Journal of Autism and Developmental Disorders, VoL 22, No. 2, t992
Language Intervention and Disruptive Behavior in Preschool Children with Autism 1 Robert L. Koegel, Lynn Kern Koegel, and Alan Surratt University of California at Santa Barbara
D&ruptive behaviors are often exhibited by children with severe disabilities during difficult teaching tasks. Because learning verbal communication can be a difficult task for nonverbal children with autism, disruptive behaviors are common during such interventions. The purpose of this experiment was to assess whether the incorporation of parameters of natural language interactions and motivational techniques might reduce disruptive behavior during language teaching tasks. Within a repeated reversals design with order of conditions and number of sessions varied within and across children, treatment was conducted for two language teaching conditions. During one condition trials were presented serially in a traditional analog clinical format where the therapist presented instructions, prompts, and reinforcers for correct responses. The other condition incorporated parameters of natural language interactions and motivational techniques, such that stimulus items were functional and varied; natural reinforcers were employed; communicative attempts were reinforced; and trials were conducted within a natural interchange. Results showed that greater improvements in responding and considerably less (often negligible) disruptive behavior occmTed during the natural language teaching conditions. Results are discussed with respect to their implications for improving language interventions, and with respect to reducing disruptive behavior without the need for specialized or severe interventions focused specifically on the disruptive behavior.
1Preparation of this manuscript was supported in part by National Institute on Disability and Rehabilitation Research Cooperative Agreement No. G0087C0234; and by U. S. Public Health Service Research Grants MH28210 and MH39434 from the National Institute of Menta~ Health. The authors thank William Frea, Christine Hurley, Amanda Darnron, Lisa Daniel, and Teresa Paslawski for their assistance during this investigation. 141 0!.62-3257/92/0600-141506.50/0 9 1992PlenumPublishingCorporation
Koegel, Koegel, and Surratt
Severe disruptive behaviors are common in the syndrome of autism. Such behaviors are particularly likely to occur during difficult teaching tasks in order to escape or avoid the task (cf. Carr, Newsom, & Binkoff, 1976), thus making it difficult to remediate some of the children's more severe deficits through teaching. For example, language tasks appear to be very difficult for autistic children to learn (Carr, 1982; Carr & Durand, 1985; Charlop & Milstein, 1989; Garfin & Lord, 1986; Koegel & Traphagen, 1982; Lovass, 1977; Shah & Wing, 1986), and seem especially associated with disruptive behaviors. Attempts to decrease or eliminate disruptive behaviors in order to facilitate teaching such tasks have been wide-ranging and abundant in the literature. One interesting approach to improve this type of language teaching has focused on curriculum issues in order to change the teaching context so that variables associated with increased motivation to respond become a part of the teaching interaction. Research in this area has shown that such efforts can facilitate language learning directly and in addition may have the potential to reduce disruptive behaviors that occur in order to escape or avoid the task. A variety of methods have been used to accomplish this objective. One avenue has been to incorporate language tasks into a natural language teaching paradigm. This approach is receiving widespread attention in the literature with a variety of different populations, including children with autism (Koegel, O'DelI, & Dunlap, 1988; Koegel, O'DelI, & Koegel, 19887; Laski, Charlop, & Schreibman, 1988) children with mental retardation (Warren & Gazdag, 1990), and young children with language delays (Camarata & Nelson, in press; Hart & Risley, 1974). The results of the above research show that such teaching techniques have a desirable effect on initial acquisition of new behaviors as well as on the generalization of treatment gains. To date, most of the literature has addressed the influence of the teaching techniques on learning language directly. However, because of the demonstrated effect on many types of generalization, and because of the potential for decreasing disruptive behavior indirectly in young children without the need for specialized or severe treatment interventions, research on the influence of such techniques on disruptive behavior has been encouraged. Therefore, the specific purpose of the present investigation was to systematically examine the effect that a natural language teaching paradigm may have on disruptive behavior. Measures were recorded on children's language use as well as on their disruptive behaviors that were not directly treated.
Language Intervention and Disruptive Behavior
Subjects Three preschool children with autism participated in this investigation. All were diagnosed by an outside agency and referred to our center for treatment. An initial intake interview confirmed the symptoms of autism in accordance with the diagnostic criteria of infantile autism described in the DSM-III-R, including a lack of responsiveness to other people, gross impairment in communicative skills, bizarre responses to various aspects of the environment, and ritualistic behavior. In addition, all demonstrated severely disruptive behavior (described in detail in the Measures section). Child 1 was 3 years 8 months at the start of the study. His disruptive behavior interfered with testing under standardized testing conditions such that examiners were unable to obtain standard scores on either verbal or nonverbal intelligence tests (e.g., Leiter International Performance Scale and Stanford-Binet Intelligence Scale). However, he scored in the 1st percentile on the Peabody Picture Vocabulary Test, giving him an age equivalent of 2 years 3 months. On the Assessment of Children's Language Comprehension, he scored 39 out of 50 on the vocabulary items, and on two, three, and four critical items he scored 70, 50, and 20%, respectively. He rarely exhibited spontaneous language. However, he was able to use a number of single-word utterances, primarily for expressing needs and desires. Occasionally he combined two words, but longer word combinations typically appeared to be echolalic. The Vineland Adaptive Behavior Scales yielded an adaptive behavior composite score of 67. His mother reported that he exhibited numerous problems at home including aggression, a sleep disturbance, teeth grinding, tantrums and defiance, running away, and being "stubborn and sullen." Child 2 was 3 years 4 months at the start of the study. She lacked any type of pointing response, and was therefore unable to achieve a score under standard testing conditions, although the administration of both language and intelligence tests was attempted. She occasionally exhibited echolalia of words and phrases, but expressive language was never used in context. She appeared to be able to understand a few simple phrases frequently used by her mother, such as "get off the table" and "come here." Her adaptive behavior composite on the Vineland Adaptive Behavior Scale was 55. She was not toilet trained, and did not sit at the table long enough to eat her meals. Additionally, her mother reported that she had a variety of other problems in the home including, frequent public masturbation, sleeping and eating abnormalities, inappropriate crying and laughing, temper tantrums, teeth-grinding, and thumb-sucking.
Koegel, Koegel, and Surratt
Child 3 was 4 years 6 months at the start of the study. In addition to exhibiting extremely disruptive behavior that interfered with testing conditions, he also lacked a pointing response, and did not respond to standardized intelligence and language tests. He had no recognizable words but made approximately five consonant sounds and one vowel sound. He appeared to be able to understand simple commands such as "sit down," "throw it away," "turn off the light," "open the door." His adaptive behavior composite on the Vineland Adaptive Behavior Scale was 64. His mother reported that he exhibited numerous maladaptive behaviors at home including self-stimulatory behavior, temper tantrums, running away, hyperactivity, sleep disturbance, and crying and laughing "too easily."
Setting All sessions were conducted in a language treatment room in a oneto-one teacher child format. The treatment provider was an advanced doctoral student, and all sessions were supervised by a licensed speech/language pathologist. Sessions were conducted in a small (10 x 15 foot) clinic room that contained a table, chairs, and stimulus materials relevant to each task. Also present in the room was a small portable video camera/recorder.
Design Data were collected in the context of a repeated reversals design (Barlow & Hersen, 1984). To control for order effects, the number and order of sessions were varied both across subjects and within subjects. Subjects were seen 1 day per week, with one to three 10-rain sessions conducted per day over a 3-month period (Child 3 also had a 9-month break before the final follow-up session with his parent).
Data Collection Data were collected on two child behaviors, disruptive behavior and language target responses. Data also were recorded on the clinician's behavior during the sessions in order to assess the amount of instruction necessary for language tasks versus disruptive behavior in the two conditions, and to insure that teaching activities were taking place in both conditions. The clinician was not told prior to the sessions that teaching behavior would be scored. All sessions were videotaped and scored later by two independent advanced doctoral students. Tapes were dubbed in 10-sec intervals and data
Language Intervention and Disruptive Behavior
were recorded continuously as to whether or not disruptive behavior occurred during each lO-sec interval for the entire session. In addition, upper-division undergraduate students transcribed both the child's and the clinician's behavior from the tapes. To facilitate recording of data, the pause button was used and the tapes were rewound as often as necessary during the data analyses.
Treatment Conditions In both treatment conditions the clinician was instructed to present instructions to the child to imitate or produce language targets, to present reinforcers to the child if the child responded correctly, and to redirect the child to the task of the child exhibited disruptive behavior (climbing on the table, leaving the seat, etc.). Differences between the conditions were defined to be consistent with previous literature on this topic (Koegel, O'DeI1, & Koegel, 1987; Koegel et al., 1988). Table I presents a summary of the differences between conditions, and the detailed procedures for each condition are described below. Analog Teaching Paradigm. The analog treatment condition consisted of attempting to evoke imitative or spontaneous responses through the use of successive trials, with each item presented serially by the clinician until the child reached criterion. On each trial the following procedure was used: (a) The clinician presented the instruction "say + (desired child utterance)"
Table I. Conditions a Analog Stimulus items
a. Chosen by clinician b. Repeated until criterion is met c. Phonologically easy to produce, irrespective of whether they were functional in the natural environment a. Clinician holds up stimulus item; stimulus item not functional within interaction a. Correct responses or successive approximations reinforced a. Edible reinforcers paired with social reinforcers
aAdapted from Koegel, O'Dell, and Koegel ( 1 9 8 7 ) .
a. Chosen by child b. Varied every few trials
Clinician and child play with stimulus item (i.e., stimulus item is functional within interaction) a. Looser shaping contingency so that attempts to respond verbally (except self-stimulation) are also reinforced a. Natural reinforcer (e.g., opportunity to play with the item) paired with social reinforcers a.
Koegel, Koegel, and Surratt
while holding up the target object. The clinician then paused and waited for the child to respond. (b) If the child failed to respond, the clinician prompted the child (e.g., by repeating the original instruction) while still presenting the stimulus item. (c) Within a shaping paradigm (e.g., Martin & Pear, 1978), if the child imitated a correct approximation to the target response, the child was reinforced. Natural Language Teaching Paradigm. The Natural Language Teaching (NLP) condition consisted of changing the variables of the analog condition to approximate factors that have been identified as basic parameters in a natural language speaking situation (see Table I). The following parameters were manipulated: (a) Instead of the clinicians arbitrarily selecting a stimulus item, the clinician selected a stimulus item from the pool of stimulus items according to the child's selection of a preferred item for any given trial. (b) Instead of an item being presented serially until the child learned its label, the pool of stimulus items was presented so that the stimulus items could be varied according to the child's interests. (c) Instead of just giving a command such as " s a y . . . , " the clinician played with the toy and modeled the target response. (d) If the child failed to respond, the clinician played with the toy and modeled the response again. (e) The reinforcement contingency was broadened so that if the child imitated either the exact correct response or a successive approximation, or made any clear verbal attempt to respond (which could be differentiated from vocal self-stimulation), the child was reinforced (this did not change the data recording procedure, which was the same in both treatment conditions). (f) Instead of the child being reinforced with edibles and praise, the child was reinforced with the opportunity to play with the instructional stimulus (and with praise).
Fidelity of Implementation To assess whether treatment differences were occurring as described above, two advanced doctoral students, familiar with both treatment techniques independently scored 16 (39%) of the 41 videotaped sessions randomly distributed across all experimental conditions and children. According to the criteria specified in Table I, the observers rated each session as conforming to either the correct implementation of the Analog condition, the correct implementation of the NLP condition, or other. To eliminate observer drift and experimenter bias, the two observers were naive to experimental conditions from which the videotapes were drawn, and the tapes were viewed in a random order. The reliability results showed that scores for all 16 sessions were in agreement as to which condition was
Language Intervention and Disruptive Behavior
being implemented, and that one or the other of the two conditions was being correctly implemented in each session (i.e., the observers never scored the "other" category). Further the observers' scores were always in agreement with the assignment of conditions determined by the experimental design described above. These results indicate that there was a clear difference between the conditions and a high degree of fidelity of implementation.
Behavioral Definitions Data were recorded for two categories of child behaviors (disruptive behavior and language use) as well as for three clinician behaviors (presentation of instructions, type of instructions presented, and presentation of reinforcers). Each of these measures is defined below. Disruptive behavior was operationally defined as any behavior that caused a disruption in the session, such as producing an interruption in the presentation of task stimuli, directing behavior away from the task, or interfeting with responding to task stimuli. The specific behaviors included in the definition were crying, yelling, echolalia or verbal self-stimulating loud enough to mask the therapist's instructions, turning head or body away from the therapist/task, slapping, grabbing, hand-flapping for more than 2 seconds, sweeping hands across the table, covering face/mouth with hands, laying down on seat or table, and getting up out of seat during stimulus presentations. Language targets were defined individually for each child based on the child's level of language development. Data were recorded on one- and two-word utterances for Child 1, who was just beginning to exhibit oneword utterances at the start of his involvement in this experiment. For Child 2, who was primarily nonverbal with a small amount of echolalia, but rarely used speech in context, data were recorded on appropriate words used in context and on vocal attempts to use words in context. For Child 3, who was completely nonverbal at the start of his involvement in the experiment, data were recorded on his use of consonant-vowel word approximation. Clinician behaviors were recorded for the number and type of instructions presented per session, and for the number of reinforcers provided per session, for every session in the experiment. Thus, every utterance the clinician made during the sessions was classified by the raters (using the transcripts from each session) as either task-related, disruptive behaviorrelated, a reinforcer, or as other (i.e., one that could not fit into any of the above categories; typically these utterances consisted of repeating what the child had said, recasting the child's utterance, and interspersing comments not related to the target tasks).
Koegei, Koegel, and Surratt
Reliability Two observers independently recorded each measure in this experiment. Percentage agreement between the two observers was calculated for each measure using the formula, agreements divided by agreements plus disagreements. An agreement was defined for each measure as the two observers recording the same response for a specific time interval, and a disagreement was defined as one observer recording one response and the other observer recording another response in that time interval. Reliability measures were obtained for 27% (i.e., 11) of the sessions in the experiment for disruptive behavior. The average percentage agreements were very high throughout the experiment, with disagreements usually occurring only for very subtle or very short duration behaviors that were difficult to observe on the video recordings. The average percentage agreement for recording occurrences of disruptive behavior was 90% (range: 33 to 100%, with only one session below 89%); and the average percentage agreements for recording nonoccurrences was 84% (range: 28 to 100%, with only two sessions below 95%). Reliability measures were obtained for 71% (i.e., 29) of the sessions for recording the targeted language behaviors. The average percentage agreement was 95.8% (range: 72 to 100%). Reliability measures also were obtained for 71% (i.e., 29) of the sessions for recording clinician behaviors. The average percentage agreement for recording clinician instructions pertaining to the target tasks was 97.3% (range: 72 to 100%). The average percentage agreement for recording instructions pertaining to disruptive behavior was 98.9% (range 91.7 to 100%). The average percentage agreement for recording the presentation of reinforcers was 97.6% (range 83.6 to 100%). Reliability measures for fidelity of implementation of the treatment conditions were obtained for 39% (i.e., 16) of the sessions. The percentage agreement for rating fidelity of implementation was 100%.
Disruptive Behavior The major question asked in this investigation was does a natural language teaching approach produce less disruptive behavior than an analog approach? The results are presented in Figure 1. The data show that during the NLP sessions all three children evidenced much less disruptive behavior than during the Analog sessions. These results are shown repeatedly in the reversal design regardless of the order of conditions or the number of sessions per condition. Specifically, Child 1 averaged 53% of the intervals with
60 40 ~ ]
< -r" U..l rn
I I 1 2 3 4 Analogue
6 7 NLP
I-z U..l O
60 4O 2
6OI a.. 80 4O 20 0p
I 10 11 12 13 14 15 16 NLP
10 11 12 13 14
./' I 1
7 8 9 10 11 SESSIONS
Fig. 1. The figure shows a comparison in the level of disruptive behavior for NLP versus Analog treatment conditions. Percentage disruptive behavior is plotted on the ordinate, and sessions are plotted on the abscissa for each child. The asterisks indicate that these final sessions were conducted by the children's parents (who had been trained in the implementation of the NLP treatment procedures).
disruptive behavior during the Analog sessions in contrast to only 4% of the intervals during the NLP sessions; Child 2 showed 65% of the intervals with disruptive behavior during the Analog sessions and 8% during the NLP sessions; Child 3 averaged 31% of the intervals with disruptive behavior during the Analog conditions and 4% during the NLP sessions. In short, the NLP condition consistently showed lower levels of disruptive behavior both within and across children, as compared to the Analog condition.
Language Behavior The above results suggest that the use of an NLP teaching format may be advantageous with respect to producing very low levels of disruptive behavior during language teaching. It also is interesting to examine the differential effects the two teaching formats had on the children's performance of the language target behaviors during the experiment. Table II shows that, consistent with previous literature in this area, the children pro-
Koegel, Koegel, and Surratt
Table II. Target Behaviors--Overall Session Average Child 1 2 3
Behavior Utterances 2 or more words Words Attempts Word approximation
NLP 82 58.6 3.4 35.1 47
Analog 72.5 58 0.28 13.5 42.5
duced more correct target behavior in the NLP conditions than during the Analog conditions. Specifically, Child 1 exhibited both more single word utterances per minute as well as more multiple word utterances in the NLP conditions; Child 2 exhibited both more correct words and more attempts to produce words in the NLP conditions; Child 3, who never progressed beyond the one-word approximation stage during this experiment, nevertheless produced more correct word approximations during the NLP conditions than during the Analog conditions.
Clinician Behavior Table III presents the results for the clinician's number of instructions used in each condition, the type of instructions used, and the number of reinforcers presented in each condition. In both conditions most of the clinician's verbalizations were task-related. However, a large difference between the conditions occurred with respect to the type of instructions the clinician was presenting. That is, in the Analog condition, because of the excessive disruptive behavior, the clinician frequently needed to redirect the child from disruptive behavior back to the target task (e.g., "sit in your chair"). This was in marked contrast to the NLP condition, where the children typically remained on task, and the clinician only rarely needed to redirect the children back to the teaching task. Table III also shows that reinforcers were presented during both conditions, but were presented slightly more often during the Analog condition. This result is consistent with Neef, Iwata, and Page (1980), and suggests that the amount of reinforcement was probably not a major variable in the results. In summary, the results show that disruptive behavior was very low during language teaching sessions for these typically very disruptive children, but this was true only when the NLP conditions were used. Further, in all cases the children produced more correct target language behavior in the NLP conditions than in the Analog conditions.
Language Intervention and Disruptive Behavior
Table 1II. Clinician Behavior--Average Percentage of Total Clinician Utterances Persession Type of clinician utterance Instruction task related Child 1 Child 2 Child 3 Instruction disruptive behaviorrelated Child 1 Child 2 Child 3 Reinforcers Child 1 Child 2 Child 3 Other Child 1 Child 2 Child 3 Total number of clinician utterances Child 1 Child 2 Child 3
62.75 81.60 39.77
39.07 44.72 16.81
,84 2.62 3.29
23.52 31.77 13.29
15.35 6.63 7.49
28.45 I3.59 10.77
21.06 19.15 49.45
8.96 9.22 59,13
840 690 660
710 620 475
The results of this investigation are consistent with previous literature showing that teaching language to autistic children in a natural teaching context typically produces more correct target behavior than an analog approach. The results also add to the literature by showing that the children exhibited considerably less disruptive behavior during the natural language teaching conditions. The results a~so have implications for an easy means of reducing disruptive behavior during difficult language teaching tasks. That is, much of the current literature suggests that disruptive behavior often is exhibited by autistic children in order to escape from difficult teaching tasks (Carr & Durand, 1985; Carr et al., t976). Thus, the low levels of disruptive behavior may have occurred because natural language approaches incorporate many variables associated with increased motivation to respond during teaching tasks. For example, the NLP conditions in this experiment incorporated the important principles of interspersing maintenance tasks (Carr et al., 1976; Dunlap, 1984; Koegel & Koegel, 1986; Neef et al., 1980), of permitting the child to have a role in the selection of stimulus materials
Koegel, Koegel, and Surratt
(Koegel, Dyer, & Bell, 19877; Dyer, Dunlap, & Winterling, 1990), of reinforcing attempts at the target behavior (Koegel et al., 1988), and of using natural reinforcers (Williams, Koegel, & Egel, 1981). Thus, it is highly likely that the children were more motivated to participate in the teaching activities during the natural language teaching conditions, and therefore were less likely to exhibit disruptive behavior as escape responses. Overall, the results show that autistic children exhibited much less (often negligible) disruptive behavior during natural language teaching conditions, without the use of specialized or severe treatment techniques specifically focused on the disruptive behavior. This result has major implications for teachers who frequently need to work on the remediation of severe language deficiencies, and who have difficulties with disruptive behavior in their classrooms.
REFERENCES Barlow, D. H., & Hersen, M. (1984). Single case experimental designs: Strategies for studying behavior change (2rid ed.). Elmsford, NY: Pergamon. Camarata, S. M., & Nelson, K. E. (in press). Treatment efficiency as a function of target selection in the remediation of child language disorders. Clinical Linguistics and Phonetics. Carr, E. G. (1982). Sign language. In R. L. Koegel, A. Rincover, & A. L. Egel (Eds.), Educating and understanding autistic children. San Diego, CA: College Hill. Carr, E. G., & Durand, V. M. (1985). Reducing behavior problems through functional communication training. Journal of Applied Behavior Analysis, 18, 111-126. Carr, E. G., Newsom, C. D., & Binkoff, J. A. (1976). Stimulus control of self-destructive behavior in a psychotic child. Journal of Abnormal Child Psychology, 4, 139-153. Charlop, M. H., & Milstein, J. P. (1989). Teaching autistic children conversational speech using video modeling. Journal of Applied Behavior Analysis, 22, 275-285. Dunlap, G. (1984). The influence of task variation and maintenance tasks on the learning and understanding of autistic children. Journal of Experimental Child Psychology, 37, 41-64. Dunlap, G., Dunlap, L. K., Clarke, S., & Robbins, F. R. (in press). Functional assessment, curricular revision, and severe behavior problems. Journal of Applied Behavior Analysis. Dyer, K., Dunlap, G., & Winterling, V. (1990). Effects of choice making on the serious problem behaviors of students with severe handicaps. Journal of Applied Behavior Analysis, 23, 515-524. Garfin, D. G., & Lord, C. (1986). Communication as a social problem in autism. In E. Schopler & G. B. Mesibov (Eds.), Social behavior in autism. New York: Plenum Press. Hart, B. M., & Risley, T. R. (1974). Using pre-school materials to modify the language of disadvantaged children. Journal of Applied Behavior Analysis, 7, 243-256. Koegel, L. K., & Koegel, R. L. (1986). The effects of interspersed maintenance tasks on academic performance in a severe childhood stroke victim. Journal of Applied Behavior Analysis, 19, 425-430. Koegel, R. L., & Traphagen, J. (1982). Selection of initial words for speech training with nonverbal children. In R. Koegel, A. Riucover, & A. L. Eget (Eds.), Educating and understanding autistic children. San Diego, CA: College Hill. Koegel, R. L., Dyer, K., & Bell, L. (1987). The influence of child-preferred activities on autistic children's social behavior. Journal of Applied Behavior Analysis, 20, 243-252.
Language Intervention and Disruptive Behavior
Koegel, R. L., O'Dell, M. C., & Dunlap, G. (1988). Producing speech use in nonverbal autistic children by reinforcing attempts. Journal of Autism and Developmental Disorders, 18, 525538. Koegel, R. L., O'Dell, M. C., & Koegel, L. K. (1987). A natural language teaching paradigm for nonverbal autistic children. Journal of Autism and Developmental Disorders, 17, 187200. Laski, K. E., Charlop, M. H., & Schreibman, L. (1988). Training parents to use the natural language paradigm to increase their autistic children's speech. Journal of Applied Behavior Analysis, 21, 391-400. Lovaas, O. I. (1977). The autistic child; language development through behavior modification. New York: Irvington. Martin, G., & Pear, J. (1978). Behavior modification: What it is and how to do it. Englewood Cliffs, NJ: Prentice-Hall. Neef, N. A., Iwata, B. A., & Page, T. J. (1980). The effects of interspersal training versus high density reinforcement on spelling acquisition and retention. Journal of Applied Behavior Analysis, 13, 153-158. Shah, A., & Wing, L. (1986). Cognitive Impairments affecting social behavior in autism. In E. Schopler & G. B. Mesibov (Eds.), Social behavior in autism. New York: Plenum Press. Warren, S. F., & Gazdag, G. (1990). Facilitating early language development with milieu intervention procedures. Journal of Early Intervention, 14, 62-86. Williams, J. A., Koegel, R. L., & Egel, A. L. (1981). Response-reinforcer relationships and improved learning in autistic children. Journal of Applied Behavior Analysis, 14, 53-60.