Journal of Autism and Developmental Disorders, Vol. 9, No. 4, 1979. The Effects of Schedule of Reinforcement on. Stimulus Overselectivity in Autistic Children 1.
Journal of Autism and Developmental Disorders, Vol. 9, No. 4, 1979
The Effects of Schedule of Reinforcement on Stimulus Overselectivity in Autistic Children 1 Robert L. KoegeP University of California at Santa Barbara Laura Schreibman Claremont Men "sCollege Karen Britten University of Kansas Richard Laitinen University of California at Santa Barbara
Recent research demonstrated that when autistic children are presented a discrimination task with multiple cues, they typically respond to an abnormally limited number, usually one, o f the available cues. This phenomenon, termed "'stimulus overselectivity, "" has been implicated as a possible basis f o r many o f the behavioral deficits characteristic o f autism. The present investigation was conducted to systematically analyze the effects o f changing the schedules o f reinforcement during discrimination training on subsequent stimulus overselectivity. Twelve autistic children were taught a discrimination involving multiple visual cues, with a C R F schedule o f reinJorcement. The children were then overtrained on either the same (CRF) schedule or on a partial (VR:3) reinforcement schedule. Subsequent overselectivity on single-cue test trials was then assessed. Results suggested that significantly less overselectivity occurred when the children were presented with the VR:3 reinforcement schedule during
'This research was supported by USPHS Research Grants MH 28231 and MH 28210 from the National Institute of Mental Health and by U.S. Office of Education Research Grant G007802084 from the Bureau for the Education of the Handicapped. The authors wish to thank John Burke, Marjorie Charlop, and Julie Williams for their assistance in this research. 2Address all correspondence to Dr. Robert L. Koegel, Social Process Research Institute, University of California, Santa Barbara, California 93106. 383 0 1 6 2 - 3 2 5 7 / 7 9 / 1 2 0 0 - 0 3 8 3 5 0 3 . 0 0 ] 0 9 1979 Plenum Publishing Corporation
384
Koegel, Schreibman, Britten, and Laitinen
overtraining. These results are discussed in terms of variables influencing overselectivity and in terms of implications for designing treatment procedures for autistic children. In the past 15 years research in behavior modification has made considerable contributions toward improving the behavior of autistic children (cf. reviews by Egel, Koegel, & Schreibman, in press; Koegel, Egel, & Dunlap, in press; Lovaas & Newsom, 1976; Rincover & Koegel, 1977; etc,). In an effort to attack the syndrome more globally, interest has increased in the role of antecedent stimuli (as compared to reinforcement stimuli) for modifying autistic behavior (cf. Koegel & Schreibman, 1974; Schreibman & Koegel, in press). This interest stems in part from the unusual type of responding autistic children make to their physical environment (e.g., Ornitz & Ritvo, 1968; Schopler, 1965). As early as 1943, Kanner noted that autistic children "tuned out," or are "in their own world." He and others (cf. Koegel & Schreibman, 1976) also noted that the children are frequently suspected of (unconfirmed) blindness or deafness. However it is apparent that the unresponsiveness of these children is highly variable. For example, a child may fail to respond to a loud sound, such as a door slamming, but may react excessively (e.g., by covering his ears) to the sound of a newspaper rustling. This variability in responding led us to focus on the role of antecedent stimuli because we wanted to (a) relate the manner in which these children respond to environmental stimuli to abnormalities in their learning, and (b) then modify their response to environmental stimuli in a manner that would make their learning more similar to that of normal children. When presented with simultaneous multiple cues, normal children typically respond on the basis of a limited number of the available cues (e.g., Eimas, 1969; Hale & Morgan, 1973; Hale & Piper, 1973; Ross, 1976; Trabasso & Bower, 1968). In addition, the number of simultaneous cues to which normal children respond generally increases with age. This allows for the more advanced levels of learning characteristic of developmental trends. Whereas normal children respond to an increasing number of cues as they mature, there is a substantial body of literature suggesting that when autistic children are presented a discrimination involving complex stimuli, they selectively respond, regardless of age, on the basis of fewer cues than normal (frequently only one). This phenomenon has been labeled "stimulus overselectivity" (Lovaas, Schreibman, Koegel, & Rehm, 1971) and has been implicated as possibly responsible for many of the behavioral characteristics of autism (Lovaas, Koegel, & Schreibman, in press; Schreibman & Koegel, in press; Schreibman, Koegel, & Craig, 1977). In the area of social behavior, it is often reported that autistic children fail to establish normal relationships with other people (cf. Kanner, 1943; Rutter, 1978). We have observed that these children learn to recognize people by very limited cues,
Effects of Schedule of Reinforcement
385
and that they have difficulty retaining a stable recognition of another person if those cues are changed (Schreibman & Lovaas, 1973). For example, a child might fail to recognize a parent if the parent's glasses are removed. A second way overselectivity has been related to autistic children's abnormalities has been through the finding that newly trained behaviors often fail to generalize to new therapists or environments (cf. Rincover & Koegel, 1975). In many cases the children " h o o k " on an unreliable cue, such as a therapist's idiosyncratic hand movements, and fail to learn about the most relevant cues for solving the task. Naturally, since most other people would not engage in the same idiosyncratic gestures, the child appears to fail to generalize. It has also been noted that many autistic children do not seem to benefit from observational learning (Varni, Lovaas, Koegel, & Everett, 1979). They frequently do not imitate and they acquire only small parts of new behaviors from observing other children or adults. In a fourth area, it has been noted that autistic children typically have difficulty learning with traditional teaching methods such as the use of additional guiding stimuli for prompts. While extra stimulus prompts are somewhat effective with the average child, many investigators anecdotally observed or empirically determined (Koegel, 1971; Koegel & Rincover, 1976; Rincover, 1978; Schreibman, 1975; Sidman & Stoddard, 1966) that autistic children do not readily learn from extra stimulus prompts. Finally, the profound deficits in language so characteristic of autism may relate to the way they respond to environmental input (Lovaas et al., 1971). To systematically determine how the autistic child's unusual overselective response to the environment may affect the problem areas mentioned above, we embarked on a line of research to investigate exactly how their unresponsiveness affects learning. In doing so, we utilized a discrimination learning paradigm, which permits the investigator to relate aspects of manipulated stimulus input directly to the child's behavior. For reviews of this type of research, the reader is referred to Fellows (1968), MacKintosh (1975), Sutherland and MacKintosh (1971), and Trabasso and Bower (1968). In the first study demonstrating this phenomenon (Lovaas et al., 1971), three groups of children (autistic, retarded, and normal) were taught to press a bar in the presence of the complex stimulus, and to withhold bar pressing in the absence of that stimulus. The complex stimulus involved the simultaneous presentation of a visual, a tactile, and an auditory cue. Once this discrimination was learned, the elements of the stimulus complex were presented individually during test trials to ascertain which components of the complex had acquired control over the children's behavior. When the components were presented individually, the autistic children typically responded to only one of the cues (e.g., visual) as opposed to the other two cues (e.g., auditory or tactile). In contrast, the normal children responded equally to all of the cues when individually presented. The retarded children responded at a level between these two extremes.
386
Koegel, Schreibman, Britten, and Laitinen
To determine if such overselectivity was a function of the simultaneous presentation of cues in different sensory modalities or a function of the multiple cues, Koegel and Wilhelm (1973) presented normal and autistic children with a complex discrimination containing only visual cues. The children were trained to consistently respond to one of two cards, each containing two visual cues (e.g., a horse and a tree vs. a chair and a lamp). They were then tested on the single cues (e.g., horse vs. lamp) to assess whether one or both components had acquired control over their responding. The normal children showed clear evidence of control by b o t h cue components of the card. On the other hand, the autistic children typically responded to only one of the two cue components. These data, along with other studies, suggest that autistic children also tend to be somewhat overselective in their responding even when multiple stimuli are presented within the same modality (Lovaas & Schreibman, 1971; Rincover & Koegel, 1975; Koegel & Rincover, 1976; Schreibman & Lovaas, 1973; Reynolds, Newsom, & Lovaas, 1974; Schreibman, 1975).
MODIFICATION OF OVERSELECTIVITY Although many studies indicate that autistic children are overselective, there is also some evidence to suggest that overselectivity can be modified (Koegel & Schreibman, 1977). One of the more optimistic lines of research involved several studies (Lovaas et al., 1971; Lovaas & Schreibman, 1971; Schreibman et al., 1977) that suggested that the amount of overselectivity might be rapidly reduced with repeated exposure to testing, that is, by alternating trials involving single components of the complex stimulus with trials involving the entire complex stimulus. The research of Schreibman et al. (1977) involved training the child on a discrimination task (similar to those employed in Koegel & Wilhelm, 1973) with a complex stimulus composed of two visual cues. Then, during the test phase of the experiment, probe trials with only single cue components were interspersed along with trials of the cue complex. In spite of the lack of reinforcement during the test trials, the amount of overselectivity decreased with continued exposure to the testing situation. Examination of the results suggested that the introduction of different schedules of reinforcement (i.e., giving test trials without reinforcement) during the testing procedure might have influenced the overselectivity reduction. This seemed likely since other research has shown that changing a reinforcement schedule broadens responding to multiple cues (cf. Trabasso & Bower, 1968). The present study systematically investigated the effects of changing the schedules of reinforcement during training on subsequent overselec-
387
Effects of Schedule of Reinforcement
tivity. If overselectivity is decreased by presenting nonreinforced (correct) trials, then changing the reinforcement schedule from continuous to partial prior to testing should result in little, if any, overselectivity during subsequent testing.
METHOD
Subjects Twelve children, eight boys and four girls, participated in this study. Their mean chronological age was 8 years 9 months (range 4 years 2 months to 15 years) and mean mental age was 3 years 4 months (as estimated by the Merrill-Palmer and Peabody Picture Vocabulary Test). One child was considered to be formally untestable with all standardized testing procedures. Five of the children did not speak and had limited receptive language skills consisting of responses to simple commands such as "sit d o w n . " Seven of the children were echolalic, repeating words or phrases without obvious communicative intent. All were diagnosed as autistic, and exhibited high levels o f self-stimulation and tantrums. The children had histories of variable responsiveness to the external environment to the extent that they were referred for testing for sensory impairment. The tests indicated no impairment.
Design The hypothesis proposed was that the presence of nonreinforced correct responses during training would produce a decrease in subsequent overselectivity. This was assessed as follows. Under one condition the children were trained on a discrimination task with a reinforcement schedule in which each correct response was reinforced (CRF). After criterion was reached, 100 overtraining trials were provided where, on the average, every third response was reinforced (VR:3). In the other condition, no change was made in reinforcement during training or the subsequent 100 trials of overtraining (i.e., the schedule remained CRF throughout). All children received training in both conditions, with the order of conditions randomly alternated across children. The prediction was that for one condition (CRF throughout training and overtraining) there should be considerable overselectivity evidenced at the start of testing since the reinforcement schedule remained constant up to that point, and that for the other set of stimuli (change from CRF during
388
Koegel, Schreibman, Britten, and Laitinen
training to VR:3 during overtraining) there should be very little overselectivity at the start of testing since the reinforcement schedule had been changed during training.
Pretraining Stimuli The children were pretrained to respond (by pointing) to one of two 10.2 • 15.2-cm white cards. Each card contained a cue complex composed of two 3.8 x 3.8-cm visual stimuli. The upper portion of Figure 1 illustrates an example of training stimuli. In this example, the two stimuli constituting the S + complex were a squirrel and a bluejay. The two stimuli constituting the S - complex were a flower and a butterfly. To ensure that the results of this study were not a function of a particular set of stimuli, a total of 31 different cue complexes were utilized for the different children. The exact cue complexes and conditions for each child are presented in Table I.
Pretraining Procedures The child was seated facing the experimenter across a 76.2 x 152.4-cm table. Trials were presented only when the child was seated quietly, displaying good eye contact and not engaged in off-task behaviors. The experimenter began a trial by simultaneously placing the two training cards (S + complex and S - complex) on the table in front of the child. The experimenter then instructed the child to " p o i n t to the correct c a r d . " If the child did not respond within 5 seconds of the instruction, the experimenter manually placed the child's hand near the S + card and rewarded the response with food (e.g., raisin or peanut) and verbal praise ( " g o o d " ) . After the child responded to the S + card on two consecutive trials without a prompt, the experimenter no longer prompted the response. Trials were presented approximately every 5 to 10 seconds. The position of the two cards was alternated according to a randomly selected Gellerman alternation order (Gellerman, 1933) so that the S + complex was on the right as often as on the left. During pretraining, correct responses were reinforced on a CRF schedule with food and praise. Incorrect responses or failures to respond were followed by a verbal " n o " and immediate removal of the stimulus materials. Training continued until the child responded correctly on 10 consecutive trials. Pretraining sessions were distributed such that no more than two sessions were conducted in 1 day and no fewer than one every 3 days. Each session lasted between 10 and 25 minutes. After children reached criterion, they progressed to one of the two overtraining schedules described below. All of the children were exposed to both overtraining schedules for different sets of stimuli, with order counterbalanced across children.
Effects of Schedule of Reinforcement
389
Fig. 1. The upper portion of the figure presents an example of a set of training stimuli used in this study. The lower portion of the figure shows examples of two of the probe trials used to
assess response to componentcues.
Overtraining with VR:3 versus C R F Schedules After reaching criterion (10 consecutive correct trials during pretraining), the child was presented with 100 additional trials with either a CRF or a VR:3 schedule of reinforcement. The sessions were distributed in the same manner as in pretraining. During the CRF condition, everything continued the same as during the pretraining. During the VR:3 condition, reinforcement was available on a schedule where, on the average, one out o f every
S+ cue complex
Butterfly and rose Squirrel and turtle Skunk and leaf Monkey and pig Tree and bug Rose and monkey Skunk and whale Elephant and turtle Orange flower and dog Pumpkin and turtle Pumpkin and turtle Pink-and-yellow flower and orange flower
Child
1 2 3 4 5 6 7 8 9 10 11 12
Blackbird and flower Butterfly and lion Man and dog Racoon and butterfly Flower and bird Pumpkin and bird Elephant and cement mixer Whale and bear Skunk and monkey Elephant and lion Elephant and lion Red flower and blue flower
S - cue complex
CRF + CRF condition
Tree and bug, Skunk and whale Tree and bug Whale and turtle Skunk and leaf Squirrel and bluejay Monkey and bear Monkey and rose Bluebird and dots Monkey and rose Monkey and rose Tree and bug
S+ cue complex
Bluebird and dots Elephant and truck Flower and bird Truck and lion Man and dog Flower and butterfly Pig and turtle Bluejay and skunk Whale and teddy bear Ladybug and banana Ladybug and banana Flower and bird
S - cue complex
CRF + VR condition
Table I. The Stimulus Materials for Each Condition for Every Child in the Experiment
g
Effects of Schedule of Reinforcement
391
three correct responses was reinforced. No consequences followed the nonreinforced correct trials. Incorrect responses were followed by a verbal " n o " and removal of the stimulus materials. As mentioned above, this partial reinforcement condition was instituted to determine whether the amount of control exerted by a given component cue would change as a function of not receiving continuous reinforcement for correct responding on these trials. Test Stimuli Once each child had acquired the pretraining discrimination and had been exposed to the various types of overtraining trials, he was then tested on the individual components of the cue complex in order to assess the control exerted by each of the two components over the child's response. Each component was presented individually, centered on a 10.2 • 15.2-cm white card. On these test trials, the child was always presented with a single S + component versus a single S - component. The lower portion of Figure 1 illustrates the stimuli for two test trials for the pretraining stimuli shown above them. Test Procedures During testing, the children were presented with two types of trials: (1) trials with the cue complex (continued training trials), and (2) trials with the individual components of the cue complex (test trials). Trials with the cue complex were identical to trials in the training and overtraining phases and were presented to maintain the original discrimination. Single cue test trials were presented to assess the amount of control exerted by the individual components over the child's response. The ratio of test to training trials was 2:1. The child was never reinforced for responses on test trials. Reinforcement was, however, continued for training trials. Thus, during this testing phase, reinforcement was available on the average of one out of every three trials (VR:3). If the child made an error on an interspersed training trial, retraining occurred to a criterion of five consecutive correct responses before testing resumed. The purpose o f the test trials was to provide an indication of the amount of stimulus overselectivity, i.e., the amount of differential control exerted by each of the single components of the cue complex as a result of the different types of training. Testing was terminated after 16 probe trials (8 with each S + component cue) in order to reduce the influence of nonreinforced trials during testing per se on overselectivity.
392
Koegel, Schreibman, Britten, and Laitinen
RESULTS Table II summarizes the results for all 12 children. The percent correct responding to each component cue, and the overselectivity scores are presented for the two conditions (CRF + CRF vs. CRF + VR training) for each child. The overselectivity score was calculated by determining the percent correct response to each of the two S + stimuli and subtracting the smaller of the two percentages from the larger. Thus, for example, if a child responded correctly to Cue 1 10007o of the time and Cue 2 75~ of the time, the overselectivity score would be 25 ( 1 0 0 - 7 5 ) . For those stimulus sets in which the children were trained on a CRF schedule throughout, 10 o f the 12 children showed some degree of overselectivity, and 6 of the 12 showed a difference of two or more correct responses (25~ or above) between the two component cues (see fourth column of Table II). The average overselectivity score for the 12 children was 24 (out of a theoretical maximum of 50). That is, consistent with the results of previous experiments, these children showed considerable overselectivity under the CRF condition. In contrast, consider the results for the stimulus sets that were used when the children were initially trained with a CRF schedule and then shifted to VR:3 during overtraining. During the subsequent test trials the children showed very little overselectivity (regardless of whether this occurred before or after a CRF + CRF condition). Five children showed no overselectivity at all (seventh column), and only 1 child (child 7) showed a difference of more than one correct response (over 12.5~ between the two component cues. The average overselectivity score for the VR condition was only 11.5. Most importantly, when compared with themselves, 7 of the 12 children exhibited higher
T a b l e II. O v e r s e l e c t i v i t y S c o r e s f o r E a c h C o n d i t i o n f o r E v e r y C h i l d in t h e E x p e r i m e n t ( R e s p o n s e t o C u e 1 C u e 2) Child
CRF + CRF condition
CRF + VR condition
1 2 3 4 5 6 7 8 9 10 11 12
100%-75% = 100%-87.5% = 87.5%-25% = 100%-87.5% = 100%-75% = 100%-87.5%= 87.5%-50% = 100%-50% = 87.5%-50% = 100%-87.5% = 100%- 100% = 100%-100% =
100%-87.5% 100%-87.5% 87.5%-75% 100%-100% 100%-100% 100%-100% 100%-37.5% 100%-87.5% 100%-100% 100%-87.5% 100%-87.5% 100%-100%
25 12.5 62.5 12.5 25 12.5 37.5 50 37.5 12.5 0 0
Average = 24.0 a aRounded
to the nearest tenth.
= = = = = = = = = = = =
A v e r a g e = 11.5 a
12.5 12.5 12.5 0 0 0 62.5 12.5 0 12.5 12.5 0
393
Effects of Schedule of Reinforcement
overselectivity in testing after only CRF overtraining. In contrast, only 2 of the 12 showed higher overselectivity scores following VR overtraining. Given that one would expect some chance correct responding in both conditions (since the child had a 50% chance of being correct by guessing on any individual trial), the above differences were tested for statistical significance. First, the difference between the CRF versus VR conditions was tested with respect to the number of children who showed overselectivity scores of one or less correct response difference for the two component cues. This was significant (p < .05) when tested with the Cochran Q Test. Second, the difference between the means of the two sets of overselectivity scores (for the VR vs. CRF overtraining conditions) was tested with the t test for paired observations. The results showed a significant difference (t~ = 1.97, p < .038, one-tailed test).
DISCUSSION The results of this investigation can be discussed in relation both to the understanding of variables that influence overselectivity per se and to the development of procedures that might improve the treatment of autism.
Variables Influencing Overselectivity Trabasso and Bower (1968) postulated that when organisms make errors in discrimination learning, their breadth of attention on subsequent trials is increased. That is, the number of cues they respond to increases. It is possible that in the present experiment the change to nonreinforced trials (in the VR overtraining condition) may have had a similar effect. That is, a child may have responded to the lack of reinforcement as indicative of an error and, therefore, increased responding to multiple cues. This interpretation is consistent with our earlier finding (Schreibman et al., 1977) that nonreinforced probe trials during testing produced an increase in the number of cues to which autistic children responded. It is interesting to consider the possible influence of overtraining per se on overselectivity. There are data (e.g., Sutherland & Holgate, 1966; Sutherland & MacKintosh, 1971) to suggest that infrahuman animals learn more about redundant cues during overtraining, and their selective responding is reduced. Others (e.g., Trabasso & Bower, 1968) report that overtraining of humans typically does not lead to increased responding to redundant cues. However, Schover and Newsom (1976) have reported a possible reduction in overselectivity with overtraining of autistic subjects. In their study, the children were given considerable exposure to nonreinforced trials both during an initial test phase (with nonreinforced probes with the com-
394
Koegel, Schreibman, Brinen, and Laitinen
ponent cues) and with exposure to numerous overtraining trials with a partial reinforcement (VR:3) schedule. These results and those of the present research suggest that overtraining will have its greatest effect on reducing overselectivity if it involves a partial reinforcement schedule, and may have little influence if a CRF schedule is used throughout training.
Development of New Treatment Techniques The results indicated that overselectivity could be significantly reduced when a partial reinforcement schedule was used during part of the training as compared to when a CRF schedule was used throughout. Thus a relatively simple procedure can be an effective treatment technique when one expects that overselectivity might become a problem (e.g., in the areas of prompting or generalization). Interestingly, anecdotal comments from our own clinics, and illustrations from the experimental literature, suggest that when partial reinforcement is a part of training, there has been a reduction in overselectivity problems. One of the advantages of a research approach to the development of new treatment techniques is that research is cumulative. It is now easy to reexamine previous reports (which may have been difficult to understand earlier) in the light of new findings. If one does this with the autism literature, the present results are particularly intriguing. For example, we have noted that one of the classic problems associated with stimulus overselectivity is that autistic children frequently " h o o k " on prompt stimuli and fail to learn much about new stimuli being trained with these prompts. However, there have also been several dramatic successes reported in the literature, in which autistic children were taught successfully with the use of prompts. In our own clinics as well as in the literature, investigators have reported being able to establish functional speech in echolalic children by using prompts (cf. Lovaas, 1977; Lovaas, Schreibman, & Koegel, 1974; Risley & Wolf, 1967). At first, these findings seemed contradictory. However, a comparison of the treatment procedures with the present experimental manipulations sheds some light on the issue. For example, the behavior modification treatment of echolalia frequently involves using the child's echolalia as a prompt. The therapist may say, "What is this?" (while holding up a toy train) and then prompt the child by saying the word train, which the child repeats. Frequently, the therapist probes to see if the child will respond without a prompt. Thus on some trials the therapist merely says, "What is this?" and waits for the child to say "train." If no response is forthcoming, the therapist may again provide a prompt. What is interesting to note about this procedure is that it is typical for the therapist to provide primary reinforcers only for the unprompted trials, and to withhold
Effects of Schedule of Reinforcement
395
or delay reinforcers for prompted trials. In view of our present results, we would expect such a partial reinforcement procedure to broaden the child's response to the multiple cues (of the prompt and the new training stimuli), and thus it would actually be puzzling if the child "hooked" on the prompt. Stimulus generalization is another area in which overselectivity has been a problem. However, there have been some successes reported when therapists used partial reinforcement. For example, Koegel and Rincover (1977) found improved responding in extratherapy environments when partial reinforcement training procedures were employed. As with the prompt studies described above, it is not possible to relate definitely the improvement to a change in overselectivity. However, the major implication of the present study is that such results should be possible. That is, partial reinforcement should reduce overselectivity, thereby improving learning related to generalization. The consistency in the literature is quite encouraging with respect to our major objective: to help normalize the learning of autistic children. This area of research appears quite promising with respect to the possibility of producing widespread behavior changes with a limited amount of therapist intervention. REFERENCES Egel, A., Koegel, R. L., & Schreibman, L. A review of educational-treatment procedures for autistic children. In L. Mann & D. A. Sabatino (Eds.), Fourth review of special education. New York: Grune and Stratton, in press. Eimas, P. Multiple-cue discrimination learning in children. Psychological Record, 1969, 9, 417-424. Fellows, B. J. The discrimination process and development. London: Pergamon Press, 1968. Gellerman, L. W. Chance orders of alternating stimuli in visual discrimination experiments. Journal of Genetic Psychology, 1933, 42, 206-208. Hale, G. A., & Morgan, J. S. Developmental trends in children's component selection. Journal of Experimental Child Psychology, 1973, 15, 302-314. Hale, G. A., & Piper, R. A. Developmental trends in children's incidental learning: Some critical stimulus differences. DevelopmentalPsychology, 1973, 8, 327-335. Kanner, L. Autistic disturbances of affective contact. Nervous Child, 1943, 2, 217-250. Koegel, R. L. Selective attention to prompt stimuli by autistic children. Unpublished doctoral dissertation, University of California, Los Angeles, 1971. Koegel, R. L., Egel, A. L., & Dunlap, G. Learning characteristics of autistic children. In W. S. Sailor, B. Wilcox, & L. J. Brown (Eds.), Methods of instruction with severely handicapped students. Baltimore: Brooks Publishers, in press. Koegel, R. L., & Rincover, A. Some detrimental effects of using extra stimuli to guide learning in normal and autistic children. Journal of Abnormal Child Psychology, 1976, 4, 59-71. Koegel, R. L., & Rincover, A. Research on the difference between generalization and maintenance in extra therapy responding. Journal of Applied Behavior Analysis, 1977, 10, 1-12. Koegel, R. L., & Schreibman, L. The role of stimulus variables in teaching autistic children. In O. I. Lovaas & B. Bucher (Eds.), Perspectives in behavior modification with deviant children. Englewood Cliffs, New Jersey: Prentice-Hall, 1974.
396
Koegel, Schreibman, Britten, and Laitinen
Koegel, R. L., & Schreibman, L. Identification of consistent responding to auditory stimuli by a functionally " d e a f " autistic child. Journal of Autism and Childhood Schizophrenia, 1976, 6, 147-156. Koegel, R. L., & Schreibman, L. Teaching autistic children to respond to simultaneous multiple cues. Journal of Experimental Child Psychology, 1977, 24, 299-311. Koegel, R. L., & Wilhelm, H. Selective responding to the components of multiple visual cues by autistic children. Journal of Experimental Child Psychology, 1973, 15, 442-453. Lovaas, O. I. The autistic child: Language development through behavior modification. New York: Irvington, 1977. Lovaas, O. I., Koegel, R. L., & Schreibman, L. Stimulus overselectivity in autism: A review of research. Psychological Bulletin, in press. Lovaas, O. I., & Newsom, C. D. Behavior modification with psychotic children. In H. Leitenberg (Ed.), Handbook of behavior modification and behavior therapy. New York: Appleton-Century-Crofts, 1976. Lovaas, O. I., & Schreibman, L. Stimulus overselectivity of autistic children in a two stimulus situation. Behaviour Research and Therapy, 1971,9, 305-310. Lovaas, O. I., Schreibman, L., & Koegel, R. L. A behavior modification approach to the treatment of autistic children. Journal of Autism and Childhood Schizophrenia, 1974, 4, 111-129. Lovaas, O. I., Schreibman, L., Koegel, R. L., & Rehm, R. Selective responding by autistic children to multiple sensory input. Journal of Abnormal Psychology, 1971, 77, 211222. MacKintosh, N. J. A theory of attention: Variation in the associability of stimuli with reinforcement. PsychologicalReview, 1975, 82, 276-298. Ornitz, E. M., & Ritvo, E. R. Perceptual inconstancy in early infantile autism. Archives of GeneralPsychiatry, 1968, 18, 76-98. Reynolds, B. S., Newsom, C. D., & Lovaas, O. I. Auditory overselectivity in autistic children. Journal of Abnormal Child Psychology, 1974, 2, 253-263. Rincover, A. Variables affecting stimulus-fading and discriminative responding in psychotic children. Journal of Abnormal Psychology, 1978, 87, 541-553. Rincover, A., & Koegel, R. L. Setting generality and stimulus control in autistic children. Journal of Applied Behavior Analysis, 1975, 8, 235-246. Rincover, A., & Koegel, R. L. Research on the education of autistic children: Recent advances and future directions. In B. B. Lahey & A. E. Kazdin (Eds.), Advances in Clinical ChiM Psychology, Vol. 1. New York: Plenum Press, 1977. Risley, T. R., & Wolf, M. M. Establishing functional speech in echolalic children. Behaviour Research and Therapy, 1967, 5, 73-88. Ross, A. O. Psychological aspects of learning disabilities and reading disorders. New York: McGraw-Hill, 1976. Rutter, M. Diagnosis and definition of childhood autism. Journal of Autism and Childhood Schizophrenia, 1978, 8, 139-161. Schopler, E. Early infantile autism and receptor processes. Archives of General Psychiatry, 1965, 13, 327-335. Schover, L. R., & Newsom, C. D. Overselectivity, developmental level and overtraining in autistic and normal children. Journal of Abnormal Child Psychology, 1976, 4, 289-298. Schreibman, L. Effects of within-stimulus and extra-stimulus prompting on discrimination learning in autistic children. Journal of Applied Behavior Analysis, 1975, 8, 91-112. Schreibman, L., & Koegel, R. L. Multiple cue responding in autistic children. In P. Karoly & J. J. Steffen (Eds.), Advances in child behavior analysis and therapy. New York: Wiley, in press. Schreibman, L., Koegel, R. L., & Craig, M. S. Reducing stimulus overselectivity in autistic children. Journal of Abnormal Child Psychology, 1977, 5, 425-436. Schreibman, L., & Lovaas, O. I. Overselective response to social stimuli by autistic children. Journal of Abnormal Child Psychology, 1973, 1, 152-168. Sidman, M., & Stoddard, L. T. Programming, perception and learning for retarded children. In N. R. Ellis (Ed.), International review of research in mental retardation (Vol. 2). New York: Academic Press, 1966.
Effects of Schedule of Reinforcement
397
Sutherland, N. S., & Holgate, V. Two-cue discrimination learning in rats. Journal of Comparative and Physiological Psychology, 1966, 61, 198-207. Sutherland, N. S., & MacKintosh, N. J. Mechanisms of animal discrimination learning. New York: Academic Press, 1971. Trabasso, T., & Bower, G. H. Attention in learning: Theory and research. New York: Wiley, 1968. Varni, J., Lovaas, O. I., Koegel, R. L., & Everett, M. L. An analysis of observational learning in autistic and normal children. Journal of Abnormal Child Psychology, 1979, 7, 31-43.