental retardation resulted from insuf- ficient sensory stimulation. He attempted to test this idea by curing the famed "wild boy of Aveyron", a boy discovered living 'wild' near Aveyron, France, in 1799 (Itard, 1962). This boy was generally indo- lent, had poorly deveIOped sensory discriminations and emotional responses, and ran wild through the fields whenever the opportunity presented itself. The boy was observed to lift potatoes from boiling water with no apparent response to the high temperature. Loud noises such as a gun shot did not cause him to start or run, yet he was able to reSpond to soft sounds such as the approach of friends. He formed a close emotional relationship with Itard's housekeeper, then with Itard, only after prolonged close social contact; but never adjusted normally to others. By a variety of ingenious methods, Itard made marked improvements in the boy's auditory, tactual, visual, thermal, olfactory, and gusta- tory discriminations; yet after five years Itard gave the project up as a failure because the boy was still unable to adjust normally to society. Seguin Edouard Seguin (1812-1900), strongly influenced by the changes Itard had effected in the wild boy, continued to develOp training k methods based upon sensory training. He agreed with Itard that mental retardation stemmed from insufficient sensory stimulation, but where Itard assumed that the lack of stimulation resulted from insufficient contact with a stimulating environment, Seguin argued that the cause lay in damage to the central or peripheral nervous systems. Conse- quently, Seguin's training program consisted of bombarding the nervous system with strong and persistent stimulation which he felt would get through to the central nervous system and eventually cause it to function normally. He used a system of muscular exercise and sensory training fit to the individual needs of the child. Montessori Montessori (1870-1952) also believed that sensory training would cure mental retardation and had great success first with retarded, then with normal, children. Her chief contribution was a series of 26 "auto-educational" devices which the children used individually to develOp all the senses except smell and taste. The new emphasis on behavior training About the turn of the century, educators began to question the utility of merely training the senses to the exclusion of practical social and vocational skills. Duncan (19A3) advocated a teaching method in which academic subject matter is correlated with shop work, crafts, and home economics. Ingram (1960) suggested a similar approach in which academic subjects are taught in a unit of activities relevant to the life of the children. Kblstoe (1956) described this new influence as a "...radical switch from stimuli to response." He sug- gested that It would appear that up until the age of about 10 or so, 5 the major curricular emphasis should be on the systematic presentation of stimuli in a sense-training program. This emphasis should gradually change until youngsters at a junior high school level are involved in a response em- phasis program. In this program general education provides the child the Opportunity of increasing his intellectual efficiency up to the limits imposed by his nature, and specific behavior training insures the development of patterns of behavior of a socially acceptable nature. (Kblstoe, 1956, Page A) The dichotomy between behavior development and sensory development The passage above is based upon a distinction between development of the intellect and develOpment of specific behaviors. In this frame of reference, specific behavior training is justified only by the need to make the child socially acceptable. Following the same line of reasoning, the proper time for behavior training is after maximal deve10pment of intellectual ability by means of sense training. The distinction between sense training and behavior training is based, how- ever, upon an untenable distinction between intellect and behavior. Kolstoe overlooks the fact that sense training consists of teaching specific behaviors. The stimulus-response learning theory approach to training the retarded recognizes that intellectual abilities are made up of specific behaviors and are develOped through their training. Druner (1960) has made a similar kind of point in his discussion of the distinction between "doing" and "understanding". According to Bruner, classic texts in problem solving draw a sharp distinction between "rote drill" and "understanding", a distinction Bruner feels is unclear. Bruner argues that understanding comes through drill, a view he finds supported by his discussions with mathematicians, who say that their understanding of mathematics came about through practice at computation. 6 It has been shown with normal and retarded children as well as with rats and monkeys that, after training on specific behaviors, sub- sequent learning of similar behaviors takes on a more sophisticated, that is insightful, character often attributed to the intellect. The procedures used in this type of learning study are all similar in that S learns many different problems of a type, one after another. Even- tually, S'begins to solve each completely new problem in a very short time, typically in one or two trials. This phenomenon of "learning to learn" is called learning set. Learning set has been studied extensively by Harlow (19A9, 1959). In his studies, Rhesus monkeys learned two-choice discrimination problems. The first problems typically required many trials to learn, but after learning hundreds of specific problems of this type, the monkeys solved each new problem (new objects) in a single trial. House and Zeaman (1963) report results similar to Harlow’s but in studies in which retardates served as Ss (MAS AA-96 mo., CAs 105—235 mo., and IQs 26-73). Their report covers five different studies, each of which presented S with a large number of short (three trials only) two-choice discrimination problems. The Ss attained from 60% to 80% correct choices on the second trial of these problems. This very simple two-choice problem is not the only type of activity in which retardates show increasing ability to learn after learning simple behaviors. Denny (1966) has develOped a training method by which severely retarded adolescents have been taught simple qualitative and relational concepts. These Ss showed the same increas- ing facility at learning new concepts as did the S3 of House and Zeaman's report. Denny estimated that as many as 18 months of diligent 7 training might be needed to make significant improvement in the learn- ing rate of some of the more severely retarded Ss, but he is confident that the rate would increase. Examples of training programs After reviewing the learning literature relevant to retardates, Denny (196A) concluded that he was Optimistic about training the retarded. As suggested by the literature, he summarized the assets and deficits of retardates as follows: (1) they seem to be deficient in inhibiting a response once it is learned, (2) they seem to be poorer than normal children on complex learning even when matched on mental age, (3) they have a deficit in verbal learning, (A) they lack verbal control of motor behavior, (5) with practice they learn nearly as well as normal persons on motor tasks, (6) rote learning tasks are nearly as easy for retardates as for normal persons if familiar materials are used, (7) their retention seems quite adequate, (8) they can learn to use verbal mediators if specifically trained to do so. Training a visual discrimination Discrimination learning is a task on which retardates show a low IQ deficit which becomes more pronounced as the discriminanda are made more similar (Denny, 1964). Denny suggested that this difficulty might be understood in terms of the retardate's difficulty inhibiting wrong responses, a difficulty which suggests that training would be more effective if steps were taken to prevent wrong responses from ever occurring at all. One way to prevent errors is to start with a simple problem and 'fade' into the more difficult prdblem in a carefully planned progression. In this transition from problem to problem S can continue to respond on the basis of the easier problem, yet at the same 8 time respond apprOpriately to the new problem. Sidman and Stoddard (1966) used fading to teach a difficult visual discrimination between a circle and an ellipse to a profoundly retarded adult (IQ<:20). The subject worked in a booth containing a response panel made of nine projection screens arranged in a 3x3 square. Circles and ellipses could be projected onto these screens, and the brightness of the backgrounds of each figure could vary from dark to bright. The first problem S had to learn was an easy brightness discrimination in which he always touched the one screen which was brightly lit. When this problem was well learned, circles, squares, and X5 were projected onto the bright screen so that when the S chose the bright screen he also chose the figure. This discrimination was then faded, in 20 brightness increments, into a discrimination of 'screen with the figure' from equally bright ’screens without a figure'. The final fading sequences involved fading in the ellipses, accomplished by projecting barely visible ellipses on all the incorrect keys. Gradually, these were made more distinctive until the §_was responding correctly to the circle without help from brightness or distinctiveness cues. Brightness variations, form positions, timing, and response recordings were all automatic. So long as the §_did not make errors he progressed step by step in a predetermined sequence, but when he erred, the equipment backed up one step from which the §_could continue forward as before. Most errors were isolated events reflecting a lapse of attention, but at difficult transitions S oscillated between two levels, unable to choose correctly in the new stimulus situation. When this happened new steps were added so that the transition was easier, 9 errors were eliminated, and progress was restored. Certain other areas of the program were so easy that steps could be removed without producing errors. After extensive revisions of this sort, a training sequence was achieved which taught the discrimination with a minimum of effort on the part of the teacher and the subject. Teaching simple concepts Simple concepts such as up-down, push-pull, in-out, over-under, rough-smooth, long-short, thick-thin, and simple reading have been taught to severely and profoundly retarded Ss by means of a training method develOped by Denny (1966). The implications of Denny's "elicitation theory" are translated into practice by the use of a teaching machine called "Mudrafa" (Egltiple Differential Response and Feedback Apparatus, Figure l) which aids in precise stimulus manipula- tion, assists in providing stimulus feedback relevant to the concept being learned, gives §_immediate knowledge of results, and prevents errors. Since all responses are made by pushing a bar, training of non-verbal Ss is quite as easy as training verbal ones. According to elicitation theory, when stimulus elements are con- sistently paired with a response, these stimulus elements will begin to elicit that response. Accordingly, during training Ss are induced to give particular responses in the presence of well controlled stimulus conditions. Careful analysis of the stimulus is made so that only relevant cues will be paired consistently with the response. All irrelevant cues must be varied often so that they do not compete with the correct stimulus. Since the SS learn concepts, words are part of Y the stimulus complex. In teaching up-down, for example, §_says 'go up", and prompts §_to move the response handle upward. 10 Rotating face plate Response handle Response slots r Rotating face plate Response handle = Slot access control panel Figure l. Mudrafa. ll Prompting of the desired response may be as simple as pointing out the direction in which the handle should be pushed or as complex as telling the §_what cues to look for. Denny suggests a systematic progression from simple prompts to complex verbal prompts which make use of previously learned concepts, for example, "it's the big one". By this means early training becomes very important to success, and continued practice on earlier concepts helps to ensure their retention. As prompting becomes very complex, S approaches the point at which he is learning in a way which is similar to ordinary classroom proce- dures where concepts are 'explained’ and responding is highly verbal. The basic components of Mudrafa are the rotating face plate, the response handle, the response slots, the slot access control panel, and the knowledge of results (KR) light in the handle (Figure l). The face plate rotates in a complete circle so that each slot can be in each of the four positions for several trials during each training session. By this means the slot cues are made irrelevant. The response handle will move into each of the four slots so that with the bare faceplate S can learn the concepts of up-down, right—left, thick- thin, and long-short. Most training is done with stimulus materials placed upon the face plate so that the S selects one of as many as four pictures placed at the end of the slots. Knowledge of results is given by lighting the bulb in the end of the response handle. Each response slot is a different width and length so that thick-thin, wide- narrow, and long-short may be taught without other stimulus materials. Each slot gives differential feedback relative to the concept it teaches. The long slot requires a long push and a short slot only a short push. In the wide slot the handle can move freely from side to 12 side within the slot, in contrast to the narrow slot which is just wide enough to accept the handle. Each slot can also be used in visual and verbal prompting, for example, "the long one, or the fat one", cueing S‘to the correct slot. The slot-access-contrOl-panel allows §_to control errors by blocking any combination of incorrect slots so that §_cannot push the handle into them. The S cannot tell that a slot is blocked unless he tries it. In early stages of concept development, all slots but the correct one can be closed so that S finds the correct one easily. Later, when §_is skilled at responding to prompts and can inhibit unrewarded responses, all slots can be Opened and errors allowed. The use of the slots for differential feedback is only one ex- ample of the machine's use for providing differential feedback. Denny suggests that such feedback be provided for all concepts taught. In teaching the concept of 'through', for example, brushes or tissue paper can be placed over the correct slot so that visual and tactual feedback, relevant to the concept ’through', occurs when the bar is pushed into the correct slot. Teaching colors The present study grew out of work on Denny's program and is as much concerned with the use of a spoken word as part of the stimulus situation as in form manipulation during color learning. Knowing colors can be operationally defined in terms of these specific behav- iors: choosing a color sample in response to the color name, saying the color name in response to a color sample, generalization of these responses to similar hues, discrimination of several hues, and accom- plishing all these behaviors despite variations of context, form, and 13 size of the colored items. _Choosing a color sample in response to a color name was the behavior taught in this study. If retarded people do not learn this skill readily, it may be due simply to a slower rate of association of stimuli and response in which case only extra training time need be invested. Pilot work suggested, however, that severely retarded adolescents have relatively more difficulty with this color task than with simple concepts and simple reading. This difficulty could stem from misdirected attention. If Ss associated color names with form cues instead of, or in addition to, color cues, responding would be inaccurate. If such were the case, the retarded person's poor ability to inhibit responses would make learning even slower because of the added difficulty of giving up the incorrect response. The role of form cues during training and transfer of skill in choosing colors is the tOpic of this paper. Three levels of form manipulation were compared in their effects upon learning and transfer of color concepts. Apparatus The apparatus was a Wisconsin General Test Apparatus (HGTA) modi- fied for portability (Figure 2). The equipment could be rapidly assembled and disassembled and carried by a nylon strap. A vertical screen prevented §_from seeing E, yet E could easily observe st choice through the Opening in which the drawer moved. Between trials the drawer was pulled to Efs side and the vertical end of the drawer blocked S's view of the stimulus materials. When the new discriminanda were in place, the drawer slid easily to the other side for §fs choice. Discriminanda Pictures. Two color photographs, 5%-x 3% inches, Of a leopard and an elephant were cut from a child’s book of animals (Haas, 1965). Geometric form". Six geometric forms were cut from red, blue, green, and yellow construction paper. The colors of the papers close- ly approximated the Munsell (1929) numbers R A/l2, B 7/3, G 6/5, and Y 8/8. The forms were: 1) circles, 2 inches in diameter, 2) squares, l-3/h inches on a side, 3) equilateral triangles, 2 inches on a side, A) pommEe crosses, 1/2 inch wide, with arms 2 inches long, 5) diamonds with axes of 2 inches and 1-1/2 inches, and 6) stars which fit tightly within a 2 inch square. Complex familiar forms. The complex forms were silhouettes of Objects likely to be familiar to young children. Each was cut from each of the four colors as had been used in the construction of the geometric forms. Each form was drawn to fit closely within the area Of a 2-inch square. All 2A such forms are reproduced in actual size in Figure 3 and listed in Table 1. 1A Hinge n for ._. ' folding * Screen, —— removable for carrying 24" Hinge Drawer for _ :- folding ' l L j d L 36" J §fs side E's side The sliding drawer Side view has been pushed toward §o Screen ,, ’ i ‘ Handle Food wells <:> Drawer §'s side Q's side The sliding drawer TOE view has been pulled toward E. Figure 2. The portable Wisconsin general test apparatus. 18 Table l. The complex forms 1. Heart 7. Tree 13. Christmas tree 19. Light bulb 2. Dress 8. Elephant 1%. Sock 20. Apple 3. Cup 9. Teddy Bear 15. Shoe 21. Guitar h. Ice Cream Cone 10. Santa Claus 16. Lamp 22. Snowman 5. Fish 11. Leaf 17. Truck 23. Flower 6. Head of a 12. Key 18. Mitten 21+. Head of a horse rabbit Procedure Each S was trained on two-choice discrimination problems in his own home. The WGTA was assembled on the living room floor. Both S and §_sat on the floor. Often parents and sometimes children were present during training, but they sat on Eds side of the apparatus. Their presence tended to be a slight distraction for most Se, but only during the first session. The E told §_that he had a game in which §_could find a lot of candy. The E then demonstrated the game by putting a piece of candy in one food well partially covered by one of the animal pictures. The §_then presented the drawer to §_and waited for S'to take the candy. At times §_had to coax §_to respond either by saying, for example, "look, here's the candy,‘ or by leading him to the food wells. A5 soon as §_began responding rapidly these demonstrations were terminated and pretraining was started. In all subsequent problems when S responded correctly, i.e., moved the positive discriminandum aside, he found a raisin, an M&M, 19 or a piece of sugared cereal in the food well of the WGTA. When S made an incorrect choice, §_withdrew the drawer immediately so that S was allowed only one choice on each trial (the non-correction method). Some SS ate the reward immediately, others saved some or all rewards in a bag. Each S took an average of 15 sessions, each 15-30 minutes long, to complete two pretraining prdblems, a test of color and color-name association, and 12 sessions of experimental problems. The experi- mental problems were conditional discrimination problems in which each of two color cues was correct an equal number of times in each problem. The §_loudly spoke the name of one of the two colors at the beginning of each trial to provide the cue for correct responding. For example, when E said "red", the red discriminandum.was correct; and when §_said "green", the green discriminandum was correct. Each step of the sequence of pretraining and experimental problems is described below and schematized in Figure h. Subjects The SS'were normal children solicited by telephone calls to their parents or suggested by parents of children already in the study. All children were living in Michigan State University apartments for married students. Twenty-two children were found. All 22 were ad- ministered a test of color and color-name association as well as two pretraining problems which trained each child in both conditional and non-conditional two-choice discrimination paradigms. These screening problems are described in detail below. From this group of children, only those children were chosen as Ss who failed the color test but passed the pretraining problems. Six children, varying in age from 3 2O First~ Pr9' Fail Drop :) training Pass Second F 1 pre- ‘ 81 ( training DrOp :) problem I Pass The experimental sessions. Each session had six problems. Session 7 Session 10 Session 2 Session 5 {Session 8 > {9531011 1) Session 3 Session 9 Figure #. A flow-chart of §‘s progress through the screening and pretraining problems and experimental problems. 21 years to H.5 years (2.: 3.6 years) qualified. The first pretraining problem, administered to all 22 potential Es. Four children failed this problem. The first pretraining problem was a simple conditional discrimination problem in which the correct_ choice was dependent upon a word spoken by §_on each trial. This problem served to train Se to listen for the conditional cue, the spoken word, then to choose one of the discriminanda for reward. The pictures of the leOpard and the elephant were the discriminanda and the words "leopard" and "elephant" were the conditional cues spoken by §_as he pushed the drawer of the WGTA toward S, The position of the correct discriminandum varied according to a Gellerman series (Geller- man, 1933). The conditional cues were chosen randomly except that E spoke each word no more than three times consecutively. Training sessions, each 2h trials long, were repeated daily until §_made seven consecutive correct responses or had completed three sessions. In the former case, §_passed on to the color test; in the latter case, §_was dropped from the study. The color test, administered to the 18 children who successfully, completed the first pretraining prOblem. Five children failed this test, that is, did not continue on to the final pretraining problem. Four knew their colors and one refused to cooperate further. Following successful completion of the first pretraining problem, the color test was administered to eliminate SS'who already knew their colors. The color test was a conditional, two-choice color discrimination problem with color words as the conditional cues. The complex familiar forms were the discriminanda. Each of the four color names (red, blue, green, and yellow) was called six times for a total of 24 trials. Each 22 of the six possible pairs of colors appeared four times in a random sequence restricted so that a specific color appeared in a maximum of three successive trials and a specific color pair appeared in a maxi- mum of two successive trials. Colors and position were selected so that left was correct as often as right. The forms used on trials 1-12 of the color test were selected randomly from the pool of 2M complex familiar forms. In trials 1-12, then, twelve forms were positive, that is, correct and rewarded, and twelve forms were negative. In trials 13-24 the complex forms were selected randomly again except that previously positive forms were paired only with other previously positive forms and previously nega- tive forms were paired only with other previously negative forms. This pairing of positive with positive and negative with negative prevented systematic transfer of choice based upon the previous posi- tive or negative value of each form cue. To qualify for the final pretraining problem and for participation in the experimental sessions, Ss had to score fewer than l8/2h correct on this test. Calculated by binomial expansion, the probability of achieving 18 or more correct out of 2M by random responding is less than 1%. One color test was constructed and used for all Ss (Table 2). The secondgpretraining problem, administered to the 13 children who passed the first pretraining problem and scored fewer than 18 out of 2h correct on the color test. Six children failed this problem leaving 7 qualified Ss. One of these later refused to c00perate and was drOpped. The second pretraining problem served to orient the Se toward color cues. This problem differed from the other problems (the first pretraining problem, the color test, and the experimental Table 2. The color test. Discriminanda Trial Left Right The correct color 1 g lamp y tree y 2 b flower r santa Claus b 3 r shoe b leaf r h g mitten y guitar y 5 y horse b Christmas tree y 6 r ice cream cone g fish r 7 b rabbit r snowman r 8 y elephant b sock y 9 y light bulb g cup g 10 g dress b apple b 11 b truck y heart b 12 r key g bear g 13 y bear g ice cream cone g 1h b snowman r apple b 15 g tree b cup g 16 y dress r lamp r 17 y Christmas tree r sock y 18 g santa Claus b heart b 19 g truck b elephant g 20 r rabbit y leaf y 21 g flower r horse g 22 y key b fish b Table 2 (cont'd.) Discriminanda Trial Left Right The correct color 23 g shoe r guitar r an r light bulb y mitten r r = red, b = blue, y = yellow, g = green. 25 problems) in that the second pretraining problem was not a conditional discrimination. The E did not speak a word and §_had to learn to select a red star consistently when presented with a red star and a green star. As in the first pretraining problem, sessions were 2h trials long and continued until §_made seven consecutive correct responses or had completed three sessions. In the former case he began the experimental problems in his next session; in the latter case he was dropped from the study. Experimental variables. Each of these 6 remaining 85 served in all conditions, receiving 2h each of three types of problems for a total of 72 problems in 12 experimental sessions. Each of these pro- blems was 7 trials long. Trials 1-6 of each problem, the training trials, used the geometric forms while trial 7, a test of transfer, used the complex familiar forms. The three problem types differed with respect to the way form cues varied in the training trials (1-6) of each problem. Typ I. In type I problems the form cues of training trials did not differ within or between trials, for example, blue square versus red square on all six training trials of the problem. The form cue used in each type I problem was selected randomly without replacement from the pool of geometric forms. This meant that in successive type I prOblems all six form cues were used once before repetition. Type I . In type II problems, form cues did not differ within a trial but varied from trial to trial, for example, red circle versus yellow circle on trial one, then red cross versus yellow cross on trial two and so on through six training trials. All six form cues were used in each type II problem. The sequence of form cues was 26 random for each type II problem. Type III. In type III problems the discriminanda on each trial differed in both form and color, for example, red circle versus green square on trial one, then red cross versus green star on trial two. All six form cues were paired with the two colors of the problem and appeared once in trials 1-6. The sequence of color and form cues was determined randomly as follows. The twelve discriminanda formed by the pairing of six geometric form cues and two color cues were shuffled, drawn two by two, and assigned in order to trial 1, trial 2, and so on through trial 6. The resulting random sequence was re- stricted, however, by eliminating pairs in which the form cues did not differ. Trials 7, the transfer trials, were comparable for all problem types. Form cues were selected at random without replacement from the pool of complex familiar forms. Twelve problems used up all the avail- able forms so that in the next twelve problems, just as in the color test, previously positive forms were paired only with previously positive forms. Throughout all experimental sessions a specific pair of forms was not repeated for a given E. Randomization_procedures. In each of the twelve experimental sessions §_received 6 problems. Each problem taught one pair of colors selected from among the six possible pairs formed from red, blue, green, and yellow. Each color pair was combined with each con- dition (problem type). Three conditions combined with six color pairs formed 18 color-problem combinations (CICs). For each S, the order of presentation of the 18 CPCs was random, with the following constraints: 1) a specific color appeared in no more than two consecutive problems; 27 2) two consecutive problems of one type was a maximum; 3) a particular CPC was not repeated until at least six other CPCs had intervened; h) each l8-problem-block (3 days) comprised all 18 different CPCS. A block of problems, then, was made up of one each of the three problem types combined with each of the six possible color pairs so that in each block, color pair and problem type were completely crossed and balanced. In the training trials (1-6) of a given problem, each color name was called three times. The sequence of color names was restricted so that the three calls of one color did not occur consecutively. Six- trial Gellerman series were used to determine position of correct choice and the sequence of color names called out. Several of these six-trial series were available, each of which was combined with every other; one series for position determination, one series for color name sequence determination. Some of these combinations resulted in completely confounded position and color cues, that is, one color was always on the same side. All of these confounded pairs were discarded, leaving a pool of #0 useful pairs. These pairs were selected randomly for every experimental problem. The position of the correct color on transfer trials was deter- mined independently of the sequence on training trials. Since a daily session had six problems with one transfer trial each, one of the six- trial series was selected at random for use each day. The color name called on trial seven was determined randomly except that within each (unlor pair each color name was called an equal number of times in each block of three days. On any one day each §_completed just one third of a block of 28 problems. This meant that unless specifically planned for, each condi- tion and color pair would not be represented equally each day over the whole group. To balance these factors each day for the whole group of Ss, CPCs were assigned to groups of three 85 in such a way that each 0’) day these 'yoked' gs completed all 18 CPCs. A a result of this pro- cedure, days of practice could be analysed as a completely crossed and balanced factor. The assignment of the 18 CPCs to the three yoked S8 was done as follows. A sequence of CICs was constructed for one S such that each of the three blocks of six problems (problems 1-6, 7-12, 13-18) could precede or follow any other block without violating sequence restrictions. Each of the three yoked Ss received the three blocks in different orders for a total of two repetitions of the 18 CPCs, making a total of 36 problems. This procedure was repeated using a new basic sequence and resulted in a total of 72 problems per S. This procedure is schematized in Table 3. Table 3. Between subject balancing of conditions each day. Sessions Subjects A B C l l 2 3 2 2 3 l 3 3 l 2 CPC blocks (6 problems each) h 2 3 l by experimental sessions. 5 l 2 3 6 3 l 2 Blocks 1-3 include all 18 color-problem combinations (CPCs). Each S_completed one block of six problems each session. These blocks of CPCs were sequenced so that each S received all 18 CPCs in the first three sessions, then re- _ ceived them again in a new order in the next three sessions. The three SS as a group completed all 18 CPCs every experi- mental session so that an analysis by sessions was possible. A similar sequencing procedure was followed on sessions 7-12 using a new basic sequence of CPCs. (See the text for details.) RESULTS Color test. The mean score on the color test was 12.6 correct out of 2h trials (chance = 12 correct). The mean number of correct responses for learners and non-learners was 15.7 and 9.7 respectively. The score of 9.7 correct attained by the non-learners is somewhat lower than chance behavior and probably reflects an error of program- ming. If, on a given trial, S tended to choose the color which was correct on the previous trial, he would be wrong more often than right. Each color was correct six times so that guessing would tend to produce about three correct choices of each color. Some Se scored higher than chance on specific colors. Each of the learners scored 5 or 6 correct on one of the colors, and one §_scored this high on two colors. In the group of non-learners, the highest score on one color was h and this occurred only once. The experimental problems. Figure 5 shows the number of correct responses in each block of problems. The two curves show the per- formance of two subgroups of Ss, three Ss who learned and three 25 who did not learn. The score assigned to each §_for the various analyses was the num- ber of correct responses. This score had to be corrected somewhat in certain instances because an error in sequencing resulted in unequal numbers of problems in the three conditions. Within a block of pro- blems the difference in number of problems amounted to one or two prdblems. The correction procedure estimated each st score had all problems occurred equally. If, for example, there were 7 trials of condition III but only 6 trials of conditions I and II, a percent correct score was computed for the deficient observations then multi- 3O 31 300 learners CORRECT RESPONSES 8 Ci non-learners IOO I 2 3 4 BLOCKS OF PRACTICE Figure 5. The effect of practice on the performance on training trials of learners and non-learners. ’V) 3:. plied by seven. This procedure produced a score for seven trials which was equal, in terms of percent correct, to st actual score based upon fewer trials. The study, as originally designed, allowed analysis by sessions of training by sequencing the problems so that in each session three Ss as a group would complete all 18 CPCs. Since too few SS learned the task, the analysis by sessions had to be replaced by an analysis by blocks of sessions. Each block comprised three sessions and in- cluded all 18 CPCs. A blocks K conditions analysis of variance of the number correct on training trials (1-6) yielded only blocks significant (Figure 6 and Table A). A similar analysis of variance on the number of correct responses on test trials (trial 7) yielded significant effects for blocks only (Figure 7 and Table 5). Table 4. Analysis of variance of the effect of blocks of practice and conditions upon the number of correct responses on training trials. Source of variance SS df FE F Blocks 81+.55 3 28.18 3.73 p <.05 Conditions 2.67 2 1.3h .18 Blocks x conditions 16.11 6 2.68 .35 Error 181.17 2h 7.55 Total 28M.SO 35 33 Table 5. Analysis of variance of the effect of blocks of practice and conditions upon performance on test trials. Source of variance SS df MS F Blocks 10.1w; 3 3.11.8 2.9.0 .10 (p <.2 Conditions 1.13 2 .56 Blocks x conditions 5.66 6 .95 Error 37.94 2b 1.58 Total 55-17 35 Performance on transfer trials was compared with performance on training trials by a block X trials analysis of variance of the number of correct responses. Blocks only were significant (Figure 8 and Table 6). In this analysis the data for trials 1 and 2, 3 and h, and 5 and 6 were combined because the curves before this grouping were very erratic. The comparison of trials 5 and 6 with trials 7 necessitated a correction of the trial 7 data to a base comparable to that of trials 5 and 6 combined. This correction was made by doubling the actual score S achieved over all trials 7. The trials effect may have been statistically non-significant only because trials 1-6 were really not very different from each other so that, in comparison, the lower scores in transfer trials did not affect the variance estimates. This possibility suggested a direct comparison of the difference between performance on trials 5 and 6 with performance on trials 7 by an analysis of variance of these two trials alone. The pooled error term computed over all blocks was used in this analysis. The difference between the two means, however, was 3h IOO‘ I (I) 111 (I) 2:: O (I. (I) 'a'é 90' .— 0 LL! (I: S 0 type I ---------. type II 80 ___._- type III I 2 3 4 BLOCKS OF PRACTICE Figure 6. The effect of practice and conditions on the performance on training trials of learners. 35 20L l5' ' type I CORRECT RESPONSES -----_-_--. type II . _——- type III IO- I 2 3 4 BLOCKS OF PRACTICE Figure 7. The effect of practice and conditions on the performance on test trials of learners. still not significant (Table 7). Table 6. Analysis of variance of the effect of blocks of practice and blocks of trials on the number of correct responses. Source of variance SS df IS F Trial blocks 25.90 3 8.63 .86 Blocks 128.73 3 42.91 u.28 p<<.025 Blocks x trial blocks 38.52 9 n.28 Error 320.67 32 10.02 Total 513.82 M7 Table 7. Analysis of variance of the difference in number of correct responses given on trials 5-6 and trials 7. Source of variance SS df MS F Trial blocks 20.16 1 20.16 2.01 p = .20 Error (pooled estimate, 320.67 32 10.02 same value as in Table 6) 37 —I—l—\_ block 4 IOO ' o 8 o 8 CORRECT RESPONSES (D O a: 7OJ "'2 3.- 4 5-6 7 (transfer) TRIAL BLOCKS Figure 8. The effect of practice and trials on the performance of learners. , DISCUSSION The color-choosing task as a learning paradigm presents two con- ceptual difficulties: how does the §_solve the problem and what is the stimulus? According to a stimulus-response analysis, §_cannot respond differentially in the learning situation unless there are two or more discriminable cue complexes which are differentially correlated with reinforcement. The §_1earns to respond one way to the first stimulus situation (Sl) and another way to 82. In the present paradigm, all visual cues present at the time of choice have been made irrele- vant, that is, any response based only upon the visual cues of color, form, or position cannot result in better than chance reinforcement. The S can solve the problem only by supplying cues himself. The dis- criminative stimulus which guides differential responding in this study must consist of the visual cue complex plus one of two or more cues added by S. This added cue is referred to in subsequent analysis as the ’cue response'. Analysis of the development and elicitation of this cue response is necessary for an understanding of the results of this study. Figure 5 shows the number of correct responses in each block for the six Ss who finished all experimental problems. Three gs learned the task, and three did not. Both learners and non-learners improved with training. The improvement of the learners' scores reflects in- creasing facility at solving the problems while the improvement of the non-learners' scores is an artifact of the design and does not reflect the development of apprOpriate responses. The performance of the non- 1earners starts well below chance in block one and rises steadily to chance performance in block four. A properly designed two-choice 38 39 discrimination problem has the discriminanda sequenced in such a way that only one pattern of responding will result in scores which are significantly different from 50% correct (guessing on two items). The unusually low scores of the non-learners in block one suggests that certain consistent patterns of response resulted in very few rewarded trials. This idea was tested by applying a consistent pattern of response (a strategy) to one of the problem sequences actually given to one S. The strategy which was tested was a 'win, stay-lose, shift' strategy based upon color cues and went as follows. On the first trial of each problem one of the color cues was chosen randomly. If it was the correct cue, it was chosen again on the next trial, and if it was the wrong cue, the other color was selected on the next trial. Consistent application of this strategy produced a score of only 25-30% correct. This test was taken as good evidence that the non-learners actually employed this strategy in the early blocks. If it is true that non-learners consistently used this strategy early in training, then it is highly unlikely that the form cues had the effect on train- ing that was originally hypothesized. Since Ss responded very con- sistently to color cues they could not have confused color and form cues. It seems clear that they had no difficulty finding the relevant dimension from the start of training. Figure 8 shows that the learners improved very little within problems. This lack of improvement with repeated trials of the same problem probably results from there having been a limited number of trials in each problem. Each six-trial problem was essentially com- posed of a random combination of trials from two different problems. For example, in a problem which taught red and green, red was correct 1+0 three times and green three times. It is not surprising that §s found it difficult to solve two concurrent conditional discriminations whose trials were randomly interspersed. Figure 8 shows that Ss tended to make f wer correct responses on transfer trials than on training trials although the effect was not statistically significant. This decrement on transfer trials may be attributed to the external inhibition of the cue response by the sudden change in the stimulus situation. The graph suggests that at the end of training the inhibitory effect is still strong. The difference between performance on trials 5-6 and trials 7 is rather constant in magnitude throughout the study with just one exception in block three. The reason for this exception is not clear. All SS in all blocks ex- cept the third had equal or lower performance on transfer trials com- pared with trials 5 and 6. In block three two Ss improved on transfer trials. There was little evidence that the scores of the learners were affected by conditions. The factor of conditions never approached significance either for training trials or for transfer trials. This is not surprising in view of the non-learners' data which indicated that even non-learners responded consistently to color cues. The learners also showed high levels of color and color-name association before training began as revealed by the high scores on the color test. These Ss already had cue responses appropriate to this task when they started training, and any effect on scores due to form manipulations was not strong enough to be detected. There is some evidence that on transfer trials, conditions inter- acted with blocks of problems. The data suggesting this conclusion Ill (Figure 7) are not strong. Most variations in the curves represent group differences of only one or two correct responses. It will be assumed, however, for the purpose of discussion that there was an interaction such that early in training transfer was best in problem types II and III. Transfer in type I problems improved steadily until, in the final block, type I problems produced the highest level of transfer observed in the entire study. This interaction of conditions with blocks will be tentatively explained in terms of the disruptive effects of external inhibition of the cue response. The explanation will make use of an analogy between certain aspects of the learning and transfer tasks and one's reaction to the sudden noise of a door slamming shut. If you have not been warned of the coming slam, the sudden noise has a strong disturbing effect upon on-going behavior (external inhibition). If you are warned, however, there may be no disturbance of behavior at all. Similarly, in this study, if the decrement on transfer trials results from the sudden change in the stimulus situation, any cues which may serve to 'warn' §_of the coming change in stimulus conditions should minimize the external inhibition and improve transfer performance. Such a warning cue is available in type I problems in the form of a pair of identical form cues on each trial preceding the test trial. Form cues in problem types II and III change continually and as a result can gain only minimal warning value compared with the unchanging form cues in type I problems. The low level of transfer in type I problems in block one can be explained similarly. Since changes occur on every trial of problem types II and III, the transfer trial is more of a change in type I problems than it is in type II and III problems. Early in training the form cues in M2 type I problems have not gained any warning effect so that without the warning, external inhibition of the cue response would be quite strong. If this analysis is accurate, we should expect that retarded Ss would not show as strong an interaction of conditions with blocks. This conclusion follows because, according to the explanation above, hig levels of transfer depend upon develOpment of an internal inhibition of strong external inhibition. According to Denny's review of learn- ing in retarded gs (196A), retardates are consistently deficient in the ability to inhibit responses. C a ‘ . r- _o _; Suggestions for Training Since §_must make a cue response to solve this type of problem, the first step in training Se to choose colors should be the develOp- ment of a reliable cue response. A very appropriate training sequence might begin by teaching the §_to repeat the color words spoken by the instructor before any attempt is made to introduce the choice response. Speaking the color name is not the only form the cue response may take, however. The effectiveness of word training should be compared with the effectiveness of training in color matching. The color matching situation is not unlike the color choosing situation in that color matching also requires that §_produce a cue response closely related to the colors. The color matching task is analogous to the conditional discrimination used in this study. In the color matching task, §_is presented with a color sample, a procedure wlich exactly parallels the use of the spoken word in this study. The color sample presented to S is a conditional cue which assumes control of S's subsequent choice of color samples. A cue response is implied in this situation for the same reason it was implied by the paradigm used in this study. There A3 is little reason to predict the superiority of either color matching or color name learning as pretraining methods. The data of this study strongly suggest that norms gs learn color concepts without difficulties due to the form cues of the discrimi- nan-a. The data also suggest, however, that early in training form variability may facilitate transfer (transfer performance in block one was higher in problem types II and III). Training programs might be improved by manipulating form cues early in training but later, as S begins to respond asprOpriately, training and transfer are likely to be more convenient and at least as effective if form cues do not vary every trial. These training swagestions may be restated in the language of attention theo y. Uien form cues are varied regularly and are not available for consistent differential choice as in problem types II and III, §fs attention will be directed to the color dimension. Such direction of attention is most important early in training; when um begins to respond correctly, we can be sure he is attending to the color dimension and the form manipulations become unnecessary. Severe- ly retarded SS are particularly likely to benefit from conditions which direct their attention to the relevant dimension. In the present study, the color word itself directed the attention of the normal children. Much stronger conditions probably would be needed for severely retarded children. LIST OF REVERENCES Bijou, S. N., Birnbrauer, J. S., Kidder, J. D., and Tague, C. Programmed instruction as an approach to teaching of reading, writing, and arithmetic to retarded children. The Psychological Record, 1966, 16, 505-522. Bruner, J. S. The Process of Education. New York: Vintage Books, 1960. Denny, M. R. Research in learning and performance. In Stevens, H. A. and Heber, R (Eds.), Mental retardation: A review of research. Chicago and London: The University of Chicago Press, 1963. Denny, M. R. A theoretical analysis and its application to training the mentally retarded. In Ellis, N. R. (Ed.), International Review of Research in Mental Retardation, vol. 2. New York and London: Academic Press, 1966. Duncan, J. The Education of the Normal Child. New York: Ronald, 19h3. Evans, T. E. The effect of a modeling technique on the learning of four tasks by retardates. Unpublished M.A. thesis, Michigan State University, 1968. Gellerman, L. N. Chance orders of alternating stimuli in visual dis- crimination experiments. Journal of Genetic Psychology, 1933, E2, 207-208. Haas, D. Especially from Thomas. Racine, Wisconsin: Whitman Publish- ing Company, 1965. Harlow, H. F. The formation of learning sets. Psychological Review, 19119, _5_6_, 51-65. Harlow, H. F. Learning set and error factor theory. In Koch, S. (Ed.), Psychology: A study of a science, vol. 2. New York, Toronto, and London: McGraw-Hill Book Company, Inc., 1959. House, B. J. and Seaman, D. Miniature experiments in the discrimina- tion learning of retardates. In Lipsitt, L. P., and Spiker, C. C. (Eds.), Advances in Child Development and Behavior, vol. I. New York: Academic Press, 1963. it )4. 5 Education of the Slow-learning Child. New York: Ingram, C. P. Ronald, 1935 (erd ed. l960). The Wild Boy of Aveyron, translated by George and Kuriel Hew York: Appleton-Century-Crofts, 1962. ('1 y l O O O (V Sensory stimulation versus speCIflc responses. Humphrey. , 2-1%. Kolstoe, 0. P. Exceptional Children, 1956, 23,1 fi*lsell Book of Colors, Munsell Color Co. Maryland: iunsell. Baltimore, 1929. perception and learning Sidman, L., and Stoddard, L. T. Programming In Ellis, N. R. (Ed.), International New York and for retarded children. Review of Des arch in Rental Retarlation, vol. 2. London: Academic Press, 1936: {EB '"'infitflu'iifljfliufufl{flinjfliiiiflflfiflmi'"