"ERR'O‘RLESS" DISCRIMINATION LEARNING AND TRANSFER IN THE MENTALLY RETARDED THESIS FOR THE DEGREE. OF M. A. MICHIGAN STATE UNIVERSITY W LAWRENCE JOSEPH KRIP’S . a — ‘ - u n . . ‘ - - . O A o * . A . . - - . . . . A. c ‘ I - .' . o . V l . ‘ _ 4 _ V . - - o ‘ r ‘ A. . I -' .- ¢ - — ' I ‘ I '- ' - . . . , . ‘ .I . . . . . - _ _ . . ' _-' V ‘ . . . - . . ' . . . . . . . _ _ I I - ' ‘ ..... . ,. . ‘ ‘ . ‘ ' ' ~ - .A , ' . .- . , _ . . v . o . . ~ - u . _ . . - - . . . - s . , - - - . .. U. ' - . - - r . , . , - .-. - ‘ - " ‘ " ' 4 o i ' , r . . v‘ . . _ v . I . l .. . _ ‘ _ . . . ,. . . . V ‘- ~ - - . . . v .. . v .- o . ’ . ' . . .' o . I LIBRARY "' I Michigan State University ABSTRACT 'ERRORLESS' DISCRIMINATION LEARNING.AND TRANSFER IN THE MENTALLY RETARDED By Lawrence Joseph Krips There are many theories postulating the learning proc- ess involved in acquired a discrimination. Acquisition usually included 5- respondhg, a method which does not al- low that stimulus to retain its neutral value, a factor which may be detrimental to the total discrimination learn- ing process. Several studies have attempted not permitting this neg- ative cue value to become assigned to 8-. This type of learning is usually achieved by physically not allowing an 8- response to occur; by not presenting 8- during acquisi- tion of 5+ responding; or by using an early introduction of 3- in the presence of 3+ with the assistance of fading. The present study is concerned with a modification of Terrace's errorless approach applied to a visual two dis- criminanda problem and transfer with twenty-nine moderately retarded children. The fading variable was a gradual in- crease or decrease along the brightness dimension. Data was collected on the number of 8- responses emitted and the latency difference between 5+ and 8- responses. The eXperiment had three sections. Original learning was compesed of two groups learning a discrimination of a projected sketch of a man (8+) and a projected sketch of a dog (5-). The control group learned with each stimulus at equal intensity for 100 trials. The experimental group faded in S- in four incremental phases during 100 trials. In transfer learning, a word replaced its respective sketch. For this task the experimental group evenly and randomly divided into those who would be presented a typical discri- mination situation and those in which 3 would be decrement- ally faded out and then 3' incrementally faded in. The posttest was simply a’25 trial (full illumination) presenta- tion of Ol.material to the experimental group. As expected, more gs learned 0L material by fading than by a typical presentation. The fading, however, showed evi- dence of producing more errors than the non-fading method. bDespite this unexpected finding, transfer by fading also produced less failures than the typical task. The experi- ment was further confounded by transfer fading g; acquiring the original discrimination faster than non-fading transfer is. The posttest strongly indicated no response disruption of 0L if the transfer material was acquired by a non-fading approach. No significant differences were found for either experimental group during this recall section. The variance between this study's findings and Ter- race's errorless learning approach were analytically dis- cussed in terms of both other studies and of inherent factors of the present experiment. The relation between this invest- igation and programmed instruction material was also discussed. ”ERRORLESS' DISCRIHINATION LEARNING AND TRANSFER IN THE MENTALLY RETARDED By Lawrence Jeseph trips A THESIS Submitted to Michigan State University in partial fulfillment ef the requirements fer the degree ef MASTER OF ARTS Department ef Psychelegy 1971 .I ""1 "x \rV’ .‘IrJ ACKNOWLEDGMENTS The author would like to express his appreciation to Dr. H. Bay Denny and Dr. Billing who served as thesis committee members. Special appreciation is extended to Dr. Lester Hyman, the chairman of this committee whose guidance and critique were invaluable. 11 TABLE OF CONTENTS List of Tables ......................................... List of Figures ........................................ Introduction ........................................... Method ................................................. Subjects .......................................... Experimental Design ............................... Procedure ......................................... Pretraining .................................. Discrimination training ...................... Results ................................................ Discussion and Conclusion .............................. List Of References eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee Appendix eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee Apparatus eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee Stimuli eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeoeeeeeeeee iii iv 11 11 11 1h 15 16 19 35 1.5 117 #7 #9 LIST OF TABLES Table l-Experimental Design ............................ Table 2-Per Cent Failure ............................... Table 3-Mean Number Incorrect Responses ................ Table b-Analysis of Covariance of iIS+)-§IS-) Latencies. Table 5-Distribution of Measures of Intelligence Among GPOUPSeeeeeeeeeeeeeeeeeeeeeeeeeeeoeooeeeee iv 12 20 28 29 bl Figure Figure Figure Figure Figure Figure Figure Figure Figure LIST OF FIGURES l-iIS+)-EIS-) Control Group Latencies during 0L . Z-Mean number of incorrect responses for the Control Group during 0L ....................... 3-iIS+)-iIS-) Experimental Groups Latencies during 0L ..................................... h-Mean number of incorrect responses for the Experimental Groups during 0L ................. 5-i}$+)-§IS-) Experimental Groups Latencies during TL ..................................... 6-Mean number of incorrect responses for the Experimental Groups during TL ................. 7-§IS+)-iIS-) Experimental Groups Latencies during the Posttest .................................. B-Nisconsin General Testing Apparatus ........... 9-Th9 StIMU11 eoeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee 2h 26 27 31 32 33 #8 50 0.00.0. p O C ’3 O I I w I O C c It P 'Errorloss' Discrimination Learning and Transfer in the Mentally Retarded Lawrence Joseph trips Hichigan State University In discrimination problems, S's response to the S- is often considered a necessary part of the task. Through nenroward, the response becomes (at least in part) extin- guished. This process, however, also neccessitatos the assigning of cue value(s) to the S- by g; S- is not neutral. It has been postulated (Holland, 1960; Bull, 1939; Bull, 1959; Terrace, 1963), however, that this value placed on the S- by §_may be detrimental to overall discrimination learning. The effect of wrong responses has been postulated to be deleterious to subsequent behavior. In this investi- gation, an errorless (i.o., little or no 8- responses) procedure was utilised with RR.children for a discrimina- tion task and a transfer task. Discrimination learning is most generally defined as learning not to genoralise. A response is made to one stimulus and not to another stimulus even though those two stimuli may have some properties in common. That is, g learns not to generalise the cues of one stimulus to that of another, and therefore, avoids indiscriminate (chance) responding to both stimuli. The response made to 8+ is correlated with reinforcement, and the response to S- is correlated with nonroinforcement. so is then said to have acquired functional control over the response associated l with it. The answer to the question of how the organism does, in fact, learn a simple discrimination problem has been at- tempted by several theories. Singlo unit S-R theories postulate a relationship between the visual stimulus and an overt response. Two-stage theories also assume that an approach response is made to one of the discriminative cues, and further assumes that §_first attends to the relevant stimulus dimensions before the occurrence of instrumental learning. That is, an orienting response must be made before any learning can occur. One theory that breaks down the process still further is elicitation theory (Denny and Adelman, 19553 Denny, 1966). According to this theory, an array elicits covert dimensional responses which produce feedback cues to which overt responses are conditioned. A consistent contiguous relationship of reward (or nonroward) with a stimulus elic- its the response. Backchaining is said to occur from the goal cues (84) back to cues in the starting region. The operant response is associated with kinesthetic cues which in turn act as an 84 for further responses in a chain. A teaching device based on the above theory has been designed by Donny. The Multiple Differential Response and Feedback Apparatus (RUDRAFA) is basically an errorless procedure tool. .Although an incorrect response may be attempted by g, the apparatus, by the use of barriers, does not allow completion of the response thereby eliminating 3 backachaining of unroinforced stimuli and eventual learning of a response to an incorrect cue. Crutch cues and immedi- ate knowledge of results are consistent with elicitation theory postulates as loading to rapid learning. .Another theory concerning acquisition of a simple discrimination problem was proposed by Bull (1939. 1953, 1950, 1952) and Spence (1936, 1937, 1937a). Their theories of discrimination learning are based on the following five postulates (Hall, 1966): 1. Ivory reinforced trial loads to an increment in oxcitatory strength for a given stimulus and its reinforced response. 2. Ivory nonroinforced trial results in an inhibitory increment to a given stimulus and its nonreinforcod response. 3. Both oxcitatory and inhibitory tendencies genoralise to stimuli along a stimulus continuum. b. There is the algebraic summation of excita- tory and inhibitory increments which result in ... 5.1A discriminatory response based upon these algebraic summations. According to these postulates then, if there are no generalised inhibitory tendencies, there should be better performance to Be than if inhibition were present. In other words, conditioning not involving discrimination learning (i.o., differential conditioning) will result in superior response operation to the desired stimulus. Gynther (1957) gave support to this prediction by differentially training one group of g, and non-differentially training another. The group to which the negative stimulus was never presented (non-differentially trained) during acquisition emitted more conditioned responses to So during test trials (both stimuli presented together) than the other group. 1. Another prediction that may be made from the Hull- Spence theory concerns the similarity of the form of the discriminanda used. The greater the dissimilarity of the stimuli, the more readily the discrimination should be learned. That is, the generalised inhibitory strength to so and the generalised oxcitatory strength to 8- should become increasingly weaker as so and 8- become more dissimilar. Hanson (1959), using four groups of pigeons each with dif- ferent and increasingly dissimilar stimuli (light hues) found that the amount of training to reach the discrimina- tion criterion varied directly with the quantitative dif- ference between 84 and 3-. That is, as the difference grew smaller, the required amount of training grow larger. Terrace (1963, 1963a, 1966) seems to have utilised both the above predictions in his work on discrimination learning without errors. The Hull-Spence theory assumes that dis- crimination learning can take place only when a response is made to 8- so that this response may be nenreinforced and, therefore, gain inhibitory strength. Terrace, however, has provided evidence that (at least with pigeons) no response has to be made to S- in order for a discrimination to be learned. This idea would be in agreement with Denny's elic- itation theory in that prevention of incorrect responding also prevents the formation of an S-R complex. Terrace's concept is an extension of (but not the same as) Gynther's differentially-non-differentially trained So. In Gynther's investigation, one group was conditioned to respond to 8+ 5 during the absence of 5-. When presented with both 8+ and 8-, they responded more to 8+ than the group who had been given a simultaneous discrimination problem during acquisi- tion. This result provides evidence for discrimination learning without errors (i.o., without responses to 8-) since indeed making a response to S- was impossible during acquisition. Since experimental evidence has shown that better re- sponding occurs when acquisition is by an errorless proce- dure than by a typical simultaneous method, and since the Hull-Spence theory predicts that as the form difference between stimuli in a discrimination problem becomes smaller, the amount of training time grows larger, it follows that if one starts with greatly dissimilar stimuli and progressively decreases the difference(s), a discrimination task may be learned in this manner. This concept is similar to that of fading as used in programmed instruction (PI) (Holland, 1960; Skinner, 1958; ansdaine, 1969; Silberman, 1962). Terrace utilised both the errorless learning idea and the fading technique to teach pigeons a red-green discrimination in the presence of both 8+ and 8-. When Terrace speaks of errorless learning or learning without errors he is not necessarily referring to 1am; correct responding. The terms refer to approaching perfect responding, that is, making fewer responses to 8- using this technique than would be made in a simple discrimination problem. It is possible, however, that responding within Terraco's framework would 6 allow more errors than an elicitation approach since the latter can provide effectively for no errors. Terrace divided his gs into four groups. In the con- stant group the brightness and duration of 8+ and 8- was the same both initially and at the end of training. The pre- gressive group, however, was initially presented an 8+ and 8- of different brightness and duration and, in three incre- mental phases, was breught up to the brightness and duration of the stimuli given to the constant group. These two groups were also each divided into early and late groups to which 8- was introduced either early, during the first cen- ditioning session, or after a number of weeks of training had already occurred in the presence of 8+. All four groups were presented the same 8+ (red) and 8- (green) of different wavelengths. The red-green discrimination was successfully acquired without the occurrence of any errors in twelve out of twelve cases in the early progressive group. Although the other groups did not fare as well, it was found that early presentation was better than late, and progressive better than constant. Terrace utilised a similar procedure for a transfer task. A vertical (5+) and horisontal (8-) line discrimination were superimposed on red and green back- grounds respectively. The red and green were then faded out. If fading was not used, or if an abrupt transfer was attempted, errors occurred. Even if the red-green discrim- ination were learned without errors, the performance was permanently impaired in terms of response latency if the 7 transfer was learned with errors. An application of errorless discrimination learning with human‘gs was made by floors and Goldiamend (1965). The stimuli used were inverted isosceles skeleton triangles. The g; were six male and female preschool children ranging in age from three to six years. The task was a delayed, three-choice matching to sample problem. The matches were rotations of the sample except for 8+ which was the same as the sample. ‘Ihen the sample was withdrawn and only the 8+ match illuminated, correct responding was easily learned. Fading-in the brightness of the negative stimuli was done in seventeen incremental steps. Varying the time of introduc- tion of the fading technique provided evidence for Terraco's concept of errorless learning. Although the task was a difficult one for the preschoolers, the problem was learned with fading and best learned (at least from the aspect of economy of time) by early progressive fading. It is inter- esting to note that even if an g'was originally presented the matches at full illumination (with its resultant chance responding), the introduction of fading immediately provided correct responding, an excellent example of stimulus control. Torraco's work then, has shown that earlier introduc- tion of 8- in the presence of 8+ with the added aid of a fading technique (the progressive group) provides greater stimulus control in a visual discrimination problem than: 1) a fading, late introduction of 8-; 2) a constant, early introduction of 8- (a typical discrimination problem); or 8 3) constant, late introduction of 8- (nondifferentially, then differentially trained as in Gynther's experiment). The same results occurred in a transfer situation based on the original stimuli. In the present investigation, the fading technique of errorless learning was applied to a visual discrimination problem using moderately retarded children. The use of an early progressive procedure might be one way of compensating for the Mflfis often hypothesised inhibition deficit and/or attentional difficulties in discrimination problems. This study is then, a modification of Torraco's work. For our purposes, fading is defined as a gradual increase or de- crease along a given dimension of a stimulus. The indepen- dent variables are brightness and time of introduction of 8-. The latter is not the same as the early and late introduc- tion of 8- as defined by Terrace. Terrace assumed that the presentation of 8- in accompaniment with 8+ was an effective stimulus even though its illumination was far lower than that of 8+. We do not make this assumption since we do not in fact 'know' when the 8- is attended to by g, we do assume, however, that at some time during the brightness increment of S-, this stimulus does become an effective one. It is further assumed that the attentional effectiveness of 8- is not the same for all as in the fading group. We are not attempting to determine the moment of such effective- ness. This moment is assumed to randomise out over all 8s. The differentiation between brightness and time of ’ 9 introduction of 8- is made not for the sake of measurement of these factors as independent variables but simply to point out that fading in this experiment actually has two components only one of which is quantifiable. The dependent variables are the number of responses omitted to 8-, and the latency difference of responses to 8+ and 8- both during and after the acquisition of the dis- crimination. In light of the above discussion, the following hypo- theses are set forth; 1. The use of a fading technique during the acquisition of a discrimination task will result in better learning of the discrimination than if this technique is not used. That is, a. The fading group will emit less responses to 8- than a non-fading group; and b. the difference between 8+ and 8- latencies of response will be greater.for the fading group than the non-fading group. Both a and b will occur during the acquisition and during testing of the discrimination. 2. Even though a visual discrimination problem is acquired with a fading technique, a transfer problem of the original task not based on fading transfer will result in poorer learning than if transfer is attempted with fading. That is; a. there will be a lesser number of responses to the new negative stimulus (S'-) by the transfer 10 fading group than by the abrupt transfer group; and b. the (8'+)-(8'-) latency difference will be less for the non-fading group both during acquisition and testing than those of the transfer fading group. 3. If a discrimination problem is learned with the aid of a fading technique, and if a transfer problem is also learned with fading, then on a review posttest of 0L, not only will less responses be made to 8-, but also the latency difference of (8+)-(8-) responses will be greater than if transfer were abrupt even though 0L was based on fading. Method Subnects. The g; were 31 moderately retarded children from the Hope School in Jackson, Michigan. The children lived at home and were bussed to and from school as are children from any normal school. The criterion for participation in this experiment was a child's inability to read even though he may have had some knowledge of letters. This prerequisite was measured by the teachers at the school. From this population, gs were randomly assigned to the different groups as outlined in the experimental design below. Only two children were discarded from the experiment. One of these was non-cooperative and the other became ill. Only 29 gs were used, therefore, to test the hypotheses. .Apparatus. Please see the appendix for a complete descrip- tion of the apparatus and the stimuli. Experimental Design. There were three main parts to the experiment (see Table 1). The first two sections consisted of fading and non-fading groups for the given tasks of original learning (0L) and transfer learning (TL). Only the fading group (Fo) of OL were used for the transfer task. These|§s were randomly divided into non-fading (FoNFt) or fading (FoFt) transfer. .These two groups were then given a posttest on OL. We see, therefore, that NFo was a control group utilised only for the original discrimination. The Fe g; were the experimental groups for OL. Fe So were also the only ones used for the remainder of the experiment. During TL, FoNFt was contrasted with FoFt. Both were then 11 12 Andean» “NV +msln Ana-z. ea eeeeeeem one Lee—nee eneeee. .1.» Ana 1mm e-eeee. eu.mm Lee use eaeaeev my.» Ann use eneaee. mu.me\n Ans nan eaeaee. my.» . Ana anu eaeeeev ms\nu.u can Ame 1m_ enedee. ewe.» Lee—use eaeaee. .e.» A__ 1. edeaeev m u.m , Ion 1s eueeee. on.» .eezv eeee Anna. anae- uuduueed nehouene Ace—Ins eaeduav +mslm .3 ocean Ana men eneeaev ens\nn1n .n eeena .en_1m~ eneaaev en nun .u eeena .enm aseee use. Ann 1— uncanny + aim ._ ocean Aeevam ..e.e. eeegenouen enxun .ee_1. eneeeev eeeeenoeun ensue Aesez. ee nee-2. ea: undsneoq.us=«u«no nudeon Heeueldnenufllu enmsh 13 compared on original learning disruption after the transfer task by means of a posttest. During OL, the only difference between the two groups was that the 8- (a picture of a dog) initially presented to Fe was not of the same intensity as the picture of the man (8+). Group Fe, however, was “brought upI to the luminosity of the other stimulus in three incremental steps. The transfer task involved the transfer of the original discrimination to that of the words (8') the line drawings symbolisod. FoNFt was simply presented the word MAN (8'+) and the word DOG (8'-) at maximum intensity, superimposed over the appropriate and equally intense pictures. The drawings were then abruptly removed leaving only the words. Fort, on the other hand, transferred by fading-in the word MAI over the picture of the man, and the word DOG over the picture of the dog. The words were initially superimposed on the proper picture at maximum intensity (G13) and the. word at partial (G3) intensity. The brightness level was increased in three phases (i.o., from GB to G7 to G10 to G13) so that at the end of the incremental fading, both the words and pictures were at full illumination. The pictures were then faded out by using the same incremental steps as the fading-in, but reversed. At the end of this series, therefore, the words remained at full intensity, and the pictures did not appear. The task at this point was the discrimination between the words MAN and DOG. It is important to note that in the above design, all nl. 1h fading steps (both incremental and decremental) for all groups have the same psychophysical intensity differences. Also, all the groups within a given learning task are pre- sented with the same stimuli by trial 76. NFo have the same stimuli initially as Fe have by trial 76 during OL. FoNFt have the same stimuli during the first 59 trials as does FoFt for trials 33-b3. The last 50 trials for FoNFt consist of the same stimuli as FoFt for trials 76-100. During the posttest, both groups (i.o., FoNFt and Felt) were again presented with the original picture discrimi- nation problem. Both 8+ and 8- were of equal intensity. According to this design, it was expected that group Fo would learn better (in terms of the dependent variables) than NFo during OL. For the transfer task, FoNFt was ex- pected not to perform as well as FoFt. It was further hypothesised that the transfer task would be more disruptive on a posttest of OL for FeNFt than for FoFt. That is, although all So in these two groups learned the original task "without errors,‘ those that learned the transfer with errors (FeNFt) would not only show poorer performance on the transfer task than the other transfer group (Felt), but would also on a review posttest show poorer discrimination and make more errors on the original stimuli than both themselves (i.o., FoNFt during 0L) and FoFt during OL and during the posttest. Procedure. After a listing of all possible as for the experiment was compiled, students were randomly assigned to 15 groups. Group Fo was to have approximately 2N and group NFo-IN, since the former were to be evenly and randomly divided for the transfer task. As their turn came, each §.was taken from the classroom by §.and accompanied outside to a trailer in which the apparatus was housed. An attempt was made by g'to be as friendly and unimposing a figure as possible. Conversation, a friendly smile, and a helpful hand were accorded to all 8,. Pretraining. Once inside the trailer, as were shown the room in which they were to be seated. It was explained that they were going to be playing a candy game. The apparatus was already projecting the pretraining stimuli, and the So were seated in front of and facing the discriminanda board. The door was closed with £_standing behind the seated‘g. The following instructions were given: 'I will show you how to play the game now. In front of you are two pictures, and you have to guess which one is correct. You make a guess simply by pressing the window with the correct picture with your finger. Go ahead and make a guess.” If §,understood the instructions, he pressed one of the windows and was told that he was playing the game correctly and that he made either a correct or incorrect response. In this pretraining series only verbal reinforcement (e.g., good, that's correct, you are doing well, etc.) or non-reinforcement (e.g., no, that is wrong, etc.) was given. If §_did not understand the instructions, seemed bewildered and did not make a response, 16 the instructions were repeated. If a response was still not made, E'toek §fs finger and pressed the appropriate window. Verbal reinforcement was given as if §.had responded cor- rectly by himself. This process continued until g did respond without prompting. All gs went through the pro- training series until it was obvious to §.that §_understoed the concept involved (five correct responses in succession). Discrimination training. Upon completion of the pretraining procedure, the door was opened by E, and §_was told, ”Now that you know how to play the gums, you can play by your- self. Evory time you press the correct picture you will get an MAM in the tray in front of you (§,peintod to the tray). Here is a bag for you (a small paper bag was opened and placed on the shelf below the reward tray) to put your candy in if you do not want to eat them now. We will begin playing in about one minute. I will tell you when to start, and you can play the game Just like before.” The door was then closed, and a new slide tray inserted into the projec- tor with tho appropriate original learning stimuli. ‘2 seated himself in front of the control panel, making sure that the timer was at sore. The proper response record sheet showing the position of the stimuli to be presented was placed in front of §_to be marked appropriately as‘g responded. ‘8 was then told, "You can start playing new. Press a window.” If at any time during the course of the experiment, an grdid not make a response within ten seconds of the beginning of a trial, he was reminded to press a 17 window. This command was repeated every ten seconds until a response was made. The first few slides presented to all So was not part of the training sequence. They were simply the words MAN and DOG presented randomly on the scroens1and responded to without reward, to determine the presence of stimulus pre- ference. If such selectivity was shown (i.o., not chance responding), §,was discarded from the experiment. A correct response was followed by delivery of an MAM and only intermittent verbal reinforcement (approximately 1/5 variable ratio schedule) for all as during the first 75 trials of any given session. For any trial in which an incorrect response occurred, only the presentation of the next slide was provided, no reward or vorbalisation being given. During criterion trials (76-100) of OL and TL and during the posttest (25 trials) only candy was given for correct responses. The amount of time spent with each‘g in this experiment was relatively small. A set of one hundred trials was run in one session. Group NFo gs, therefore, were started and finished in one day. To 8s, however, took at least three days to run. Criterion performance was 20 8+ responses out of the last 25 trials. If this level was not achieved by any given §_during either-0L or TL, then either one or two 1 All stimuli were randomly assigned to one of the two screens in all groups. The assignments were such, however, that the same stimulus never appeared in the same screen for more than three successive trials. I. 18 repetitions of one hundred trials were given to that‘g on the material failed. One set was presented on day one following the failure and the other set on day two after the failure, if criterion was not met during the first repeti- tion. If by this procedure, an §.could not meet criterion on the original discrimination, he was not allowed to con- tinue and was deleted fron the experiment. If criterion by the above procedure was not met during transfer learning, however, g’was allowed to go on to the posttest. Using this technique, the maximum time for each g in group Fe was seven days. No more than one hundred trials were given on any one day except for the first day in which a pretraining series was also administered as described above. An attempt was made to allow no days to intervene between sessions. Meek- ends were the exception, of course, since there was no school on those days. All NFo gs were run before group Fo 8s. The latter woee put through the experiment as they were randomly picked. That is, some gs were finished before others had begun, and different 89 could be on different parts of the experiment during any given day. ll'lllll'llalllII Results Evidence in support of hypotheses one and two are provided by a comparison of the number of 8s learned versus go not learned between groups during OL and TL (see Table 2). On the original discrimination, 36.e$ of the fading group So failed versus 50% of the non-fading So not learning the discrimination. During the transfer task, all So in the fading group (FoFt) learned the discrimination within the first one hundred trials whereas only four out of six FoNFt ‘gs learned within that period. On the second presentation of one hundred trials, only the remaining two FoNFt g; were run, one of which learned, the other did not learn in three hundred trials. At best, then, the FoNFt group had 16.35 failures while FoFt had no failures. Considering only the first one hundred trials on TL increases the rate of failure for FoNFt to 33.3$.while FoFt remains with no failures. According to the above gross analysis, therefore, we have strong support for hypotheses one and two. A more detailed analysis, however, places certain qualifications upon this conclusion. All groups were analysed in terms of 8+ and 8- laten- cies. The mean latencies for block of five trials were determined. The latency difference between the two types of responses was then found (i.o., 3(8+)-Y(8-)). Thus one statistic represents two varying measures. When a discrim- ination has been acquired, a characteristic range of this. statistic was found. For example, if no 8- responses are 19 20 «e 3.2 eels-"---" Ax“ v .— Ams .me. e .xu .en. n need Ax". _ mmm «on .s_. _ Aesvlh H me an.nn 111111111: . Ax_ .eev u Ax: .euv e Ian .eu. m as: .mu. eeee undcusoa nouosona .ennndsh evocnooa no: evocnsed Hooch .onauqem .uoceson no: .uonnsoa ceuesecoooam .oasnusm .uonasod 9oz .uonnsod coduevnooonm .ousndoh «denuded «oz .Uonusea nodesosoeonm Ann .en. eezee «8.: men .933: Ann .em. 5 Ana .en. n .eeeeeea eea Asa .an. «— Axe .en. n .eeneeea :3 .3: ea :3 .3. eh: unneeded Hecauuho eesaeee eeeo eeanu eases ll‘ll'llvrtl .el'I 21 made i(8+)-f(8-) is simply a measure of i(8+) response latencies. Since correct responding is our definition of a learned discrimination, we may assume that this level is definitive of 3(8+)-i(8-) learning. Even if 8- responses have been made, the difference between the latencies of 8+ and 8- can still be in the direction and magnitude of just 8+ latency measures. No matter what this level is, it is probable that no difference between the latencies (i.o., 2(8+)-§(8-)-0) or a difference in the direction of longer 8- latencies (i.o., -x seconds) represents no learning or learning of the incorrect stimulus. Allowing for an occassional error, an analysis of I(8+)-i(8-) latencies associated with either no or one 8- rospense per five trials was made. The mean latency differ- ence was found to be +13.2 seconds with Ob2.6 seconds. Three standard deviations either side of this mean would include most trials associated with ”errorless“ responding. The range is, therefore, 5.h seconds to 21.0 seconds. That is, responding consistently in this range is representative of learned responding. Corroborating evidence is found in the corresponding 8- frequency distribution. We may say that for our population, i(8+)-i(8-) latencies above the +5.5 second level is an operational definition of dis- crimination. The first analysis we can make is within the control group. Five of the‘gs learned and five did not learn. An examination of the latency differences (see figure 1) shows 22 AO.M:wu:p momoco.cq Q5080 mouaaoO AImVNTn+vaIP osamah m~dwn9 00F no 00 mm on on ox. no ocbm Oahu 0». on. on ma ON m. 0- m J a e\. ‘- pocudoq no.2 .3" -u pocucoq QOFI. 06.1 HQO oh.T "Qo94. Mohp.+ 0.3 + an+. (spuoeaS) sagouezeq 23 the difference between the two groups. The learned group consistently had higher latency differences (above +5.“ seconds) than the not learned g; (below 5.5 seconds). Failing gs oscillate about the soro difference between 8+ and 8- latency. They did not discriminate the difference between the two stimuli. The 8- frequency curves (see figure 2) corroborate this conclusion. It is interesting to note that, roughly speaking, the higher the latency curves the lower the 8- frequency curves and vice versa. For example, during acquisition, the learned control group made no errors during trials 26-30 and hl-SO. A glance at the latency chart would predict that such would be the case. The relationships within and between these two graphs may be used as a template for analysis of the rest of the groups since we have here an absolute visual difference between go that learn and go that do not learn a discrimination under typical discrimination circumstances. In comparing the control group and the experimental group on OL we find that of those g; in both groups that learned the discrimination, acquisition lasted only until about trial 30 (for the control group) and trial 35 (for the experimental group). Contrary to expectation, however, fading acquisition was not "errorless” and, in fact, more errors were produced by the fading group throughout the 100 trials than by the control group. The upward acquisition slope of the latency graph and downward slope of the 8- frequency chart are more characteristic of acquisition during 24 MS a: Q a . on: no.“ momGOQmMQ somaanGMOGw wmmceuflmnuncoméum oasuuh .1 . . . mgcwah on: no om mm ow mm 0n no 00 mm on mu 0... mm on mm on m. o. m I {W . lllllee o F m- . ”mahqu floz sllssssessessssse. pounced . _ . . o m 106110011! JO .ta'qmnN 8881101186.! 25 a typical discrimination problem than an errorless approach. On the other hand, the control group shows a flatter graph of latency that would normally be more closely associated with ”errorless” learning. An examination of the original learning data for FoFt (see figures 3 and h) further complicates the findings because of a marked difference between the two groups on the same task. Both groups show the upward acquisition slope in the latency curves and the downward slope in the incorrect frequency curves. The difference between the two groups is in the rate of this acquisition. FoFt learned the original discrimination faster than FoNFt. We can see that FoFt had definitely learned the discrimination by trials 21-25. FoNFt did not learn until trials 66-70. The two graphs in this case serve to corroborate each other and the above results could be easily determined by an examination of either set of data. In comparing the data of the control group and the experimental groups during OL, we can see that the first hypothesis is not as strongly supported in these terms as was originally determined (see Table 3). That is, although fading in the 8- may allow more 8; to learn a discrimination, such an approach does not necessarily provide better dis- crimination performance than a situation in which 8+sS- throughout acquisition. The experimental groups had a total mean error of 16.7 as compared to the learned control group with h.0 average errors. During acquisition this 26 . . menu: menu-Gouda mam-5.1.0 AduGM$Ma0QKWU°ImUMIA+mVNlm oazmah mgcwafi 0900 Do mm on 2. on no on mm on me ov me on mm cm 9 0.. m odpl Oh I ' '1 ee "I. 1 co m \\\\\oooao ’51 m \.N 99 . .... m \ ...... 0.0 1... \" ‘ O a \ x \ s \ ‘ ooo \l). x / \. J . I b.‘ to s .\ .\ eon-see: .eeeee . O P + 6 \... I t \ eeee coffee eeee MW \. mem- .eem~koeeeee 000000 OOOOC'OOOOO 0..“ 000‘00 . We ..... ...... on... s .- e.e ( ..... 8 N + a am Oh ........ I-.. V unflzorm 06 N + one .mzou 2'7 mafia: munch d dofiuaonx on: no.“ m.om:M®moa.auw Uuw wonfisa Somme/TV ousmah afioona . . mAdMaB COP no 00 mm on on Ox. no 00 no 00 mm Cu no on ma 0N mp OF 0 .. ,. . ........... ...... x ..... o oo oo eooo ooo ooobeeo oooeeoo oooo ...... . .... 00’. ooeoe : n .... . o. @- emom ......... ahZHOh ON 10 .169qu zoeasoou; sesuodses 28 n.c «huh “O“O“ o o e o o OGOON each new— . 0.5 m n.~ n.o n w.“ «ea c b.n 5.0 N h.m n.w o Heeea eeee eezee Hevcoldao um ooononeom eooaneocH noel}: coextn onnsh eeeeeeea unannoed nehocsus N ““00 [NOW bk Hones pounced undcheed Hendudho «hack 9.6“ 0e: ceau 0.9 m.u~ «.9 o.m_ .e.e men" men uoz pounced "chance «.0 vhzoh «omen... O O O O O dunno. .90“:— .ooulmn .ns 1—m .on one .mm 1_ .oo—In .oo—uwm ens 1—n .on new .mmlnhl anode. 29 relationship hold true throughout. The contamination of results due to the differences between the two groups on the same task during 0L necessi- tated an analysis of covariance of i(8+)-Y(8-) latencies between 0L and TL for each group (see Table h). The analy- sis was performed on the four blocks of twenty-five trials and on the whole one hundred trials. This analysis tells us the mean magnitude of i(8+)-i(8-) latency deviation free the average between 0L and TL and the direction of this deviation between these two tasks. For the first twenty- fivo trials, then, we may say that both groups varied with comparable magnitude in the same direction for the two tasks. Although both varied in similar ways during trials 26-50, the magnitude of this direction was greater for the FoFt group. Trials 51-75 Provide a complete reversal of this situation. During criterion trials, the results again reverse with FoFt showing the greater magnitude of devia- tion. For the entire series of trials, however, we can see that FoFt varied negatively while FoNFt was positive. Since Table h-Analysis of Covariance of i18+)-§(8-) Latencies FoNFt FoFt trials' 1- 25: +0.2 +0.“ 26-,50: +0.5 +3.3 51" 75' ¢8o2 *0o6 76-1008 +0.7 +7.3 1-1008 +2.2 -1.5 30 both groups covaried in the same direction with comparable magnitude during the first twenty-five trials we may remove the effect of overlearning from OL to TL as a cause of TL learning differences. As predicted by hypothesis 2, not only did the transfer fading group have less 8- responses both initially, during acquisition, and test trials than FoNFt, but also the f(8+)-§(8-) latency difference was less for the non-fading group throughout the 100 trials (see figures 5 and 6). The only two marked exceptions from this prediction occurred during trials hl-NS and 76-80 in which FoFt did not make a discrimination between the two stimuli. Recovery was imme- diate, however, in both instances. Group FoFt provided no slope of acquisitoon for either dependent variable whereas FoNFt did provide more of a slope, although not as steep as would be expected of a typical transfer situation. We may attribute this latter phenomenon to the OL based on fading. The posttest, administered one day after completion of the transfer task and consisting of twenty-five trials of full presentation of the original problem, provided strong evidence for the null hypothesis. Both groups achieved a high level of proficiency in discriminating the stimuli (see figure 7). The actual number of incorrect responses for both groups was so low as to approach perfect performance. A total of three errors were made, one §_of FoFt made two 8- responses during trials 1-5, and one §,of FoNFt made one 31 AH mcuusu . mowoaoacq mascaU Acucghaaonxfi AlvaTa+vaID oasmah mfidwufi . 00-00 00 mm 00 on Oh no 00 no 00 mm 0v mm. 00 mm 0N up 0- n 0.0 I .. . o.o . .. ... .... .. . e. .. .. ... ... .. . . m s a « owl. o s o o .. .. .. » .. .. .. m m a s . .. . .. .u 00 — + ee ee eee e o see so’ eeee as t O eo oeoo ooeoee eso- eeeeeooe seeeeeoe oooo e.oee oooboo on oeoeessssoeeee‘ee oeee ee . 0.0 P + 0| a... 0.0m + (813110088) 8910118191 32 FoNFt fill-III. sesuodsex 1ooxaoou; go JoqumN 1015 20 25 3O 35 40 45 50 55 60 65 7O 75 80 85 90 95100 5 Trials Figure B—Nfian numbei'aof incorrect reop‘gnses for the. Groups du r i n’g xper imen 33 smosemom one mcaaop monocowoq mazoaO ~oengfifiaonxfi HIWVMTA+mvNT> odomwm mndnufi. mm ON m- or m 0.0. + 0.0. + amok ..... QONI. uhzom (spuooeS) segoue1eq error on trials 21 -25 . 3h Discussion 222 Conclusion The results demonstrate that for a visual discrimina- tion problem a fading-in of the brightness cue for a given stimulus allows more moderately retarded So to acquire this discrimination than if a simple simultaneous method is employed. The utilisation of a fading-in technique, howev- er, does not assure that less 8- responses will be omitted than a non-fading approach. The i(8+)-f(8-) latencies accordingly show acquisition for the fading group and no such slope for the non-fading group. That is, the typical discrimination paradigm may be a closer approximation to errorless learning than fading for a moderately retarded population during a non-transfer visual discrimination task. During a transfer task, a fading transfer is also beneficial as to the number of people who can accomplish it. Using a fading technique, however, is not reflected as being significantly better with respect to either dependent variable. The results further provide evidence of no response disruption of original learning if the intervening transfer task is accomplished with a non-fading approach. That is, although original learning is acquired by fading, an abrupt transfer task provides no significant differences from a fading transfer on recall of OL.material as measured by the dependent variables. As may be seen, the results vary significantly from what may be logically extrapolated from Terrace's studies. 35 36 The disagreement between theory and empirical testing occur in two main areas. First, fading the brightness cue is not the only or is not a prerequisite to errorless learning. Secondly, despite the fact that fading original learning is not errorless more So learn a transfer through fading than by abrupt transfer. The failure of fading to provide errorless learning during OL.-ay be due to one difference between Terraco's experiments and the present one. That is, the latter did not concern itself with the duration of the presentation of stimuli. There could be some importance attached to the length of viewing of stimuli. It may be that if the dura- tion of the cues during presentation are longer, more atten- tion (for example) would be paid to them. The assumption involved in this hypothesis is one that may not necessarily be borne out. Shmply because the stimuli are presented for a longer period of time does not mean that an §,will attend to them that much more. Evidence for this may be found in the experiment just presented. In it, the moment of effec- tiveness of S- as a stimulus was assumed to randomise out over all So. It could be that we have here a skewed sample which did not randomise between groups FoNFt and FoFt whore, during 0L, there was such a large difference between these groups on the same task. Essentially the same problem is involved in both concepts-how can an §,be "forced“ to attend to a stimulus. .A longer period of presentation would not seem to solve the problem. 1 III I! I lull Illlllllll II ell" III III III. ..II I. I I III I I | |. 37 One way to assure the effectiveness of the stimuli is to have self-determining number of trials to criterion. That is, whether the stimulus is effective during any given trial is not important. What is important is that effective- ness does result after a number of trials. Allowing all So as many trials as needed to reach a given criterion (that is, to learn a discrimination) and then fading-in the next brightness level would (at least theoretically) be an an- swer. Adopting this technique to the present experiment would be a relatively easy matter. The same general design could be used, but instituting a criterion of five (5) consecutive correct responses before the next fading incre- ment (or decrement). Utilisation of this method would also solve the problem of variable overlearning trials that was inherent in the present experiment. Another explanation of the discrepancy between the results of Torrace's and the present experiment may be due to factors inherent in the apparatus of this experimenter's investigation. Every trial may have been unwittingly rein- forcing duo to a satisfaction of gfs manipulatory drive. Each g'had control over the stimuli presented to him. He could shut off the projector and turn it on again simply by pressing a panel (either panel). The room was dark in the interim and the possibility of new stimuli were imminent. The sound of the projector was also under their control. All this could have served as a conditioned reinforcement. Another confounding problem evolves from this theory in that 38 §pis asked to make not only a visual discrimination, but also a reinforcement differentiation. The greater reward occurred with an 8+ (or 8'+) response because this was accom- panied by a primary physical (MAM) and secondary auditory (click of the MAM dispenser and occassional verbal rein- forcement) reward. An objective determination had to be made as to which reward was greater. A subjective determi- nation also was forced upon §,in that he must determine if one set of reinforcement was better, worse, or made no difference. The problem raised here may be partially solved by the use of a non-automated UGTA. A.simple sliding drawer with food wells covered by physical stimuli would somewhat lessen the secondary reinforcement inherent in the automated appa- ratus. This device would not, of course, provide the high degree of time measurement accuracy as the automated appara- tus, but this fact would perhaps be compensated by a lesser degree of confounding. A second factor, intrinsic to the present design that may have accounted for the results, was the ITI. Between trials a period of 6.8 seconds elapsed. This time length may have been too long for the population involved. A shorter ITI may produce less of a strain on their short term memory (STM). We had assumed that the carry over from trial to trial would be great enough for a discrimination to result. The results may be indicative of the invalidity of this assumption in that a long 8TH functioned to counteract 39 the fading crutch. Any further experimentation in this area should, therefore, be preceded by a determination of the most beneficial time between trials for the given popula- tion. Not only may STM be strained, but too short an ITI could confuse the subject. STM may also be a factor involved in the discrepancy of the oxperimenter's results and those of Moore and Goldiamond (discussed previously). In the present investigation STM was utilised when fife eyes traveled between the two windows and the ITI. Moore and Goldiamond also had STM between trials. But within the trial the time involved for STM was longer and forced. That is, an S’had to attend to the sample until it was removed and then to the matches which were then presented. Since it was impossible to refer back to the sample, the §_was forced to rely on his memory of the cues in the sample, That is, there was no reference back beyond the initial surveying of the sample stimulus. The differences between the SIM in the two experiments may have been compensated by the differences in the I.Q.'s of the populations. He should remember, however, that they uti- lised younger g; with higher I.Q.'s than the present study with older children but lower I.Q.'s. The mental ages, therefore, might be equatable. Since data on intelligence was not provided, no I.Q. comparisons between the two exper- iments may be made. An analysis of intelligence data from the present experiment provide some insight into its problematic results to (see Table 5). 1.0., M.A., and C.A. distributed in a highly equatable fashion between the control and experimental groups. Within the control group, the non-learners were those with a lower C.A.. In fact, the low C.A. measures of the non-learners in the control group and the deleted mom-- bers of the experimental group were equatable as were the learners in the respective groups. That is, we may postu- late a C.A. deficit as a contributing factor in not learning independent of method of acquisition. More specifically, those 2; with a mean C.A. of 7.0 or less did not learn to respond to 8+ to criterion no matter what visual crutches were utilised. He can now understand why a greater percentage of the fading g; learned during OL. The g. were so distributed as to have a lower percentage of low C.A. g. in the fading group than in the non-fading group. The fading, therefore, had no effect. During transfer, however, fading did provide less errors than no fading. This effect occured despite of an M.A.fO.A. deficit for the fading transfer group. We may conclude, therefore, that fading during a transfer situation does indeed assist in learning even though such learning is not based on errorless OL. This latter conclusion is our only remaining contradiction to Terrace's theory. We mayffind a clue to this predicament by returning to an analytic comparison between the present study and that of Moore and Goldiamond. These investigators utilised #1 ... ..Hu ... ..E 3.5 1.4.0 m.e u.<.o 353 31m Leta. omen Amen. «amen Tn EMF. a. n 1.4.: “6 ...Hum 2n... ..<..l a.» unmuw a. m u.¢.o ~.e e. .o e.ne .4. o SEE 35H... 2......“ Rem .~_uz. veneeea Anna. eeeeaen Anuzv veneeeq Anna. eezeeeq eez a.n naupm m.n unmum web .oE Qoh 'o o ....E .....m Les-2. en Ae_nz. eez macho Houanoo macaw Heucoldam Ha enoono unoad ooneudndovsH Mo oohseoe: no :odusndnvudnln oupes 52 seventeen (17) fading steps while the present study (as in Torraco's experiments) had three (3). The former study also accomplished good stimulus control. There was little dis- turbance in the ongoing learning process with seventeen fading steps. The present study did, however, find disturb- ance with the 3 “jumps.” This problem points out an important question involved in any study concerned with a fading variable. That is, after a fading-in step, how much of a new task is being presented to g: In the present study, the non-fading trans- fer group (FoNFt) was given an abrupt transfer (or very large sise-of-stop fade-in) after trial fifty. Their laten- cies were consequently disturbed and the number of incorrect responses immediately increased. It would seem that there is a negative correlation between the amount of disturbance, in so far as the number of incorrect responses are concerned, and the number of fading steps (in this case 1 vs. 7). For group FoFt only a few of the fading-in steps caused great latency disruption. The last fade-in (trial s6venty-six) was the most notable exception. Obviously, the change in stimuli was too great. The fading-in caused a 7.5 second average latency. This figure is far different from any other latencies for either group during a fading-in step. The last fading-in for group FoFt can now be interpreted as much more of a presentation of a new task than previous "jumps.“ This process was similar to the one abrupt trans- fer step (trial fifty-one) of group FoNFt in which this ”3 latency equaled that of their first trial (a new task). Although disturbance occurred for FoFt during TL, there was less such disturbance. Therefore, fading during transfer assists the moderately retarded S'to learn a visual discri- mination. A continuum of fading steps to find the maximum number of ”jumps” needed for the least latency disturbance and number of 8- responses would be one extrapolation of this study. It is expected that the greater the number of stops, the less the disturbance until an asymptote is reached. Beyond this point the profitability of further fade-in's would decline. This approach would be similar to that of Sidman and Stoddard (1966) in which a program is changed by the experimenters as it is evaluated by thelgs. The present experiment suggests that not only the number of incorrect responses be used in evaluation (to determine criterions before fading-in steps), but also the latencies before and after the ”jump.“ Too low a latency difference below a predetermined standard would indicate a mistake in the program, and a need for less of a change. This approach could mean, therefore, that not all steps should have equal psychophysical and/or quantitative changes. The relation between this study and programmed instruc- tion has already been alluded to. The results of this experiment point the way toward a subjectively determined (by population and individual) variability of a general program for any given subject matter or theoretical study. an The unique capability of simultaneously teaching and evalu- ating the behavior of the mentally deficient is inherent in this approach. Terminal behavior and eventual potential may thusly be reached. The malleability of a given program, in addition, is such that many levels of retardation may be dealt with, without a satiation effect. The most succinct method may be used, provided no informational value is lost, by both profoundly and moderately retarded. The author, of course, realises the limitations of this mode of learning presentation. The potentialities, however, are not known. LIST OF REFERENCES LIST OF REFERENCES Denny, M. R. A theoretical analysis and its application to training the mentally retarded. Chapter in N. R. Ellis (Ed.), International Review 2; Research pp.Mental Retardation, Vol. 2, 19 Denny, M. R. and Adolman, H. M. Elicitation theory; I. An analysis of two typical learning situations. Peychol. RCVe’ 1955’ 22-, 290-296e Gynther, M. D. Differential eyelid conditioning as a func- tion of stimulus similarity and strength of response to the Geo lo OXEo P.1Ch91e, 1957’ 22, “08-“!6o Hall, J. F. The Psychology 25 Learning. New York: Lippin- cott, 19 Hanson, H. M. Effects of discrimination training on stimu- lu; generalisation. g. exp. Psychol., 1959. 22, 321- 33 . Holland, J. G. Teaching machines: An application of prin- ciples from the laboratory. g, exp. Anal. Behav., 1960, 1, 275-287. — "‘— Hull, 0. L. The problem of stimulus equivalence in behavior th.orYe P.13h01e R.'e, 1939’ 2-6., 9-30o Hull, C. L. Principles p£.Behavior. New York; Appleton- Century-Crofts, 1953. Hull, C. L. Stimulus intensity dynamism (V) and stimulus generalisation. Psychol. Rev. 1999, 2Q, 67-76. Hull, C. L. Behavior postulates and corollaries-1959. Psychol. Rev., 1950, 21, 173-180. Hull, C. L. A,Behavior system. New Haven! Yale, 1952. Lumsdaine, A. A. Some issues concerning devices and pro- grams for automated learning. In A. A. Lumsdaine and R. Glaser (Eds.) Teaching Machines and Programmed Loarnin . Washington: National Education Association, 1960. “5 #6 Moore, R. and Goldiamond, I. Errorless establishment of visual discrimination using fading procedures. g, exp. Anal. Behav., 196%, 1, 269-272. Sidman, M. and Stoddard, L. T. Programming perception and learning for retarded children. In Norman R. Ellis (Ed.) International Review 2; Research pp’Montal Retardation, Vol. 2, 1955. Silberman, H. F. Self teaching devices and programmed “RtCPARIQo ROVe'EdUCo Roao, 1962, 32, 179-193o Skinner, B. F. Teaching machines. Sci nce, 1958, 128, 969-977o Spence, K. M. The nature of discrimination learning in animals. Psychol. Rev., 1936, h , “27-hh9. Spence, K. H. The differential response in animals to stimuli varying within a single dimension. Psychol. RCVo, 1937’ 22, “30-53“. Spence, I. U. Analysis of the formation of visual discrim- ination habits in chimpanzee. g, Comp. Psychol., 1937., 22. 77-100. Terrace, H. 8. Discrimination learning with and without errors. g. exp. Anal. Behav., 1963, 6, 1-27. Terrace, H. 8. Errorless transfer of a discrimination across two continue. {1. exp. Anal. Behav., 1963a, 6, 223-232. Terrace, H. S. Stimulus control. In I. K. Honig (Ed.) Operant Behavior: Areas 2; Research and Application. New York: Appleton-Century-Crofts, 1966. APPENDIX APPENDIX Apparatus. A semi-automated version of the two discrimi- nanda Wisconsin General Testing Apparatus (UGTA) was used. The §_sat in a semi-soundproof metallic gray cubicle facing a similarly colored wooden board. On the board were con- tered two transluscent screens which could be pushed by S’in order to make a response. Centered beneath the screens was a reward tray under which was located a small shelf. These objects were placed so as to be within easy reach of 8 (see figure 8). Behind the discrimination board were the power supplies and electronic equipment (not seen by §_during the actual running of the experiment). A projector was aimed at the transluscent screens. An electric timer, accurate to 1/100 seconds, was connected to both the projector and to the screens. The presentation of the stimuli on the windows (trial initiation) was accompanied by the start of the electric timer. A trial was terminated by pressing one of the screens, thereby stopping the clock and projection of stimuli for 6.8 seconds (the interstimulus interval). This cutting-off procedure was accomplished by two micro- switches, one placed behind each of the windows, which relayed to the clock, the projector, and the experimenter's panel lights. In this manner, the latency of responses was accurately measured. gpwas provided with a panel which indicated (by lights) which screen, left (L) or right (R), “7 moccacnod- waauoob monocomv canuooor’le oaswsh . 0 3o Cr" unwaoro a o>O M30 Aocom :0 oz one canon: nonxm you: who “GM chdgmll IIIIII “IIIIWI.IIM vioofl m anus o :5 . hos Goon om. . 5...: 4-.. . .nawmpom U ““3 I Olulldllmqm we." ow acoom: H m: an. ..aomaoamann 282 . moo o coasgauoma ...oanwfinwzo U m 0 . . .Q hfiansm: a: a» o“ no3om . . monk. )C. IIIII I...‘I..'I. . 6h “me . - 8 PH.- . ---..........m ...... ...... .....uouooHOum mfimohom. uaooo5~o£ modem one o «God: a onxfi....\ ® “9 had been pressed by S. A panel key also allowed 2 to pro- vide a reward from the MdM dispenser which was located in back of and above the screens. Two keys for manual cpera- tion of the projector, forward (F) and reverse (R) completed the apparatus. Stimuli. The stimuli were projections of line drawings of a man striding, a dog standing on all fours, the word MAN in capital letters, and the word DOG also in capital letters (see figure 9). These stimuli were presented by means of h25 black and white slides. These slides were actually hand-mounted negatives of gray tempera color drawings on clear acetate. The darker the gray color against an almost white background, the more light was allowed through that area on a slide when projected. The independent variable of brightness was thereby controlled. The grays were mixed from black and white tempera colors and matched by eye to a psychophysically determined gray series.1 This continuum consists of 16 equidistant grays ranging from one unit above pure white (G1) to one unit below pure black (G16). 01 was used as the background upon which the forms were photographed. The drawings them- selves utilisod G#, G7, G10, and 013 of the series, each representing a higher degree of light intensity as a pro- jected negative. Four etchings were made of every object, one in each 1 This gray series was developed by Dr. William T. Stellwagen of Michigan State University. 50 “:5: am oSHIG 0.2...th Q... 4 &m 0:50amxoom o 3 oaza chommnamnz #0 1:0. on oadn .deonawmfld mg. 000 MfiU TU o>aaomoc o no ooocanmaan unannouonu 51 group painted with a different color gray according to the above prescriptions. Uhen photographed, one figure was simply overlayed (or not) onto its matching word according to the conditions of the experiment. The photographs were taken with a copy set up and a Pentax camera with a 55mm. lens, loaded with Pan-X film ASA 32. Two 3,2000floodlights in 10” reflectors maintained an intensity of 375 foot candles at the site of the windows. The film was exposed for 1/15 second at a 5.6 stoppage. The height from the lens to the surface was #0”, the drawings being separated by 16' from their centers. This same tech- nique was used for the pretraining stimuli which consisted of a picture of jello and a picture of an airplane both taken from a magasine. In this case, however, indoor color film (ASA 25) was used. EEB 1 5 1931; 3 ILDIIIIZIIIJILIILIIJLI WIN 0