Perceptual and Motor Skills, 1979,49, 171-180.
@ Perceptual and Motor Skills 1979
BEHAVIORAL CONTRAST DURING THE ACQUISITION OF SPEECHREADING' C. MERLE JOHNSON Central Michigan University
AND
JAMES H. KAYE
Kalamazoo Valley Intermediate School District
Sumnaasy.-Two hearing-impaired children were taught to speechread the names of 18 objects divided into three sets. A multiple baseline across sets of objects was obtained for each subject. Tokens presented contingent upon correct responses to certain sets of names increased the frequencies of those correct responses. Both subjects exhibited positive behavioral contrast during the acquisition of speechreading; correct responses to one set increased above the previously established level when the contingencies for a second set changed from token reinforcement to extinction. The obtained contrast effect was conducive to learning and the procedure was effective in teaching the children co speechread the training words.
Speechreading or lipreading, the technique of discriminating the content of spoken language by attending to visual stimuli provided by the speaker's Lips and face, is an integral part of the oral method of education for deaf or hard-ofhearing children (Farwell, 1976). Some deaf children fail to learn speechreading, especially if' they have concomitant problems (Berger, 1972; Myklebust, 1960). Alternatives to speechreading include manual communication which utilizes fingerspelling and American Sign Language (Ameslan), or total communication which utilizes a combination of manual communication with the oral method (Freeman, 1976). Such alternate methods are not always available in public schools, and consequently some of these children are deprived of educational services because of their inability to function in oral method classrooms ( Kohl, 1966). A behavioral approach in speechreading training suggests early disctimination training using differential reinforcement. Johnson and Kaye ( 1976) demonstrated that a deaf multihandicapped child could be taught through token or social reinforcement to discriminate the names of objects based on both voiced and nonvoiced stimuli. The present study sought to explore and extend behavioral training procedures useful with severely hearing-impaired children. Specifically, two multiply handicapped children were trained to speechread the names of three groups of objects. 'This research was supported. in part, by an ESEA Title VI-B grant awarded to the Kalamazoo Vallev Intermediate School District. Tames H. Kaye, Project Director. The authors wish to thadk Paul S. Wollam, ~uperintenhent, and ~ n n l e eDecent, Director of Special Education, as well as Gerald L. Shook, Coordinator of the Kalamazoo Valley Multihandicap Center for providing an environmenr which fosters the developmenr of applied research within the educational process. W e also wish to thank Jack Michael, Brian Iwata, and James Cowart of Western Michigan University for their editorial assistance and consultaoon. Reprints may be obrained from C: Merle Johnson. Department of Psychology, Central Michigan University, Mount Pleasant, Mlchigan 48859.
C. M. JOHNSON
172
&
J. H. KAYE
Subjects
Two hearing-impaired children enrolled in the Kalamazoo Valley Multihandicap Center served as subjects. Both children had failed to respond effectively to previous speechreading techniques. However, they could fingerspell and had a small repertoire of American Sign Language. Brian was a 12-yr.-old boy who was diagnosed as hearing impaired, speech impaired, and mentally retarded. His unaided hearing in the better ear averaged 55 dB (ISO, 1964) over the frequencies of 250 to 4000 Hz. His spontaneous speech was hypernasal and unintelligible due to a severe cleft lip and palate. H e resided at a nearby state institution for children and actended school weekdays at the Center. Lori was a 10-yr.-old girl who had been diagnosed as deaf, mentally retarded, and emotionally disrurbed. She evidenced a profound sensorineural hearing loss ranging from 90 dB to 110 dB (ISO, 1964) from 250 to 4000 Hz. Lori had entered the Center at age 8 and had exhibited hyperactive, destructive, and self-abusive behaviors. Apparatus and Setting
Training was carried out within a sound-attenuated booth measuring 1.5 X 1.8 m. A Panasonic closed circuit television camera with an 8.5-mm wideangle lens was located on one wall near the ceiling. A Sony F-98 microphone was suspended through the ceiling and out of reach. The experimenter and subject sat facing each other with a table located between them. At one end of the cable was a Gerbrands (Model B) token dispenser. Three sets of six objects were utilized as training stimuli and are represented in Table 1. In the first set TABLE 1 SETS OF OBJECTS ARRANGEDIN A 3
banana pear
2 HORIZONTAL ARRAY
Group B
Group A apple peach
x
orange plum
blue purple
green red
Group C orange yellow
1
2
4
5
3 6
were six plastic fruits, in the second six 5- X 5-cm cards with names of colors printed in their respective colors. The third set contained six 5- X 5-cm cards with numbers printed on them. Prior to baseline, a single training session was used to teach the subjects to point to objects whenever the experimenter vocalized an object name. Correctness of a response was not a factor for social reinforcement. Reinforcement was presented in the form of praise in sign language, e.g.. the sign for "good." The criterion for reinforcement was for the subject's hand or extended fingers to come in contact with one of the six objects within 5 sec. after the experimenter's vocalization.
BEHAVIORAL CONTRAST I N SPEECHREADlNG
173
A three-component multiple baseline (Baer, Wolf, & Risley, 1968) containing withdrawal phases (Hersen & Barlow, 1976) was obtained for each subject. The training objects of each set were placed, one set at a time, in a 3 x 2 horizontal array on the cable between the subject and experimenter. The presentation order was Set A, Set B, and then Set C. The child left the booth for a drink of water at the cqmpletion of each set at which time the experimenrer changed the objeccs on the table. Before each trial the experimenter tossed a die into a bowl held under the table. Each face of the die corresponded to one of the six objects in that set. After noting which face came up, the experimenter established eye contact with the subject and then vocalized the name of the corresponding o b j m . The objecr name was verbalized only once per trial and normal conversation levels were used at all times. The dependent variable was the subject's pointing to one of the six stimuli on the table afrer each vocal presentation. Pointing was defined as the subject's hand or extended f~ngerstouching one of the six stimuli. Each of the pointing responses was recorded as correct or incorrect. In the case of mulriple responses on the same trial, only the first response was counted. A correct response was defined as pointing to the appropriate object within 5 sec. following the vocalization of an object name by the experimenter. Incorrect responses were defined as incorrect matches between objects and object names or the absence of pointing responses within the 5-sec. limited hold. Each object was presented six times in a random fashion each session. Component responding was defined as the total number of correct responses by the subject to each of the three sets. The responses were tallied on a piece of paper placed out of the child's view. Throughout the experiment social interaction between the subject and experimenter was minimal. During reinforcement phases tokens were delivered for correcc responses via the token dispenser to limit social interaction that may have confounded the magnitude of reinforcement. Tokens were saved by the subjects and traded for candy or toys at the Center's store at the end of the day. Bu~e1ine.-During baseline the experimenter vocalized an object name and the subject pointed to one of the six stimuli in front of him. The experimenter simply recorded the subject's response, giving no indication whether or nor it was correct. Phase I.-The first manipulation was the contingent delivery of a token to the s u b jecc for each correct response emitted to Sers A and B. The experimenter activated the token dispenser for each correct response and then terminated eye contact with the subject by recording the data for that trial. Incorrect responses to Sers A and B, as well as all responses to Set C, were not reinforced. Phase 11.-During this phase correct responses to Set A were reinforced with tokens while incorrect responses to Set A were not reinforced. Neither correct nor incorrect responses to Sets B and C were reinforced. Phase Ill.-Correct responses to Sets A and B were reinforced again with tokens. Incorrect responses to Sets A and B and both correct and incorrect responses to Ser C were not reinforced. Phase 1V.-The reinforcement contingency applied to correct responding to Set B was changed from reinforcement to extinction a second time. This phase was reinstated for Lori to determine if the effects of Phase I1 could be replicated. P h a ~ eV.-Reinforcement was applied for correct responding to Sets A and B for a third time. Phase V1.-Correct responses to all three sets were reinforced with tokens and no incorrect responses were reinforced. This manipulation was made to determine the effect of reinforcing correcc responses to Set C which had previously been in baseline conditions. Phase VI1.-In this phase all 1 8 objects from the three sets were placed on the table
C. M. JOHNSON
&
J. H. KAYE
FIG. 1. The number of correct speechreading responses emitted by Brian to each set of stimuli during each session in a 6 x 3 array. Any object could have been presented on any given trial although the data were recorded in three separate components as before. A second die was added to randomize sets. All correct responses were reinforced with tokens while incorrect responses were not reinforced. Phase VII1.-The final manipulation was carried out without any of the two or three-dimensional objects used previously. At the start of the first session of this phase, Lori was given instructions in American Sign Language to sign the name of the object instead of pointing. N o further instructions were provided and the table was cleared except for the token dispenser. The contingency in this phase required the subject to sign correctly the name of the object wirhin 5 sec. following the vocalization of an object name. Correct responses were reinforced with tokens and incorrect responses were not.
Reliability Interobserver reliability checks were taken about every nine sessions. The observer watched the entire session through a closed circuit television monitor' in a separate room and recorded data in the same manner as the experimenter. A trial-by-trial comparison was used and reliability scores were computed by dividing the number of agreements by the total number of agreements plus disagreements. Four interobserver reliability checks were taken by three different observers during Brian's experiment. Reliability ranged from 96% to 100% with a mean of 97 5% Twenty reliability checks were taken by 11 different
BEHAVIORAL CONTRAST I N SPEECHREADING
a
I
. m
a
9
D
NUMBER OF CORRECT RESPONSES
m
176
C. M. JOHNSON & J. H. KAYE
observers during Lori's experiment. Reliability during this experiment ranged from 94% to 100% with a mean of 98.9%. RESULTS Fig. 1 depicts the levels of correct responding to each set by Brian. Brian exhibited a falling baseline to Set A and stable baselines to Sets B and C. In Phase I, Brian demonstrated a rapid acquisition of speechreading the objects in Set B and on session 29 he achieved 100% correct for that set. Brian showed a slight increase from the end of baseline for Set A in Phase I, but he did not exceed the correct responding level that he had exhibited during the first few days of baseline. However, when the contingencies for correct responding to Set B were reversed from reinforcement to extinction in Phase 11, correct responding to Set A immediately increased and continued to rise until the end of the phase. Correct responding to Set B showed a slight decrement initially in Phase I1 but remrned to previous levels by the end of the phase. Correct responding to Set C remained stable throughout the experiment. Unfortunately, at this point Brian was transferred to another state institution, necessitating an early and unforeseen termination of his participation in the experiment. Fig. 2 portrays the mean number of correct responses by Lori. Her baseline levels were generally lower than Brian's and she exhibited a falling baseline to Set A. During Phase I correct responding to Set A increased rapidly and then stabilized at about 16 correct responses per session. Correct responding to Set B made slower increments during Phase I but stabilized at about 19 correct responses per session. Responding to Set C remained stable during the first phase. When the contingencies for correct responding to Set B went from reinforcement to extinction in Phase 11, correct responding to Set A immediately increased from about 17 in Phase I to about 23 in Phase 11, even though there was no contingency change for correct responding to Set A. Correct responding to Set B showed a slight decrement when going from reinforcement to extinction in Phase 11. This behavioral interaction was similar to that shown by Brian during the same manipulation. In addition, correct responding to Set C remained stable. In Phase 111, correct responding to Set B increased and later stabilized at about the same level as correct responding to Set A. Throughout Phase 111 correct responding to both Sets A and C remained stable. When reinforcement was withheld for all responses to Set B in Phase IV, the effects of the first reversal in Phase I1 were replicated. Correct responding to Set A immediately increased from about 23 in Phase I11 to about 27 and continued to rise to about 30 by the end of the phase. Correct responding to both Sets B and C remained stable during Phase IV. In Phase V correct responding to Sets B and C remained unchanged and correct responding to Set A became slightly more variable. In other phases similar to this one, correct responding to Set B had increased.
BEHAVIORAL CONTRAST IN SPEECHREADING
177
However, the data showed that Lori was consistently missing red and green in that set, resulting in no improvement. When all three components were reinforced in Phase VI, correct responding increased in each set. Component responding to Set A increased first and attained 100% correct during Session 137. Next, component responding to Set B increased and attained 100% correct during Session 146. However, it should be noted that during the days of Sessions 135 and 136 a substitute therapist had worked with Lori on identifying colors outside the experimental setting and one of those colors was red. The data from previous sessions showed that Lori had consistently missed both red and green for over 20 sessions. The effects of this mishap are not completely known, but it may have accounted for some of the increase in correct responding to Set B. Regardless, a rapid increment in correct responding to Set C ensued and all three components remained near the 100% correct level until the end of the phase. When all three groups of objects were presented simultaneously in Phase VII, Lori exhibited a slight decrement in correct responding to each set. However, by the end of the phase she recovered most of the loss. When all of the objects were removed in the last phase, Lori apparently had no difficulty with the response generalization since she maintained component responding near the 100% correct level.
DISCUSSION The multiple baseline containing withdrawal phases indicated that in no instance did the number of correct responses to any given set of stimuli improve during a given phase unless tokens were presented contingent upon correct responses. Furthermore, the obvious discrepancies in the number of correct responses between the reinforced sets ( A and B) and the nonreinforced set (C, up to Phase VI for Lori) for both subjects also demonstrated the effectiveness of token reinforcement. Interestingly, substantial changes in correctness were obtained using tokens in relative isolation; great care was taken to eliminate or greatly reduce the likelihood of social reinforcement during experimental sessions, and no primary reinforcement was employed in order to control for spurious or momentary changes in reinforcement magnitude or quality. It would seem likely that even more dramatic and rapid gains would result from combining social, primary, and token reinforcement in a more traditional therapeutic-educational setting (Johnson & Kaye, 1976). The low, steady rate of correct responding to Set C for both subjects until tokens became available for such responding suggests: ( a ) limited impact of possible maturational variables, ( b ) little or no practice effect, and ( c ) little or no generalization to Set C from correct responding to Sets A and B. On the other hand, the rapid acquisition of Set C correct responding for Lori in Phase
178
C. M. JOHNSON
&
J. H. KAYE
VI, as well as the slightly less dramatic improvement in responding to Set B in that phase, may have resulted from the development of a learning set (Harlow, 1949; Kimble, 1961) during correct response acquisition to Set A. Lori's previous learning history with Set A may have made her acquisition of the behaviors related to Set C easier and possibly to Set B as well. However, an error analysis of Lori's responding provides a more reasonable account of the initial sharp rise in correct responding to Set B in Phase VI. Errors occurring to Sets A and C were observed to be random. The additional uncontrolled training Lori received for red in Set B probably was primarily responsible for the initial rise in her correct responding to colors in this phase, although subsequent gains could be related to the transfer of training through learning sets. A between-sessions analysis showed that those gains came just after correct responding to Set A reached asymptote. Similarly, the dramatic gains in correct responding to Set C came just after correct responding to Set B reached asymptote. Such gains would be predicted by the facilitative effect of learning sets. When correct responding to Set B changed from reinforcement to extinction, correct responding to Set A increased above previously established levels even though there was no contingency change for responding to this set. This interaction between responding to Sets A and B appeared to be similar to the phenomena Reynolds ( 1961b) defined as positive behavioral contrast. Reynolds defined behavioral contrast as a change in the rate of responding during one procedurally constant component of a multiple schedule in a direction away from the rate of responding generated by a procedural manipulation during a second component. When the rate change in the constant component is an increase, it is called positive contrast; when it is a decrease, it is called negative contrast (Reynolds, 1961a). Contrast effects have been previously demonstrated with humans. O'Brien (1968) demonstrated behavioral contrast with two mentally retarded girls pushing buttons on a multiple variable-interval extinction schedule of reinforcement. Waite and Osborne (1972) reported behavioral contrast with school children exposed to similar schedules. Fagen ( 1979) found that 3- to 4mo.-old infants exhibited positive behavioral contrast during footkick conditioning when switched from multiple CRF-CRF to multiple CRF-EXT schedules of reinforcement-similar to the present study. Taylor ( 1970) and Couture ( 1976) have demonstrated behavioral contrast with college students key pressing on multiple VI EXT schedules of reinforcement. The contrast in O'Brien's ( 1968) experiment diminished and is called transient behavioral contrast whereas in Waite and Osborne's (1972) experiment the contrast did not disappear and is referred to as sustained behavioral contrast (Nevin & Shettleworth, 1966). The contrast in the present study was sustained by both subjects throughout the phases in which it appeared. In addition, this increase in correct responding
BEHAVIORAL CONTRAST IN SPEECHREADING
179
levels to Set A during Phase I1 was replicated in Phase IV in Lori's experiment. However, the modest decrease in correct responding to Set B during Phase I1 was not present in Phase IV. Behavioral contrast, demonstrated by an increase in responding in one procedurally constant component, does not necessitate a decrease in responding in the changed component (Halliday & Boakes, 1974; Wilkie, 1971; Fagen, 1979). Since the subjects were learning a new skill, it is not surprising that component responding in Set B did not decrease when placed on extinction, as would be expected in most contrast situations. Once the children had learned to speechread certain items, they continued to respond correctly without token reinforcement and this irreversibility of behavior resulted in therapeutic gain. The present study differed from other studies finding behavioral contrast with human subjects because previous experiments were concerned with measures of performance such as key pressing. Behavioral contrast was evidenced in this experiment while both subjects were learning a complex task such as speechreading. It appears as if schedule interactions such as behavioral contrast may be important factors in applied research such as the present experiment and that applied researchers should be aware of such phenomenaZ (Sulzer-Azaroff & Mayer, 1977). For instance, Lovaas and Simmons (1969) noted that self-destructive behavior by an autistic child increased with non-punishing adults after decreasing with a punishing adult, even though there was no contingency change with the non-punishing adult. The increase in correct responding to Set A during contrast periods demonstrated that such contrast effects can be conducive to learning. Since the contrast training procedure seemed to be effective in increasing behavior, some consideration should be given to its possible utility. Therapeutic gain more closely aligned with successful language functioning was demonstrated by Lori in the last rwo phases, wherein she transferred responding from pointing to objects to signing the name of the objects in sign language, a more common and functional set of responses for the hearing impaired. At the completion of this experiment, generalization procedures to other people and further discrimination training to other words were implemented, resulting in even more functional skills for Lori. REFERENCES BAER,D. M., WOLF,M. M., & RISLEY,T. R. Some current dimensions of applied behavior analysis. loitrnal o f Applied Behavior Analysis, 1968, 1 , 91-97. BERGER, K. W . Speechreding. Baltimore: National Educational Press, 1972. COUTURE, E. T. Behavioral contrast: a human replication. Unpublished Master's thesis. State Univer. of New York at Geneseo, 1976.
7. S. Bailey,
A handbook of research methods in applied behavior analysis. lished book manuscript, 1977)
(Unpub-
'
180
C. M. JOHNSON
& J. H . KAYE
FAGEN,J. W. Behavioral contrast in infants. Infant Behavior and Development, 1979, 2, 101-112. FARWELL,R. M. Speechreading: a research review. American Annals o f the Deaf, 1976, 121, 19-30. FREEMAN,R. D. T h e deaf child: controversy over teaching methods. Jorrrnal o f Child Psychology and Psychiatry, 1976, 17, 229-232. HALLIDAY,M. S., & BOAKES,R. A. Behavioral contrast without response-rate reduction. Journal o f the Experimental Analysis o f Behavior, 1974, 22, 453-462. HARLOW,H. F. T h e formation of learning sets. Psychological Review, 1949, 56, 5 1-65. HERSEN,M., & BARLOW,D. H. Single-case experimental designs: strategies for studying behavior change. New York: Pergamon, 1976. JOHNSON, C. M., & KAYE,J. H. Acquisition of lipreading in a deaf multihandicapped child. Journal o f Speech and Hearing Disorders, 1976, 41, 226-232. KIMBLE,G. A. Hilgard and Marquis' 'Conditioning and learning.' New York: Appleton-Century-Crofts, 1961. KOHL, H. R. Language and education for the deaf. New York: Center for Urban Education, 1966. LOVAAS,0 . I., & SIMMONS. J. Q. Manipulation of self-destruction in three retarded children. Journal o f Applied Behavior Analysis, 1969, 2 , 143-157. MYKLEDUST,H. R. T h e psychology o f deafness, sensory deprivation, learning and adjustment. New York: Grune & Stratton, 1960. NEVIN, J. A., & SHETTLEWORTH,S. J. An analysis of conrrast effects in multiple schedules. Jownal of the Experimental Analysis o f Behavior, 1966, 9, 305-315. O'BRIHN, F. Sequential contrast effects with human subjects. journal o f the Experimental Analysis o f Behavior, 1968, 11, 537-542. REYNOLDS,G. S. An analysis of interactions in a multiple schedule. Journal o f the Experimental Analysis of Behavior, 1961, 4, 107-117. ( a ) REYNOLDS,G. S. Behavioral contrast. Journal o f the Experimental A ~ a l y s i s o f Behavior, 1961, 4, 57-71. ( b ) SULZER-AZAROFF,B., & MAYER, G. R. Applying behavior-analysis procedures with children a d youth. New York: Holt, Rinehart & Winston, 1977. TAYLOR,R . W. Behavioral contrast in human subjects. Unpublished doctoral dissertation, Indiana Univer., 1970. WAITE, W . W . , & OSBORNE,J. G. Sustained behavioral contrast in children. Journal of the Experimental Analysis o f Behavior, 1972, 18, 113-117. W I L K I E ,D. M. Delayed reinforcement in a multiple schedule. Journal of the Experimental Analysis o f Behavior, 1971, 16, 233-239.
Accepted June 19, 1979.
@ Perceptual and Motor Skills 1979
BEHAVIORAL CONTRAST DURING THE ACQUISITION OF SPEECHREADING' C. MERLE JOHNSON Central Michigan University
AND
JAMES H. KAYE
Kalamazoo Valley Intermediate School District
Sumnaasy.-Two hearing-impaired children were taught to speechread the names of 18 objects divided into three sets. A multiple baseline across sets of objects was obtained for each subject. Tokens presented contingent upon correct responses to certain sets of names increased the frequencies of those correct responses. Both subjects exhibited positive behavioral contrast during the acquisition of speechreading; correct responses to one set increased above the previously established level when the contingencies for a second set changed from token reinforcement to extinction. The obtained contrast effect was conducive to learning and the procedure was effective in teaching the children co speechread the training words.
Speechreading or lipreading, the technique of discriminating the content of spoken language by attending to visual stimuli provided by the speaker's Lips and face, is an integral part of the oral method of education for deaf or hard-ofhearing children (Farwell, 1976). Some deaf children fail to learn speechreading, especially if' they have concomitant problems (Berger, 1972; Myklebust, 1960). Alternatives to speechreading include manual communication which utilizes fingerspelling and American Sign Language (Ameslan), or total communication which utilizes a combination of manual communication with the oral method (Freeman, 1976). Such alternate methods are not always available in public schools, and consequently some of these children are deprived of educational services because of their inability to function in oral method classrooms ( Kohl, 1966). A behavioral approach in speechreading training suggests early disctimination training using differential reinforcement. Johnson and Kaye ( 1976) demonstrated that a deaf multihandicapped child could be taught through token or social reinforcement to discriminate the names of objects based on both voiced and nonvoiced stimuli. The present study sought to explore and extend behavioral training procedures useful with severely hearing-impaired children. Specifically, two multiply handicapped children were trained to speechread the names of three groups of objects. 'This research was supported. in part, by an ESEA Title VI-B grant awarded to the Kalamazoo Vallev Intermediate School District. Tames H. Kaye, Project Director. The authors wish to thadk Paul S. Wollam, ~uperintenhent, and ~ n n l e eDecent, Director of Special Education, as well as Gerald L. Shook, Coordinator of the Kalamazoo Valley Multihandicap Center for providing an environmenr which fosters the developmenr of applied research within the educational process. W e also wish to thank Jack Michael, Brian Iwata, and James Cowart of Western Michigan University for their editorial assistance and consultaoon. Reprints may be obrained from C: Merle Johnson. Department of Psychology, Central Michigan University, Mount Pleasant, Mlchigan 48859.
C. M. JOHNSON
172
&
J. H. KAYE
Subjects
Two hearing-impaired children enrolled in the Kalamazoo Valley Multihandicap Center served as subjects. Both children had failed to respond effectively to previous speechreading techniques. However, they could fingerspell and had a small repertoire of American Sign Language. Brian was a 12-yr.-old boy who was diagnosed as hearing impaired, speech impaired, and mentally retarded. His unaided hearing in the better ear averaged 55 dB (ISO, 1964) over the frequencies of 250 to 4000 Hz. His spontaneous speech was hypernasal and unintelligible due to a severe cleft lip and palate. H e resided at a nearby state institution for children and actended school weekdays at the Center. Lori was a 10-yr.-old girl who had been diagnosed as deaf, mentally retarded, and emotionally disrurbed. She evidenced a profound sensorineural hearing loss ranging from 90 dB to 110 dB (ISO, 1964) from 250 to 4000 Hz. Lori had entered the Center at age 8 and had exhibited hyperactive, destructive, and self-abusive behaviors. Apparatus and Setting
Training was carried out within a sound-attenuated booth measuring 1.5 X 1.8 m. A Panasonic closed circuit television camera with an 8.5-mm wideangle lens was located on one wall near the ceiling. A Sony F-98 microphone was suspended through the ceiling and out of reach. The experimenter and subject sat facing each other with a table located between them. At one end of the cable was a Gerbrands (Model B) token dispenser. Three sets of six objects were utilized as training stimuli and are represented in Table 1. In the first set TABLE 1 SETS OF OBJECTS ARRANGEDIN A 3
banana pear
2 HORIZONTAL ARRAY
Group B
Group A apple peach
x
orange plum
blue purple
green red
Group C orange yellow
1
2
4
5
3 6
were six plastic fruits, in the second six 5- X 5-cm cards with names of colors printed in their respective colors. The third set contained six 5- X 5-cm cards with numbers printed on them. Prior to baseline, a single training session was used to teach the subjects to point to objects whenever the experimenter vocalized an object name. Correctness of a response was not a factor for social reinforcement. Reinforcement was presented in the form of praise in sign language, e.g.. the sign for "good." The criterion for reinforcement was for the subject's hand or extended fingers to come in contact with one of the six objects within 5 sec. after the experimenter's vocalization.
BEHAVIORAL CONTRAST I N SPEECHREADlNG
173
A three-component multiple baseline (Baer, Wolf, & Risley, 1968) containing withdrawal phases (Hersen & Barlow, 1976) was obtained for each subject. The training objects of each set were placed, one set at a time, in a 3 x 2 horizontal array on the cable between the subject and experimenter. The presentation order was Set A, Set B, and then Set C. The child left the booth for a drink of water at the cqmpletion of each set at which time the experimenrer changed the objeccs on the table. Before each trial the experimenter tossed a die into a bowl held under the table. Each face of the die corresponded to one of the six objects in that set. After noting which face came up, the experimenter established eye contact with the subject and then vocalized the name of the corresponding o b j m . The objecr name was verbalized only once per trial and normal conversation levels were used at all times. The dependent variable was the subject's pointing to one of the six stimuli on the table afrer each vocal presentation. Pointing was defined as the subject's hand or extended f~ngerstouching one of the six stimuli. Each of the pointing responses was recorded as correct or incorrect. In the case of mulriple responses on the same trial, only the first response was counted. A correct response was defined as pointing to the appropriate object within 5 sec. following the vocalization of an object name by the experimenter. Incorrect responses were defined as incorrect matches between objects and object names or the absence of pointing responses within the 5-sec. limited hold. Each object was presented six times in a random fashion each session. Component responding was defined as the total number of correct responses by the subject to each of the three sets. The responses were tallied on a piece of paper placed out of the child's view. Throughout the experiment social interaction between the subject and experimenter was minimal. During reinforcement phases tokens were delivered for correcc responses via the token dispenser to limit social interaction that may have confounded the magnitude of reinforcement. Tokens were saved by the subjects and traded for candy or toys at the Center's store at the end of the day. Bu~e1ine.-During baseline the experimenter vocalized an object name and the subject pointed to one of the six stimuli in front of him. The experimenter simply recorded the subject's response, giving no indication whether or nor it was correct. Phase I.-The first manipulation was the contingent delivery of a token to the s u b jecc for each correct response emitted to Sers A and B. The experimenter activated the token dispenser for each correct response and then terminated eye contact with the subject by recording the data for that trial. Incorrect responses to Sers A and B, as well as all responses to Set C, were not reinforced. Phase 11.-During this phase correct responses to Set A were reinforced with tokens while incorrect responses to Set A were not reinforced. Neither correct nor incorrect responses to Sets B and C were reinforced. Phase Ill.-Correct responses to Sets A and B were reinforced again with tokens. Incorrect responses to Sets A and B and both correct and incorrect responses to Ser C were not reinforced. Phase 1V.-The reinforcement contingency applied to correct responding to Set B was changed from reinforcement to extinction a second time. This phase was reinstated for Lori to determine if the effects of Phase I1 could be replicated. P h a ~ eV.-Reinforcement was applied for correct responding to Sets A and B for a third time. Phase V1.-Correct responses to all three sets were reinforced with tokens and no incorrect responses were reinforced. This manipulation was made to determine the effect of reinforcing correcc responses to Set C which had previously been in baseline conditions. Phase VI1.-In this phase all 1 8 objects from the three sets were placed on the table
C. M. JOHNSON
&
J. H. KAYE
FIG. 1. The number of correct speechreading responses emitted by Brian to each set of stimuli during each session in a 6 x 3 array. Any object could have been presented on any given trial although the data were recorded in three separate components as before. A second die was added to randomize sets. All correct responses were reinforced with tokens while incorrect responses were not reinforced. Phase VII1.-The final manipulation was carried out without any of the two or three-dimensional objects used previously. At the start of the first session of this phase, Lori was given instructions in American Sign Language to sign the name of the object instead of pointing. N o further instructions were provided and the table was cleared except for the token dispenser. The contingency in this phase required the subject to sign correctly the name of the object wirhin 5 sec. following the vocalization of an object name. Correct responses were reinforced with tokens and incorrect responses were not.
Reliability Interobserver reliability checks were taken about every nine sessions. The observer watched the entire session through a closed circuit television monitor' in a separate room and recorded data in the same manner as the experimenter. A trial-by-trial comparison was used and reliability scores were computed by dividing the number of agreements by the total number of agreements plus disagreements. Four interobserver reliability checks were taken by three different observers during Brian's experiment. Reliability ranged from 96% to 100% with a mean of 97 5% Twenty reliability checks were taken by 11 different
BEHAVIORAL CONTRAST I N SPEECHREADING
a
I
. m
a
9
D
NUMBER OF CORRECT RESPONSES
m
176
C. M. JOHNSON & J. H. KAYE
observers during Lori's experiment. Reliability during this experiment ranged from 94% to 100% with a mean of 98.9%. RESULTS Fig. 1 depicts the levels of correct responding to each set by Brian. Brian exhibited a falling baseline to Set A and stable baselines to Sets B and C. In Phase I, Brian demonstrated a rapid acquisition of speechreading the objects in Set B and on session 29 he achieved 100% correct for that set. Brian showed a slight increase from the end of baseline for Set A in Phase I, but he did not exceed the correct responding level that he had exhibited during the first few days of baseline. However, when the contingencies for correct responding to Set B were reversed from reinforcement to extinction in Phase 11, correct responding to Set A immediately increased and continued to rise until the end of the phase. Correct responding to Set B showed a slight decrement initially in Phase I1 but remrned to previous levels by the end of the phase. Correct responding to Set C remained stable throughout the experiment. Unfortunately, at this point Brian was transferred to another state institution, necessitating an early and unforeseen termination of his participation in the experiment. Fig. 2 portrays the mean number of correct responses by Lori. Her baseline levels were generally lower than Brian's and she exhibited a falling baseline to Set A. During Phase I correct responding to Set A increased rapidly and then stabilized at about 16 correct responses per session. Correct responding to Set B made slower increments during Phase I but stabilized at about 19 correct responses per session. Responding to Set C remained stable during the first phase. When the contingencies for correct responding to Set B went from reinforcement to extinction in Phase 11, correct responding to Set A immediately increased from about 17 in Phase I to about 23 in Phase 11, even though there was no contingency change for correct responding to Set A. Correct responding to Set B showed a slight decrement when going from reinforcement to extinction in Phase 11. This behavioral interaction was similar to that shown by Brian during the same manipulation. In addition, correct responding to Set C remained stable. In Phase 111, correct responding to Set B increased and later stabilized at about the same level as correct responding to Set A. Throughout Phase 111 correct responding to both Sets A and C remained stable. When reinforcement was withheld for all responses to Set B in Phase IV, the effects of the first reversal in Phase I1 were replicated. Correct responding to Set A immediately increased from about 23 in Phase I11 to about 27 and continued to rise to about 30 by the end of the phase. Correct responding to both Sets B and C remained stable during Phase IV. In Phase V correct responding to Sets B and C remained unchanged and correct responding to Set A became slightly more variable. In other phases similar to this one, correct responding to Set B had increased.
BEHAVIORAL CONTRAST IN SPEECHREADING
177
However, the data showed that Lori was consistently missing red and green in that set, resulting in no improvement. When all three components were reinforced in Phase VI, correct responding increased in each set. Component responding to Set A increased first and attained 100% correct during Session 137. Next, component responding to Set B increased and attained 100% correct during Session 146. However, it should be noted that during the days of Sessions 135 and 136 a substitute therapist had worked with Lori on identifying colors outside the experimental setting and one of those colors was red. The data from previous sessions showed that Lori had consistently missed both red and green for over 20 sessions. The effects of this mishap are not completely known, but it may have accounted for some of the increase in correct responding to Set B. Regardless, a rapid increment in correct responding to Set C ensued and all three components remained near the 100% correct level until the end of the phase. When all three groups of objects were presented simultaneously in Phase VII, Lori exhibited a slight decrement in correct responding to each set. However, by the end of the phase she recovered most of the loss. When all of the objects were removed in the last phase, Lori apparently had no difficulty with the response generalization since she maintained component responding near the 100% correct level.
DISCUSSION The multiple baseline containing withdrawal phases indicated that in no instance did the number of correct responses to any given set of stimuli improve during a given phase unless tokens were presented contingent upon correct responses. Furthermore, the obvious discrepancies in the number of correct responses between the reinforced sets ( A and B) and the nonreinforced set (C, up to Phase VI for Lori) for both subjects also demonstrated the effectiveness of token reinforcement. Interestingly, substantial changes in correctness were obtained using tokens in relative isolation; great care was taken to eliminate or greatly reduce the likelihood of social reinforcement during experimental sessions, and no primary reinforcement was employed in order to control for spurious or momentary changes in reinforcement magnitude or quality. It would seem likely that even more dramatic and rapid gains would result from combining social, primary, and token reinforcement in a more traditional therapeutic-educational setting (Johnson & Kaye, 1976). The low, steady rate of correct responding to Set C for both subjects until tokens became available for such responding suggests: ( a ) limited impact of possible maturational variables, ( b ) little or no practice effect, and ( c ) little or no generalization to Set C from correct responding to Sets A and B. On the other hand, the rapid acquisition of Set C correct responding for Lori in Phase
178
C. M. JOHNSON
&
J. H. KAYE
VI, as well as the slightly less dramatic improvement in responding to Set B in that phase, may have resulted from the development of a learning set (Harlow, 1949; Kimble, 1961) during correct response acquisition to Set A. Lori's previous learning history with Set A may have made her acquisition of the behaviors related to Set C easier and possibly to Set B as well. However, an error analysis of Lori's responding provides a more reasonable account of the initial sharp rise in correct responding to Set B in Phase VI. Errors occurring to Sets A and C were observed to be random. The additional uncontrolled training Lori received for red in Set B probably was primarily responsible for the initial rise in her correct responding to colors in this phase, although subsequent gains could be related to the transfer of training through learning sets. A between-sessions analysis showed that those gains came just after correct responding to Set A reached asymptote. Similarly, the dramatic gains in correct responding to Set C came just after correct responding to Set B reached asymptote. Such gains would be predicted by the facilitative effect of learning sets. When correct responding to Set B changed from reinforcement to extinction, correct responding to Set A increased above previously established levels even though there was no contingency change for responding to this set. This interaction between responding to Sets A and B appeared to be similar to the phenomena Reynolds ( 1961b) defined as positive behavioral contrast. Reynolds defined behavioral contrast as a change in the rate of responding during one procedurally constant component of a multiple schedule in a direction away from the rate of responding generated by a procedural manipulation during a second component. When the rate change in the constant component is an increase, it is called positive contrast; when it is a decrease, it is called negative contrast (Reynolds, 1961a). Contrast effects have been previously demonstrated with humans. O'Brien (1968) demonstrated behavioral contrast with two mentally retarded girls pushing buttons on a multiple variable-interval extinction schedule of reinforcement. Waite and Osborne (1972) reported behavioral contrast with school children exposed to similar schedules. Fagen ( 1979) found that 3- to 4mo.-old infants exhibited positive behavioral contrast during footkick conditioning when switched from multiple CRF-CRF to multiple CRF-EXT schedules of reinforcement-similar to the present study. Taylor ( 1970) and Couture ( 1976) have demonstrated behavioral contrast with college students key pressing on multiple VI EXT schedules of reinforcement. The contrast in O'Brien's ( 1968) experiment diminished and is called transient behavioral contrast whereas in Waite and Osborne's (1972) experiment the contrast did not disappear and is referred to as sustained behavioral contrast (Nevin & Shettleworth, 1966). The contrast in the present study was sustained by both subjects throughout the phases in which it appeared. In addition, this increase in correct responding
BEHAVIORAL CONTRAST IN SPEECHREADING
179
levels to Set A during Phase I1 was replicated in Phase IV in Lori's experiment. However, the modest decrease in correct responding to Set B during Phase I1 was not present in Phase IV. Behavioral contrast, demonstrated by an increase in responding in one procedurally constant component, does not necessitate a decrease in responding in the changed component (Halliday & Boakes, 1974; Wilkie, 1971; Fagen, 1979). Since the subjects were learning a new skill, it is not surprising that component responding in Set B did not decrease when placed on extinction, as would be expected in most contrast situations. Once the children had learned to speechread certain items, they continued to respond correctly without token reinforcement and this irreversibility of behavior resulted in therapeutic gain. The present study differed from other studies finding behavioral contrast with human subjects because previous experiments were concerned with measures of performance such as key pressing. Behavioral contrast was evidenced in this experiment while both subjects were learning a complex task such as speechreading. It appears as if schedule interactions such as behavioral contrast may be important factors in applied research such as the present experiment and that applied researchers should be aware of such phenomenaZ (Sulzer-Azaroff & Mayer, 1977). For instance, Lovaas and Simmons (1969) noted that self-destructive behavior by an autistic child increased with non-punishing adults after decreasing with a punishing adult, even though there was no contingency change with the non-punishing adult. The increase in correct responding to Set A during contrast periods demonstrated that such contrast effects can be conducive to learning. Since the contrast training procedure seemed to be effective in increasing behavior, some consideration should be given to its possible utility. Therapeutic gain more closely aligned with successful language functioning was demonstrated by Lori in the last rwo phases, wherein she transferred responding from pointing to objects to signing the name of the objects in sign language, a more common and functional set of responses for the hearing impaired. At the completion of this experiment, generalization procedures to other people and further discrimination training to other words were implemented, resulting in even more functional skills for Lori. REFERENCES BAER,D. M., WOLF,M. M., & RISLEY,T. R. Some current dimensions of applied behavior analysis. loitrnal o f Applied Behavior Analysis, 1968, 1 , 91-97. BERGER, K. W . Speechreding. Baltimore: National Educational Press, 1972. COUTURE, E. T. Behavioral contrast: a human replication. Unpublished Master's thesis. State Univer. of New York at Geneseo, 1976.
7. S. Bailey,
A handbook of research methods in applied behavior analysis. lished book manuscript, 1977)
(Unpub-
'
180
C. M. JOHNSON
& J. H . KAYE
FAGEN,J. W. Behavioral contrast in infants. Infant Behavior and Development, 1979, 2, 101-112. FARWELL,R. M. Speechreading: a research review. American Annals o f the Deaf, 1976, 121, 19-30. FREEMAN,R. D. T h e deaf child: controversy over teaching methods. Jorrrnal o f Child Psychology and Psychiatry, 1976, 17, 229-232. HALLIDAY,M. S., & BOAKES,R. A. Behavioral contrast without response-rate reduction. Journal o f the Experimental Analysis o f Behavior, 1974, 22, 453-462. HARLOW,H. F. T h e formation of learning sets. Psychological Review, 1949, 56, 5 1-65. HERSEN,M., & BARLOW,D. H. Single-case experimental designs: strategies for studying behavior change. New York: Pergamon, 1976. JOHNSON, C. M., & KAYE,J. H. Acquisition of lipreading in a deaf multihandicapped child. Journal o f Speech and Hearing Disorders, 1976, 41, 226-232. KIMBLE,G. A. Hilgard and Marquis' 'Conditioning and learning.' New York: Appleton-Century-Crofts, 1961. KOHL, H. R. Language and education for the deaf. New York: Center for Urban Education, 1966. LOVAAS,0 . I., & SIMMONS. J. Q. Manipulation of self-destruction in three retarded children. Journal o f Applied Behavior Analysis, 1969, 2 , 143-157. MYKLEDUST,H. R. T h e psychology o f deafness, sensory deprivation, learning and adjustment. New York: Grune & Stratton, 1960. NEVIN, J. A., & SHETTLEWORTH,S. J. An analysis of conrrast effects in multiple schedules. Jownal of the Experimental Analysis o f Behavior, 1966, 9, 305-315. O'BRIHN, F. Sequential contrast effects with human subjects. journal o f the Experimental Analysis o f Behavior, 1968, 11, 537-542. REYNOLDS,G. S. An analysis of interactions in a multiple schedule. Journal o f the Experimental Analysis of Behavior, 1961, 4, 107-117. ( a ) REYNOLDS,G. S. Behavioral contrast. Journal o f the Experimental A ~ a l y s i s o f Behavior, 1961, 4, 57-71. ( b ) SULZER-AZAROFF,B., & MAYER, G. R. Applying behavior-analysis procedures with children a d youth. New York: Holt, Rinehart & Winston, 1977. TAYLOR,R . W. Behavioral contrast in human subjects. Unpublished doctoral dissertation, Indiana Univer., 1970. WAITE, W . W . , & OSBORNE,J. G. Sustained behavioral contrast in children. Journal of the Experimental Analysis o f Behavior, 1972, 18, 113-117. W I L K I E ,D. M. Delayed reinforcement in a multiple schedule. Journal of the Experimental Analysis o f Behavior, 1971, 16, 233-239.
Accepted June 19, 1979.
