JOURNAL OF APPLIED BEHAVIOR ANALYSIS 1993,269,389-398 NUMBER 3 (FALL 1993) AN EVALUATION OF TWO METHODS FOR INCREASING SELF-INITIATED VERBALIZATIONS IN AUTISTIC CHILDREN JOHNNY L. MATSON, JAY A. SEVIN, MARGARET L. Box, AND KELLEY L. FRANas LOUISIANA STATE UNIVERSITY AND BART M. SEVIN AUBURN UNIVERSITY Three children with autism and mental retardation were treated for deficits in self-initiated speech. A novel treatment package employing visual cue fading was compared with a graduated time-delay procedure previously shown to be effective for increasing self-initiated language. Both treatments induded training multiple self-initiated verbalizations using multiple therapists and settings. Both treatments were effective, with no differences in measures of acquisition of target phrases, main- tenance of behavioral gains, acquisition with additional therapists and settings, and social validity. DESCRIPTORS: autism, language, time delay Communication training has been an important cedure in which a verbal model is first paired withfocus in treating children with autism for many and later faded from a nonverbal discriminativeyears (Matson, 1989). However, one important stimulus, while also fading reliance on verbalaspect of language, spontaneous or self-initiated prompts. In previous studies using time delay, self-speech, has received little emphasis in the literature. initiated target phrases increased in all participants.Charlop, Schreibman, and Thibodeau (1985) de- Generalization to additional stimuli, settings, andfined self-initiated verbalizations as verbal responses persons occurred. Also, two studies demonstratedto nonverbal discriminative stimuli in the absence maintenance of acquired gains several months laterof verbal discriminative stimuli. (In contrast to the (Ingenmey & Van Houten, 1991; Matson et al.,definition of verbal stimuli given by Skinner, 1957, 1990).this study does not include printed words as verbal Given the importance of self-initiated speech,stimuli.) Given that much of normal daily com- the development of additional strategies for treatingmunication is not verbally prompted, teaching chil- this aspect of language may be useful. Studyingdren to initiate language is an important goal. alternative procedures may increase our understand- There have been a limited number of studies on ing of the variables associated with the acquisitionincreasing self-initiated speech in autistic children of self-initiated speech. Therefore, we examined a(e.g., Charlop et al., 1985; Charlop & Walsh, treatment package consisting of modeling, multiple1986; Ingenmey & Van Houten, 1991; Matson, exemplars and reinforcers, and a visual-cue/stim-Sevin, Fridley, & Love, 1990). These studies used ulus-fading procedure for increasing self-initiateda treatment package consisting of modeling, pos- verbalizations. In addition, we compared this al-itive reinforcement, and graduated time delay. ternative strategy to time delay, the only successfulGraduated time delay is a stimulus-shaping pro- treatment to date. We compared the effects of the two procedures on acquisition of target responses. Address correspondence and reprint requests to Johnny L. We also induded maintenance, generalization toMatson, 236 Audubon Hall, Department of Psychology,Louisiana State University, Baton Rouge, Louisiana 70803- other therapists and settings, and treatment ac-5501. ceptability (social validity). 389
390 JOHNNY L. MATSON et al. METHOD orating reports by their parents. Toys and edible items served both as stimulus cues for verbalizationsSubjects and as reinforcers for correct verbalizations. During Research participants were 3 boys with autism. the visual-prompt procedure, cue cards (4 in. byEach participant met DSM-III-R criteria for autism 6 in. index cards with target words written in dif-and was severely autistic as rated by the Childhood ferent colored markers) were used.Autism Rating Scale (Schopler, Reichler, & Renner,1988). Randy was 5 years old. He had an IQ of Target Behaviors and62 (Stanford Binet). His expressive language was Selection Proceduresequivalent to 2 years 4 months (Vineland Adaptive Parents of participants compiled a list of self-Behavior Scale). Chris was 4 years old. He had an initiated (nonverbally prompted) words or phrasesIQ of 71 (Stanford Binet). His expressive language considered important for treatment. Target vocal-was equivalent to 2 years 4 months (Vineland were selected from these lists to ensure thatAdaptive Behavior Scale). Darryl was 4 years old. izationsHe had an IQ of 47 (Stanford Binet). His ex- the verbalizations would be socially meaningful.pressive language was equivalent to 1 year 1 month Four targets were selected for each participant. For(Vineland Adaptive Behavior Scale). Randy and Chris the verbalizations were "hello," These participants were selected for participation "excuse me," "thank you," and "play with me."because of their severely limited use of self-initiated For Darryl, the verbalizations were the same, exceptlanguage. Specifically, language for Randy consist- "help me" was substituted for "play with me."ed entirely of immediate and delayed echolalia and The participants had not previously used thesesimple noun labeling in response to the question, phrases (as reported by parents and confirmed dur-"What is this?" Chris named objects and family ing direct observations of participants in multiplemembers when asked, and occasionally asked for settings and during baseline sessions). During thedesired foods. Darryls speech was severely delayed experiment, parents did not indude these targetand consisted almost entirely of contextually in- phrases in any home-based training.appropriate, delayed echoing of a few phrases (e.g., A correct response occurred when the participant "comb your hair, comb your hair"). Thus, verbal said the target phrase within 10 s after presentationbehavior of all 3 participants consisted almost en- of a nonverbal stimulus cue and prior to the pre-tirely of echoing modeled phrases and limited re- sentation of a verbal model. Nonverbal stimulussponding to verbal prompting. cues differed for each target word: For "hello," the stimulus consisted of the therapist entering the room. Setting and Materials Slight variations were introduced by having the Treatment occurred in a university clinic, with therapist enter the room waving, not waving, or a therapist and a behavioral rater present. Sessions entering after knocking. These variations were used were approximately 20 min long and occurred four to promote generalization of the verbalizations be- times per week. Setting generalization probes oc- ing trained. For "help me," stimulus cues involved curred in the homes of Randy and Chris and in a placing a desired toy or edible item out of the classroom for Darryl. participants reach, in a dosed container that the Stimulus materials present during sessions in- participant could not open, or in another inacces- cluded toys, edible items, childrens books, and cue sible part of the room. For "play with me," the cards. The toys (cars, tops, liquid bubbles, stuffed participant was shown games or toys requiring 2 animals), edible items (dry cereal and M&Ms®), players (e.g., ball to be tossed back and forth). For and childrens books were selected because they "excuse me," stimulus cues induded placing a de- appeared to be preferred by the participants during sired toy or edible item on a table or chair and two unstructured sessions prior to baseline, corrob- then blocking the participants path to it. For "thank
INCREASING LANGUAGE 391you," stimuli involved handing the participant any The delay interval between stimulus presentationof a number of desired toys or edible items. and verbal model (by the therapist) was initially 2 Thus, stimulus cues included object presentation, s and was increased to 4 s after two consecutivemanipulation of objects, and specific social scenar- sessions in which self-initiated phrases or correctios. Each target used multiple stimulus cues. In imitations of a target occurred for at least 80% oforder to enhance generalization to other similar the trials (four of five). The interval was graduallysituations not induded in training (training suffi- increased by 2-s durations in this manner. Self-cient exemplars, Stokes & Baer, 1977), these var- initiated verbalizations and correct imitations of theious cues and scenarios were rotated across trials. model were reinforced, but only self-initiated ver- balizations (before the model) were considered cor-Design rect. A multiple baseline design across behaviors was Reinforcers were rotated within and across ses-used. For each participant, two target phrases were sions and induded praise (for all targets), edibletrained with time delay, and two were trained using items if used as a stimulus cue (for "excuse me,the visual-cue procedure. No 2 participants received "help me," and "thank you"), access to toys whenthe same treatment for an identical pair of target toys were used as stimuli, and 30-s contingent playphrases; treatments for target phrases were coun- with the therapist (for "play with me").terbalanced across participants. Acquisition of a target response was defined as For each participant, treatment began simulta- three consecutive sessions in which self-initiated ver-neously on two target behaviors, one treated using balizations occurred during at least 80% of the trialstime delay and one treated using visual cues. When for the target.acquisition criteria for both of the initial verbali- Generalization probes and sequential modi-zations were met, treatment began on the remaining fication. Generalization probes occurred duringpair of target phrases. baseline for two additional therapists and one ad- Each participants design included baseline, ditional setting. Before successful acquisition (de-treatment, and follow-up phases. Generalization fined above) of target phrases with the first ther-probes occurred in the baseline phase to assess gen- apist, generalization probes identical to baselineeralized acquisition to additional therapists and one sessions occurred (i.e., no models). Once successfuladditional setting. In addition, generalization probes acquisition with this therapist was attained, trainingoccurred during the treatment phase. began with the second therapist. After acquisition criteria were reached with this therapist, trainingExperimental Conditions: began with the third therapist in different settings.Time Delay Thus, generalization consisted of a sequential mod- Baseline. Each session induded five consecutive ification procedure (Stokes & Baer, 1977).trials for each of four target phrases (a total of 20 Follow-up. Maintenance was assessed at 10stimulus prompts). The order of behaviors trained months for Randy and Chris and at 2 months forwas rotated across sessions. During baseline trials, Darryl. During follow-up sessions, stimulus pre-the therapist presented only nonverbal stimulus cues sentation was identical to treatment sessions. Self-(e.g., the therapist knocked and entered the room). initiated verbalizations were reinforced, but mod-Correct self-initiated target responses were rein- eling and graduated time delay were not used.forced with rewards intrinsic to the situation. Re-sponses other than the target behaviors were not Experimental Conditions:reinforced. Visual Cues Treatment. Treatment began simultaneously for Baseline. Stimulus cues were identical to thosethe first pair ofbehaviors. The time-delay procedure described for time delay (e.g., the therapist enteredwas identical to that used by Matson et al. (1990). the room) with one addition: During stimulus pre-
392 JOHNNY L. MATSON et al. Baseline Treatment Follow-up 100% i1 co0 80 8 C 60 O 2 40 (0) 1 20 4)a: o 100% W Q 80 *-I.50 660 C > O C Lu 2 40 . 0.Cl 20 A 000000000*iiift-. 10* ...... 5 10 15 20 25 a90 35 40 45 Sessions Figure 1. Percentage of correct self-initiated verbalizations during baseline, treatment, and follow-up sessions for Randy.Duration of the delay interval (e.g., 4 s, 6 s) and the phase of visual cue fading (e.g., Step 3, Step 4) employed in eachsession are noted by an arrow at each progression.
INCREASING LANGUAGE 393 Baseline Treatment Follow-up to 0 a 9 0 a) C2 I 100% 0 80 c 40 (a 4) fi2 30- 20 0 0)100% *_ 0 0s C E so 0OF T 4a2 LuW 40 20 =1 20 A .. .~~~. . . ... . -A-A Jl ......... 5 10 15 20 26 30 35 40 45 Sessions Figure 2. Percentage of correct self-initiated verbalizations during baseline, treatment, and follow-up sessions for Chris.Duration of the delay interval (e.g., 4 s, 6 s) and the phase of visual cue fading (e.g., Step 3, Step 4) employed in eachsession are noted by an arrow at each progression.
394 JOHNNY L. MATSON et al.sentation, the therapist held a flash card (4 in. by The third therapist began generalization program-6 in.) with the target word (e.g., "hello") printed ming immediately after the second, and these ther-on it. Correct self-initiated verbalizations were re- apists did not use cue cards during generalizationinforced. (Reinforcement procedures were identical programming.in both procedures.) Follow-up. Self-initiated verbalizations were re- Treatment. Each trial included stimulus-cue inforced, and modeling and visual cues were notpresentation followed by a 10-s interval for self- used.initiated responding. As in the time-delay proce-dure, if the target phrase was not self-initiated, the Social Validitytherapist modeled the target phrase; the model was Social validity was assessed in two ways. First,always followed by a 10-s delay to assess correct a group of 15 persons made up of special educationimitations. A five-step stimulus-fading procedure teachers, doctoral students in psychology, and par-was used over the course of treatment. In Step 1, ticipants parents considered vignettes describingthe stimulus included the cue card only (i.e., the both treatments. These persons individually ratedtherapist showed the participant a card with the each procedure using the Treatment Evaluation In-word "hello" printed on it). In Step 2, the cue ventory-Short Form, a measure of treatment ac-card paired with a nonverbal stimulus was used ceptability (Kelley, Heffer, Gresham, & Elliot,(i.e., the therapist entered the room holding the 1989; Miller & Kelley, in press). The order in whichcard). Step 3 was identical to Step 2 except the they rated the two procedures was balanced acrosssize of the card was reduced by one half; in Step the 15 raters. Differences in acceptability of the two4, the size of the card was again reduced by one procedures were analyzed using paired difference thalf. In Step 5, the card was completely eliminated; tests. Respondents rated both procedures favorably.only the nonverbal stimulus (e.g., entering the room) No significant differences were found.was presented. Modeling was used at each step if The second procedure was conducted to assessthe participant did not respond spontaneously dur- whether improvements in speech would be notice-ing the initial 10-s interval. The criterion for mov- able to average observers unfamiliar with autism.ing from step to step was three consecutive sessions Twenty undergraduate psychology students, blindwith correct self-initiated verbalizations (not imi- to experimental conditions, rated baseline andtations) occurring at rates of 80% or greater. Self- posttherapy videotapes of the participants. T testsinitiated phrases and correct imitations were rein- were used to examine rated differences in targetforced. behaviors and three untreated behaviors before and Generalization probes and sequential modi- after treatment. (See Matson et al., 1990, for morefication. Generalization procedures were identical details on this procedure.) Per comparison alphato those used in the time-delay procedure. Probes level was adjusted using the Bonferroni correctionoccurred prior to acquisition of target responses. procedure. Raters noted significant improvementsProgramming (sequential modification) occurred in all four target behaviors for all 3 participants,during treatment phases. The second therapist be- with the exception of "play with me" for Chris,gan generalization programming after the partici- which was significant only at the .05 level. Nopant had successfully reached Step 4 of visual cue significant differences in the three untreated behav- fading with the primary therapist. At the conclusion iors were noted. of treatment for the first 2 participants, we ques- tioned whether the gradual fading of the card was Interobserver Agreement a necessary component of treatment. Thus, we con- Interobserver agreement for target responses wasducted fading and generalization procedures for calculated for 37% of all sessions. Agreement oc-Darryl more rapidly. The intermediate steps of fad- curred when both raters scored a trial identicallying the visual cue (Steps 3 and 4) were eliminated. (i.e., no response, correct imitation, incorrect self-
INCREASING LANGUAGE 395initiated response, or correct self-initiated response). fective as time delay, the only treatment previouslyReliability was calculated by dividing the total found to be effective. No differences with regardnumber of agreements by the number of agree- to acquisition, maintenance, or treatment accept-ments plus disagreements and multiplying by 100%. ability were noted.Interobserver agreement was 99% to 100% for all In attempting to train self-initiated language,targets and participants. several difficulties are encountered. Self-initiated language often occurs in response to internal (i.e., physiological) or complex social cues. The nature RESULTS of these stimuli may in part account for the difficulty Figure 1 displays data for Randy. Both time in training self-initiated speech. First, of practicaldelay and visual cues were effective in increasing significance, these stimuli are often difficult to ma-self-initiated responding. Using time delay, Randy nipulate in natural settings. Second, stimulus ov-reached acquisition criterion for "play with me" erselectivity and lack of social awareness, commonwithin 14 treatment sessions. Using visual cues, in autistic children, may interfere with identificationacquisition of "hello" at Step 1 occurred within of and attention to events that serve as discrimi-nine treatment sessions. The card was completely native stimuli for self-initiated language. The vi-eliminated by the 21st treatment session. Acqui- sual-cue procedure was therefore intended to trainsition criteria for the remaining two target phrases verbalizations to occur in response to salient stimuliwere met within 10 treatment sessions. Training (e.g., large brightly colored cards). We hoped thatby additional therapists and in a second setting on pairing cue cards with more complex social cuesall target phrases produced almost immediate and would increase the salience of the social cue. Stim-accurate use of target phrases. Gains were main- ulus control is at the heart of this cueing procedure.tained at the 10-month follow-up. Few differences Cards were successfully established as discrimina-in the effectiveness of the two procedures occurred tive stimuli for target phrases (Step 1) prior tofor Randy. introducing nonverbal cues (Step 2). Thus, stim- Chris showed similar gains (Figure 2). Although ulus control progressed from verbal models to cardsearly sessions showed variable responding, initial and from cards to nonverbal stimuli.acquisition criteria were met for both verbalizations This extrastimulus prompting procedure repre-by the 16th treatment session. Acquisition of the sents one method of coping with the social andsecond set of verbalizations occurred before the attentional problems associated with autism. There10th treatment session. Gains with additional ther- is, of course, no special relevance to using coloredapists and setting occurred rapidly across all phases. cards. Any neutral stimulus with salient featuresWith the exception of the target phrase "play with might have sufficed. Cards have the advantage ofme" at the first follow-up session, gains were main- being small, easy to move into the participantstained after 10 months. Both procedures were also visual field, and easy to carry to new settings. Also,effective for Darryl (Figure 3). All four behaviors given the reading deficits of our participants, usingwere relatively stable at high levels of responding cue cards with words might have the benefit ofby the end of treatment. teaching children to read a few phrases. Some preliminary comparisons between the vi- sual-cue and graduated time-delay procedures were DISCUSSION attempted. Both treatment packages used verbal A treatment package consisting of verbal mod- modeling, positive reinforcement, rotation of re-eling, visual cues and fading, and training with inforcers, multiple stimulus cues, and stimulus-multiple exemplars and reinforcers was effective in shaping procedures. The chief differences in theincreasing self-initiated verbalizations in 3 children procedures induded (a) the use of verbal modelswith autism. The visual-cue procedure was as ef- versus verbal models with extrastimulus visual
396 JOHNNY L. MATSON et al. Baseline Treatment Follow-up I.ft pt _A I C) igRg~ ~ 2nd therapist training 1. U) x 100% X 80 m- E 80 OF a& 540 eon 20 0) 0 100% 00O 80 U)e 60 0)- 0. o 40 C 20 0 ioo 1 00%; k 80 00 E 4a 2C I 15 20 Sessions Figure 3. Percentage of correct self-initiated verbalizations during baseline, treatment, and follow-up sessions for Darryl.Duration of the delay interval (e.g., 4 s, 6 s) and the phase of visual cue fading (e.g., Step 3, Step 4) employed in eachsession are noted by an arrow at each progression.
INCREASING LANGUAGE 397prompts and (b) the use of 2-s graduated versus had been untested, our goal was not primarily to10-s constant time delay. The present study cannot study generalization but to evaluate the rapiditybe considered a pure comparison between gradu- with which acquisition could be achieved whenated and constant delay procedures, given the ad- treatment was implemented by new persons and inditional difference between the two treatment pack- new settings. Several steps were taken to ensureages. Also, there was perhaps a procedural bias in that naturalistic elements were included in training.favor of the visual-cue procedure because partici- Despite the clinical setting, scenarios were con-pants had less time to respond (only 2 s) in initial structed to occur as they would in a natural setting,time-delay sessions. Nevertheless, the effectiveness variations in the scenarios were introduced, andof the visual-cue procedure suggests that, at least reinforcers intrinsic to the language tasks were in-under some circumstances, graduated delay inter- duded. In addition, in the final phases of treatment,vals are not essential components for training self- target responding in natural settings was demon-initiated speech. Perhaps other methods of stimulus strated. A second potential limitation is that thereshaping might be substituted. Similarly, as noted may have been cross-treatment interference, givenwith Darryl, some stimulus fading steps may not that the treatments were introduced simultaneous-always be necessary. Future research should focus ly. However, because the integrity of the multipleon creating a technology for determining the level baseline was maintained (i.e., the second pair ofand speed at which stimuli need to be faded to target behaviors did not improve until treatmentmaximize generalization while minimizing the began), we consider cross-treatment interference tonumber of sessions. be unlikely. Some autistic children selectively respond to vi- No studies have focused on training autistic chil-sual stimuli (Rincover & Koegel, 1975). Given the dren to initiate language in response to internal cuesphenomenon of stimulus overselectivity, there ap- (e.g., physiological events). Target behaviors mightpears to be a sound theoretical rationale for in- include "Im hungry," "Im tired," and "I feelcluding visual components in language training for sick." It may be possible to manipulate some in-some children. Although a need for alternative ternal states and then teach children to label themtreatments has not been shown by demonstrated (e.g., to label "Im tired" after exercise or "Imweaknesses in the time-delay procedure, visual full" after eating a meal). A second approach mightcueing might be a back-up technique in cases in involve teaching children to verbalize internal stateswhich verbal prompting is not feasible (e.g., hear- that are difficult to manipulate but that are frequenting-impaired clients). The procedures used in this naturally occurring events (e.g., hunger, anger). Instudy could easily be adapted for use with signing light of the usefulness of behavioral approaches forand total communication. More important, devel- these verbalizations, more research is warranted.oping new techniques for increasing self-initiatedlanguage may lead to greater understanding of the REFERENCESprocesses that underlie current techniques. Com-parisons of different treatment techniques may lead Charlop, M. H., Schreibman, L., & Thibodeau, M. G. (1985). Increasing spontaneous verbal responding into the identification of components that are both autistic children using a time delay procedure. Journalcommon across effective treatments and essential of Applied Behavior Analysis, 18, 155-166.for positive outcomes. Charlop, M. H., & Walsh, M. E. (1986). Increasing autistic Our results should be interpreted in light of childrens spontaneous verbalizations of affection: An as- sessment of time delay and peer modeling procedures.several limitations. One consists of our use of se- Journal of Applied Behavior Analysis, 19, 307-314.quential modification to program generalization. Ingenmey, R., & Van Houten, R. (1991). Using timeTreatment was implemented by additional thera- delay to promote spontaneous speech in an autistic child. Journal of Applied Behavior Analysis, 24, 591-596.pists and in an additional setting. Self-initiated gen- Kelley, M. L., Heffer, R. W., Gresham, F. M., & Elliot, S.eralization was not evaluated. Because visual cueing N. (1989). Development of a modified treatment eval-
398 JOHNNY L. MATSON et al. uation inventory. Journal of Psychopathology and Be- Schopler, E., Reichier, R. J., & Renner, B. R. (1988). The havioral Assessment, 11, 235-247. childhood autism rating scale. Los Angeles: WesternMatson, J. L. (Ed.). (1989). Chronic schizophrenia and Psychological Services. adult autism: Issues in diagnosis, assessment and treat- Skinner, B. F. (1957). Verbal behavior. Englewood Cliffs, ment. New York: Springer. NJ: Prentice-Hall.Matson,J. L., Sevin,J. A., Fridley, D., & Love, S. R. (1990). Stokes, T. F., & Baer, D. M. (1977). An implicit tech- Increasing spontaneous language in three autistic chil- nology of generalization. Journal of Applied Behavior dren. Journal of Applied Behavior Analysis, 23, 227- Analysis, 10, 349-367. 234.Miller, D. L., & Kelley, M. L. (in press). Treatment ac- ceptability: The effect of parent gender, marital adjust- Received October 13, 1991 ment, and child behavior. Journal of Child and Family Initial editorial decision January 24, 1992 Behavior Therapy. Revisions received April 10, 1992; July 27, 1992;Rincover, A., & Koegel, R. L. (1975). Setting generality November 3, 1992; March 18, 1993 and stimulus control in autistic children. Journal of Ap- Final acceptance March 18, 1993 plied Behavior Analysis, 8, 23 5-246. Action Editor, Susan Fowler