The Analysis of Verbal Behavior

135

All children with autism have some level ofdifficulty communicating with other children andadults. The level of communication skill variesacross children, however, in many cases chil-dren never learn to talk (i.e., lack vocal verbalbehavior). Teaching vocal verbal behavior toyoung children with autism is difficult becausemany of the children initially do not have theability to imitate vocal sounds (echoic control).When children with autism do not begin imitat-ing vocal sounds, parents and teachers oftenbegin teaching the child to communicate viasign language or picture symbols. Althoughthese alternative forms of verbal behavior canbe used to communicate effectively, it does notnecessarily lead to an increase in vocalizationsfor children with autism. Moreover, few proce-dures exist for teaching a child with autism toimitate vocal sounds. A lack of a technologyfor teaching vocal imitation presents a chal-lenging situation to teachers who are not ableto get children to produce sounds so that theycan be shaped and directly reinforced.

Although few procedures exist for teachingvocal imitation to nonvocal children, the pro-cess by which children begin producing vocalsounds has been theorized by Skinner (1957)

Using Stimulus-Stimulus Pairing and Direct Reinforcement toTeach Vocal Verbal Behavior to Young Children With Autism

Regina A. Carroll and Kevin P. KlattUniversity of Wisconsin, Eau Claire

In this study the effect of a stimulus-stimulus pairing procedure was used as part of a clinical investigationto increase vocalizations for two young children diagnosed with autism. This procedure involved pairinga vocal sound with a preferred stimulus (e.g., toy) to condition automatic reinforcement. In addition, thisstudy assessed the effects of a direct reinforcement procedure to bring a vocalization under echoic controlfollowing the pairing procedure for 1 of the participants. The results showed the pairing procedure wasused successfully to increase a vocalization for one of the participants, and the vocalization was broughtunder echoic control.Key Words: autism, stimulus-stimulus pairing, automatic reinforcement

2008, 24,135–146

We thank the University of Wisconsin, Eau ClaireOffice of Research and Sponsored Programs for fi-nancial support, Sara Weinkauf for help in collectingdata, and Nicholas Vanselow for help with previousdrafts. Address correspondence to Kevin Klatt,Psychology Department, University of Wisconsin,Eau Claire, Eau Claire, Wisconsin, 54702;e-mail: klattkp@uwec.edu

and later by Vaughan and Michael (1982). Thisprocess could involve an infant receiving aninitially neutral stimulus (parent’s vocal sounds)at the same time as an already established rein-forcer (e.g., food, rocking). The parent’s vocalsounds may acquire reinforcing properties as aresult of being paired with the reinforcing stimu-lus. The infant may begin to say similar vocalsounds, despite never being directly reinforcedfor saying the sounds, engaging in vocaliza-tions because saying them produces reinforc-ing stimuli. This process might begin to explainwhy infants can be heard to engage in vocalsounds with no apparent reinforcement (e.g.,upon waking in a crib).

Research has been conducted to investigatewhether vocalizations can be conditionedthrough automatic reinforcement. For example,Smith, Michael, and Sundberg (1996) evaluatedthe vocal verbal behavior of two typically de-veloping infants. In this study, vocal soundswere paired with a reinforcing stimulus for oneparticipant and with neutral or aversive stimulifor another participant. The results showed anincrease in the frequency of vocalizations whena reinforcing stimulus was presented, and adecrease when an aversive stimulus was pre-sented. The frequency of the vocalization didnot increase when the neutral stimulus wasused; thereby demonstrating the vocalizationwas not under echoic control.

A number of studies have found that a stimu-lus-stimulus pairing procedure, in which a pre-ferred stimulus is paired with a vocal sound,effectively increases the frequency of vocal-izations in the absence of social consequences

136 REGINA A. CARROLL and KEVIN P. KLATT

(Sundberg, Michael, Partington, & Sundberg,1996; Yoon & Bennett, 2000; Miguel, Carr, &Michael, 2002). Sundberg et al. investigated theeffects of an adult pairing a vocal sound withthe delivery of a reinforcing stimulus on the fre-quency of vocalizations for young children withlanguage delays and a typically developing child.The target sounds identified for pairing werenovel vocalizations (or had not been emittedduring a prepairing condition). The results ofthis experiment showed that all of the childrenspontaneously emitted new vocal responses af-ter pairings, although not necessarily the targetvocalizations. The researchers concluded thatthe new vocal responses occurred due to auto-matic reinforcement because the new vocal re-sponses were not directly reinforced.

Yoon and Bennett (2000) conducted two ex-periments to further investigate the stimulus-stimulus pairing procedure. In the first experi-ment, the effects of pairing a reinforcing stimu-lus with novel and low-frequency vocalizationswas investigated for preschool children withsevere language and communication delays. Theresults showed that the pairing procedure waseffective in conditioning target vocalizations asreinforcers, although the increase in the fre-quency of vocalizations was shown to be tem-porary. In the second experiment, the pairing pro-cedure was compared to direct reinforcement ofthe vocalization (echoic training) using the samereinforcing stimulus. The results from this studyshowed the pairing procedure was more effec-tive than echoic training to increase the fre-quency of vocalizations.

Similarly, Miguel et al. (2002) investigated theeffects of a pairing procedure on the frequencyof vocalizations for three young children withautism. The target sounds were low-frequencyvocalizations emitted by participants during pre-vious observations. The results of this experi-ment showed a temporary increase in vocaliza-tions for 2 of the 3 participants. The participantfor whom the pairing procedure was ineffectivehad a more complex vocal verbal repertoire priorto the beginning of the study. The researchersspeculated that the pairing procedure might notwork as well with children who have some vocalverbal abilities such as manding.

Although the results of previous studies sug-gest that stimulus pairing may be an effectiveprocedure for increasing vocal verbal behavior,two recent studies have failed to replicate theseeffects for children with autism (Esch, Carr, &

Michael, 2005; Normand & Knoll, 2006). Esch etal. attempted to increase vocal sounds, using apairing procedure, for three participants diag-nosed with autism. In Experiment 1, the research-ers paired a vocal stimulus with a reinforcingstimulus and planned to immediately reinforcethe vocalizations. In this experiment, however,vocal sounds did not increase through pairing,thus vocal sounds could not be directly rein-forced. A second experiment involving pairingwithout direct reinforcement was then conductedand, again, vocal sounds did not increase.

More recently, Normand and Knoll (2006) in-vestigated the effects of stimulus-stimulus pair-ing on spontaneous vocalizations for a youngchild diagnosed with autism. The investigatorsrepeatedly paired preferred stimuli with two low-frequency target sounds and recorded the fre-quency of the target sounds immediately priorto and immediately following the pairing ses-sions and during a follow-up period. Resultsfrom this study showed phonemes did not in-crease following the pairing sessions.

The results from these experiments suggestthat vocal sounds can be increased throughan unidentified automatic reinforcement mecha-nism for some but not all participants. Althoughresults from the previous studies indicate thatstimulus-stimulus pairing may increase vocal-izations for children with autism and typicallydeveloping children and infants, none of theexperiments have included young children di-agnosed with autism under the age of 3. Onepurpose of the present experiment, therefore,was to investigate the use of a stimulus-stimu-lus pairing procedure to increase both low-fre-quency and novel vocalizations for two chil-dren under age 3 diagnosed with autism. Eachchild had a different verbal repertoire as as-sessed by the Behavioral Language Assess-ment (BLA) (Sundberg & Partington, 1998). Asecond purpose of the present study was toinvestigate the use of the stimulus-stimuluspairing procedure as part of clinical servicesprovided to young children diagnosed withautism, as opposed to a purely systematicevaluation of the procedure’s effectiveness.

GENERAL METHOD

Participants

The participants in this study were Mary, age22 months, and Max, age 23 months. Both chil-

137USING STIMULUS-STIMULUS PAIRING

dren had been diagnosed with autism by a pe-diatrician independent of the study. The BLA(Sundberg & Partington, 1998) was used to as-sess participants’ verbal repertoires and deter-mine a classification profile from Level 1 (lowverbal repertoire) to Level 5 (high verbal reper-toire). The language repertoire of Mary was atLevel 1. She occasionally emitted a few vocalsounds, but had no vocal imitative skills. Shedid not have a mand nor tact repertoire at thestart of the study. The language repertoire ofMax was at Level 3. At the start of the studyMax had strong echoic and mimetic imitativeskills, was able to mand for preferred stimuliand could tact simple objects, but had a limitedintraverbal repertoire.

Setting and Materials

Training sessions were conducted in a 3.5 mx 3.5 m therapy room. The room contained achild sized table and chairs. The materials usedthroughout the session included preferredstimuli (e.g., food, toys, and videos).

Preference Assessment

Each participant was observed for 45 min inthe therapy room with a variety of foods andtoys available to them. The percentage of 1-min intervals that the participants were en-gaged with each particular item was recorded.Engagement in an item was defined as any-thing the participant touched, picked up, ma-nipulated, consumed, or an occasion of look-ing at the television (for at least 2 s) in a 1-mininterval. The five items that Max engaged infor the highest percentage of 1-min intervalswere selected for a stimulus preference assess-ment. Mary did not engage with the toys oractivities that were available to her. Repeatedobservations revealed the two stimuli Marywould engage with were edibles and videos.Thus, both edibles and videos were paired to-gether and used as preferred stimuli through-out the pairing sessions.

Prior to each session for Max, a single-arraymultiple-stimulus preference assessment wasconducted (Higbee, Carr, & Harrison, 2000). Theexperimenter placed the five items selectedthrough direct observation in front of Max. Thefirst item that Max touched or pointed to wasselected as the preferred stimulus for the sub-sequent session. Two independent observers

in an adjoining observation room recorded thepreferred stimulus selected by the participantduring the preference assessment.

Target Response and Measurement

In order to identify target responses, Maxand Mary were observed during 1-hr behav-ioral therapy sessions (unrelated to the study).Max was observed for six 45-min sessions andMary was observed for eight 45-min sessions.During these sessions, two independent ob-servers recorded all vocalizations emitted bythe participants. For both participants, the tar-get vocalizations selected for the study includedboth a known and novel target sound. Theknown target sound was the lowest frequencyone-syllable vocalization produced by partici-pants throughout observations and the noveltarget sound was a sound that the participantnever emitted during observations. Through-out the study, target sounds were defined asthe production of any target syllable thatmatched or was a close approximation to themodel presented. For Mary the known targetsound was /ts/ and the novel target sound was/buh/. For Max, the known target sound was /m/ and the novel target sound /I/.

In Experiment 1, the frequency of targetsounds emitted by the participant before andafter each stimulus-stimulus pairing session wasmeasured to evaluate the effects of a stimulus-stimulus pairing procedure on the occurrenceof known and novel target sounds. In addition,for Max the effects of a direct reinforcementprocedure on the frequency of target soundswere also evaluated. In Experiment 2, echoiccontrol of the known and novel target soundsfor Mary was evaluated using variations of adirect reinforcement procedure. The percent-age of correct vocal responses (Procedures 1,2, 4, and 5) and the frequency of target sounds(Procedures 1 and 3) emitted by Mary were re-corded during direct reinforcement sessions.

Interobserver Agreement

To assess interobserver agreement on theoccurrence of the dependent variable, two in-dependent observers recorded data during 77%of randomly selected sessions for Mary and100% of sessions for Max. Interobserver agree-ment on frequency of target sounds was calcu-lated by dividing the number of observer agree-

138 REGINA A. CARROLL and KEVIN P. KLATT

ments on response occurrence recorded in eachsession by agreements plus disagreements andmultiplied by 100 to yield a percentage of agree-ment. For direct reinforcement sessions,interobserver agreement on the percentage cor-rect of target sounds was calculated by divid-ing the number of observer agreements on cor-rect target sounds in each session by agree-ments plus disagreements and multiplied by 100to yield a percentage of agreement. Theinterobserver agreement averaged 91% for Mary(range 82% to 100%) and was 100% for Max.

Independent Variable Integrity

Two independent observers recorded theimplementation of the stimulus-stimulus pair-ing, control, and direct reinforcement proceduresin 54% of sessions for Mary and 100% of ses-sions for Max. Each session was scored as ei-ther correct (100% accuracy) or incorrect (lessthan 100%). In general the experimenter’s pre-sentation of the target sound and preferredstimulus were scored. For example, a stimulus-stimulus pairing session was scored as correctif the experimenter (a) presented 5 target soundsduring a trial, (b) presented the sounds within 5s, (c) presented the preferred stimulus after thepresentation of the first 3 target sounds, and(d) did not present any other stimulus during atrial. The number of correct sessions was di-vided by the number of incorrect plus correctsessions and multiplied by 100 to generate apercentage of independent variable integrity(IVI). The independent variable integrity was96% for Mary and 100% for Max.

EXPERIMENT 1: STIMULUS-STIMULUSPAIRING

The purpose of Experiment 1 was to assessthe effects of stimulus-stimulus pairing on thefrequency of target sounds. The frequency ofthe known target sound for Max did not in-crease following stimulus-stimulus pairing;therefore, the effect of a direct reinforcementprocedure on the frequency of target soundswas also assessed for Max.

Experimental Design

A multiple-baseline design across target vo-calizations was used to assess the effects ofstimulus-stimulus pairing on target sounds.

Procedure

Presession and postsession observations.Pre- and postsession observations (5 mineach) were conducted immediately before andafter each baseline, control, stimulus-stimu-lus pairing, and maintenance conditions. Dur-ing these observation sessions, participantswere allowed to walk around the room and haveaccess to various toys while the frequenciesof the target sounds were recorded. There wasminimal interaction between the experimenterand participant during these observations.

Baseline. Baseline sessions were con-ducted the same as pre- and postsession ob-servations and were 5 min in duration. Thebaseline condition was conducted to recordparticipants’ vocal repertoires in the absenceof an independent variable.

Control. During control sessions a trial con-sisted of the experimenter saying the targetsound five consecutive times and then waiting20 s to present a preferred stimulus to the par-ticipant. The participant was then allowed toengage with the preferred stimulus for 20 s be-fore it was removed and a new trial began. If theparticipant emitted the target sound during the20 s interval, the timer was reset and the pre-sentation of the preferred stimulus was delayed20 s. This correction procedure was used tocontrol for adventitious reinforcement of tar-get sounds. Each session consisted of 20 trialsfor Max. Since Mary engaged in problem be-havior during the control condition, an aver-age of 10 trials were presented throughout eachcontrol session. The control sessions for Marywere terminated if she engaged in problem be-havior (e.g., crying, screaming) for three consecu-tive trials. For Mary, 50% of the control sessionswere terminated before 20 trials were presented.

Stimulus-stimulus pairing. During stimulus-stimulus pairing sessions a trial consisted ofthe experimenter saying the target sound threeconsecutive times, then immediately present-ing the preferred stimulus while simultaneouslysaying the target sound two more times. Theparticipant was allowed access to the preferredstimulus for 20 s and then the next pairing trialwas presented. A correction procedure was alsoimplemented in this condition to control for ad-ventitious reinforcement. If the participant emit-ted the target sound the next trial was delayed20 s. Each session consisted of 20 stimulus-stimulus pairing trials.

139USING STIMULUS-STIMULUS PAIRING

Maintenance. Following the stimulus-stimulus pairing condition for Mary, main-tenance sessions were conducted to recordthe frequency of the target sound /ts/ inthe absence of the independent variable.

During maintenance sessions, pre- andpostsession observations, 5 min each, wereconducted immediately before and after a 1-hr therapy session that was conducted in-dependent of the study.

Figure 1. The response per minute of Mary’s target sounds (/ts/ in upper panel and /buh/ in lower panel)during pre- and postsession observations.

140 REGINA A. CARROLL and KEVIN P. KLATT

Direct reinforcement of echoic responses.The stimulus-stimulus pairing procedure didnot result in an increase in the frequency ofMax’s target sound. Therefore a direct rein-forcement procedure was implemented to as-sess its effects on the frequency of his targetsound. During the direct reinforcement condi-tion, an echoic trial began by the experimentersaying the target sound once (e.g., “say m”). Ifthe participant emitted the target sound within3 s a preferred stimulus was immediately deliv-ered. The participant was given access to thepreferred stimulus for 20 s. After 20 s hadelapsed, the stimulus was removed and the nexttrial was presented. If the participant did notemit the target sound within 3 s the preferredstimulus was not delivered and the next trialwas presented. A total of 20 trials were pre-sented in each session. Response frequencyof the target sound /m/ was recorded duringpre- and postsession observations immediatelybefore and after each direct reinforcement ses-sion. The percentage correct of vocal respond-ing within each session (the 20 direct reinforce-ment trials) was also recorded.

RESULTS

The frequency of Mary’s target vocalizationsduring pre- and postsession observations isshown in Figure 1. In the baseline and controlconditions the known target sound, /ts/ (upperpanel), averaged 0.13 responses per minute(range, 0 to 0.4) during presession observationsand did not occur during postsession observa-tions. During the stimulus-stimulus pairing con-dition the target sound /ts/ averaged 0.65 re-sponses per minute (range, 0 to 3.4) duringpresession observations and 1.55 responses perminute (range, 0 to 3.8) during the postsessionobservations. In the maintenance condition thetarget sound averaged 2.17 responses per minute(range, 0 to 4.2) during presession observationsand 2.63 responses per minute (range, 0.4 to 5.2)during the postsession observations. The lowerpanel of Figure 1 shows no increase in the noveltarget sound /buh/ following the baseline, con-trol, and stimulus-stimulus pairing conditions.

The frequency of Max’s target vocalizationsduring pre- and postsession observations areshown in the top two panels of Figure 2. In thebaseline and control conditions the known tar-get sound, /m/ (upper panel), averaged 0.03 re-sponses per minute (range, 0 to 0.2) during

presession observations and did not occur dur-ing postsession observations. During the stimu-lus-stimulus pairing condition, the frequency ofthe target sound increased to 1.0 response perminute during postsession observations for onesession, but then returned to zero for the subse-quent sessions. The upper panel of Figure 2 alsoshows that there was no increase in the fre-quency of the known target sound /m/ followingthe direct reinforcement condition; therefore, thefrequency of the novel target sound /I/ was notassessed following the baseline sessions.

The percentage of correct vocal responsesfor the target sound /m/ in the direct reinforce-ment condition for Max is shown in the bottompanel of Figure 2 (these data correspond di-rectly to the four data points in the direct rein-forcement condition in the first panel). In thiscondition, the average percentage of correctresponses was 90% (range, 75% to 100%).These data show that although Max was re-sponding accurately in the direct reinforcementcondition (Panel 3), there was no increase inthe frequency of the sound during pre- orpostsession observations (Panel 1).

EXPERIMENT 2: ECHOIC CONTROL

The stimulus-stimulus pairing procedureresulted in an increase in the known targetsound /ts/ for Mary in Experiment 1. The nextquestion, then, was whether the target soundcould be brought under echoic control. Thepurpose of Experiment 2 was to assesswhether the known target sound previouslyincreased with stimulus-stimulus pairingcould be brought under echoic controlthrough direct reinforcement.

Participant

Mary participated in Experiment 2. Max didnot participate in Experiment 2 because an in-crease in the frequency of the known targetsound was not obtained after the stimulus pair-ing procedure and the target sound was alreadydemonstrated to be under echoic control.

Experimental Design

A comparison design was used to investigatethe effects of several variations of a direct rein-forcement procedure on echoic control of thetarget sounds.

141USING STIMULUS-STIMULUS PAIRING

Procedure

Baseline: Direct reinforcement probes. Thepurpose of the baseline condition was to as-sess Mary’s ability to imitate the experimenter’s

presentation of the target sound prior to theimplementation of direct reinforcement proce-dures. Each baseline session consisted of 20probe trials. Each trial consisted of the experi-menter presenting the target sound once (e.g.,

Figure 2. The response per minute of Max’s target sound during pre- and post-session observations (panels1 & 2) and percentage correct of direct reinforcement trials (panel 3).

142 REGINA A. CARROLL and KEVIN P. KLATT

“say ts”). If Mary approximated or matchedthe target syllable within 5 s (i.e., emitted thesound /t/ or /s/ separate or in combination),the probe was recorded as correct and the nextprobe was presented. If Mary did not emit thetarget sound within 5 s, the probe was recordedas incorrect.

Following the intervention for the knowntarget sound, the baseline procedures wereslightly revised for the novel sound. The pro-cedure was revised to provide Mary with alonger interval of time in which to respond toan echoic trial. For /buh/, the experimenter pre-sented the target sound every minute in a 5-min block during the baseline condition. A to-tal of five trials were presented throughoutthe baseline session. If Mary approximated ormatched the target sound within 15 s the probewas recorded as correct and the next probewas presented. If Mary did not emit the targetsound within 15 s the probe was recorded asincorrect. No preferred stimulus was deliveredfor correct or incorrect responses during thebaseline condition.

Direct reinforcement procedures. Five varia-tions of a direct reinforcement procedure wereimplemented in Experiment 2 (the direct rein-forcement procedures were altered by chang-ing the antecedent stimulus when a change inbehavior was not observed). In general, a di-rect reinforcement trial began with the experi-menter saying the target sound once (e.g., “sayts”). If Mary emitted the target sound within 5 s(Procedures 1 and 2) or 15 s (Procedures 4 and5) a preferred stimulus was immediately deliv-ered. Mary was given access to the preferredstimulus for 20 s. After 20 s had elapsed theitem was removed and the next trial was pre-sented. If Mary did not emit the target soundwithin 5 s (or 15 s) of the experimenter’s presen-tation of the sound the preferred stimulus wasnot delivered and the next trial was presented.

Procedure 1: Stimulus-stimulus pairingand direct reinforcement. In this condition the5-min pre- and postsession observations andstimulus-stimulus pairing trials were conductedthe same as in Experiment 1. The procedureconsisted of presenting five stimulus-stimu-lus pairing trials followed by 20 direct rein-forcement trials. The percentage correct ofvocal responding during the 20 direct rein-forcement trials and the frequency of targetsound during pre- and postsession observa-tions was recorded.

Procedure 2: Direct reinforcement withoutpairing. In this condition 20 direct reinforce-ment trials were presented without stimulus-stimulus pairing. The percentage correct ofvocal responding during the direct reinforce-ment trials was recorded.

Procedure 3: Direct reinforcement contin-gent on the target sound. This condition in-cluded a 5-min pre- and postsession observa-tion and a 30-min (on average) direct reinforce-ment session in which all target sounds emit-ted by Mary were directly reinforced withedibles. If the target sound was emitted byMary at any time throughout the session, theexperimenter immediately delivered the pre-ferred stimulus while simultaneously present-ing the target sound once. The frequency ofthe target sound during pre- and postsessionobservations and during the direct reinforce-ment sessions was recorded.

Procedure 4: Direct reinforcement of anyvocalization. In this condition the experimenterpresented 15 direct reinforcement trials. A trialwas presented every minute, in three 5-minblocks distributed throughout an hour longtherapy session (e.g., a block was presented inthe beginning, middle and end of the therapysession). If Mary emitted any vocalizationwithin 15 s of the experimenter’s presentationof the target sound (e.g., “say ts”) she immedi-ately gained access to a preferred stimulus.

After the fourth session of this condition, a5-s delay to reinforcement for nontarget vocal-izations was added. If Mary emitted the targetsound at any time she gained immediate accessto the preferred stimulus. If Mary emitted a non-target vocalization, however, the response wasnot reinforced for 5 s. Thus, Mary was pro-vided immediate access to a preferred stimulusfor emitting the target sound and delayed ac-cess to the preferred stimulus for saying anyother vocal sound following the experimenter’spresentation of the sound. In this condition thepercentage correct of vocal responding duringthe 15 direct reinforcement trials was recorded.

Procedure 5: Direct reinforcement of the tar-get sound. In this condition Mary could gainaccess to the preferred stimulus only if she emit-ted the target sound following theexperimenter’s presentation of the sound. Atotal of 20 direct reinforcement trials were pre-sented in a session and the percentage correctof vocal responding during the 20 direct rein-forcement trials was recorded.

143USING STIMULUS-STIMULUS PAIRING

RESULTS

The upper panel of Figure 3 shows the per-centage correct of vocal responses for the tar-

get sound /ts/ during baseline, direct rein-forcement Procedure 1, Procedure 2, Procedure4, Procedure 5, and maintenance sessions.During the baseline condition and direct rein-

Figure 3. The percentage correct of direct reinforcement trials for the target sound /ts/ during baseline, directreinforcement procedure 1, 2, 4, and 5, and maintenance conditions (upper panel) for Mary. The response perminute of the target sound /ts/ during direct reinforcement Procedure 1 and Procedure 3 (lower panel) for Mary.

144 REGINA A. CARROLL and KEVIN P. KLATT

forcement Procedures 1 and 2, the averagepercentage of correct target responses was3.13% (range 0% to 10%). During Procedure 4Mary emitted target and nontarget vocal re-sponses for an average of 77% (range 47% to90%) of trials. In this condition the averagepercentage of target vocal responses was38.5% (range 13% to 73%) and the averagepercentage of nontarget vocal responses was38.5% (range 13% to 65%). After the imple-mentation of Procedure 5, the average percent-age of correct target responses increased toan average of 81% (range 45% to 100%). Theupper panel of Figure 3 also shows the per-centage of trials with correct target responsesfor the 10 weeks that maintenance sessionswere conducted. During maintenance ses-sions, responding remained at 100% correctacross all trials.

The bottom panel of Figure 3 shows the fre-quency of the target sound /ts/ during directreinforcement Procedures 1 and 3. During Pro-cedure 1, the target sound averaged 1.93 re-sponses per minute (range 0.6 to 3.3) duringpresession observations and averaged 0.6 re-sponses per minute (range 0 to 1.4) duringpostsession observations. During Procedure3, the target sound averaged 0.77 responsesper minute (range 0.2 to 1.4) during presession

observations, 0.48 responses per minute (range0.2 to 1.02) during the direct reinforcement ses-sion, and 1.03 responses per minute (range 0.2to 2.2) during postsession observations.

Figure 4 shows the percentage of correctvocal responses for the novel target sound/buh/ during baseline and direct reinforcementProcedure 5. During baseline, the target re-sponse did not occur and during Procedure 5the average percentage of correct target re-sponses was 15.4% (range 0% to 80%).

GENERAL DISCUSSION

The purpose of these two experiments wasto investigate whether a low-frequency andnovel vocalization could be increased using astimulus-stimulus pairing procedure, and thenbrought under echoic control for 2 young chil-dren diagnosed with autism. The results ofExperiment 1 show the stimulus-stimulus pair-ing procedure produced an immediate increasein the frequency of the known target sound /ts/ for Mary. These results are consistent withprevious studies in which the use of a stimu-lus-stimulus pairing procedure was followedby an increase in unprompted target vocaliza-tions (Sundberg et al., 1996; Miguel et al., 2002;Yoon & Bennett, 2000).

Figure 4. The percentage correct of direct reinforcement trials for the target sound /ts/ during direct rein-forcement Procedure 5 for Mary.

145USING STIMULUS-STIMULUS PAIRING

Conversely, the frequency of Mary’s targetvocalizations in Experiment 1 maintained inboth the pre- and postsession observationswell after the pairing procedure was removed.This result differs from previous research inwhich the frequencies of target vocalizationsreturned to baseline levels once stimulus-stimulus pairing was removed (Miguel et al.,2002). Determining why Mary’s vocalizationmaintained, however, is difficult consideringthe maintaining reinforcer was not identified.Her target vocalization may have maintainedbecause saying the sound now producednonsocially mediated reinforcement. Anotherexplanation is that Mary engaged in the vo-calization because it was inadvertently rein-forced on an intermittent schedule of reinforce-ment by family members or therapists.

The stimulus-stimulus pairing procedurewas ineffective in increasing the frequency ofthe novel sound /buh/ for Mary. Although thenovel sound did occur within a few sessionswhen a direct reinforcement procedure wasused, it remained inconsistent across sessions.Miguel et al. (2002) suggested the need forfurther investigation of the differential effectsof stimulus-stimulus pairing with already ex-isting vocalizations compared to novel vocal-izations. The results from these experimentssuggest it may be easier to increase existingrather than novel vocalizations.

The stimulus-stimulus pairing procedure didnot produce an increase in the frequency ofMax’s target sound /m/ in Experiment 1. Thisfailure to increase a low-frequency vocaliza-tion is similar to the results obtained by Miguelet al. (2002) in which a pairing procedure wasineffective in increasing target vocalizationsfor one of the participants. Given that thestimulus pairing procedure did not increasethe low-frequency target sound for Max, noattempt was made to use the pairing proce-dure on the novel sound.

The results from these two experimentsverify what has been reported in previous ex-periments, namely that the stimulus-stimuluspairing procedure can be used effectively toincrease vocalizations for some but not all chil-dren with autism. Miguel et al. (2002) specu-lated the pairing procedure might work differ-ently depending on the pre-existing verbalrepertoire of the child. For example, in theirstudy, the stimulus-stimulus pairing proce-dure increased the frequency of target vocal-

izations for participants with no existing vo-cal imitation repertoire; however, it did notincrease the frequency of target vocalizationsfor the participant that had a vocal imitationrepertoire at the start of the study. This find-ing is consistent with the results of the studyconducted by Normand and Knoll (2006) inwhich a pairing procedure did not increasevocalizations for a participant that was as-sessed at a Level 4 on the BLA (Sundberg &Partington, 1998) prior to implementing stimu-lus pairing. The results from the current studyalso found the pairing procedure worked forthe child with the most limited vocal imitationrepertoire. To illustrate, prior to the study Maxdemonstrated vocal imitation skills and wasassessed at Level 3 on the BLA, whereas Maryhad no verbal imitation skills and assessed atLevel 1 on the BLA. Further research needsto be conducted to conclude that the pairingprocedure works better for children with verylimited repertoires.

In addition to an increase in the low-fre-quency target vocalization for Mary, the re-sults of Experiment 2 showed the target sound/ts/ also came under echoic control. Anecdot-ally, once the target sound was brought underechoic control, 12 other vocalizations werethen also brought under echoic control, al-though this was not a formal part of eitherstudy nor experimentally controlled. Nonethe-less, this result suggests that future researchshould continue to investigate whether gen-eralization of vocalizations can be obtainedonce a target vocalization has been broughtunder echoic control.

The results from these two experimentsshould be interpreted with caution due to sev-eral limitations. First, the preferred stimuli usedfor Max in Experiment 1 were determined froma multiple-array stimulus preference assess-ment. The stimuli were not established rein-forcers because a stimulus-response relation-ship was never directly assessed. The pre-ferred stimuli used for Mary were determinedthrough direct observation, so whether thosestimuli functioned as reinforcers is unknown.It should be noted, however, that for Mary thestimuli used during pairing had been used toincrease other behaviors unrelated to thisstudy as part of her overall therapy.

Another limitation to this study was the lackof experimental control with the multiplebaseline design in Experiment 1 and the use of

146 REGINA A. CARROLL and KEVIN P. KLATT

a comparison design in Experiment 2 to investi-gate variations of a direct reinforcement proce-dure to establish echoic control over Mary’starget sound. A comparison design was usedbecause the study was conducted in a clinicalsetting and family members preferred tryingvarious direct reinforcement procedures toquickly identify what would bring the targetsound under echoic control. To bring the tar-get sound under echoic control, several varia-tions were implemented until the vocalizationincreased. There are, however, several draw-backs to a comparison design including the lackof experimental control necessary to concludethat the direct reinforcement procedure was re-sponsible for the behavior change (Bailey &Burch, 2002). In Experiment 2, Mary’s correctresponses to direct reinforcement trials in-creased during direct reinforcement Procedures4 and 5. The results suggest these procedurescould be beneficial in establishing echoic con-trol; however, further investigation is neededto identify which direct reinforcement proce-dures are most effective in establishing echoiccontrol over vocalizations.

A third limitation of this study, or more accu-rately identified as a potential limitation of thestimulus-stimulus pairing procedure, is the po-tential for setting the occasion for problem be-havior. In this study, for example, the controlcondition involved the experimenter saying thetarget sound several times and then waiting 20s before presenting a preferred stimulus. Fur-thermore, in the pairing procedure conditionthe experimenter repeatedly removed a preferredstimulus at the end of each trial. In this case apreferred stimulus is either delayed or removed,which could produce problem behaviors thathave a history of being reinforced by access totangible items. Future research should be con-ducted to verify this relation.

These experiments extend the current under-standing of the role of automatic reinforcementin the development of vocal verbal behavior.The results of this study provided preliminarysupport for the use of a stimulus-stimulus pair-ing procedure to increase an existing, but low-occurring vocalization for a child as young as 2years. Furthermore, and perhaps more impor-tantly, the results were obtained for the partici-pant who initially began with the most limitedverbal skills. The implications from these find-ings is potentially great as teachers and par-ents most likely teach alternative forms of

communication to children with the most lim-ited verbal skills. The pairing procedure couldconceivably be used in addition to alternativeforms of communication to increase a vocal-ization that could then be brought underechoic control.

REFERENCES

Bailey, S. J., & Burch, R. M. (2002). ResearchMethods in Applied Behavior Analysis.Thousand Oaks, CA: Sage Publications, Inc.

Esch, B. E., Carr, J. E., & Michael, J. (2005).Evaluating stimulus-stimulus pairing anddirect reinforcement in the establishment ofan echoic repertoire of children diagnosedwith autism. The Analysis of Verbal Behav-ior, 21, 43–58.

Higbee, T. S., Carr, J. E., & Harrison, C. D. (2000).Further evaluation of the multiple-stimuluspreference assessment. Research in Devel-opmental Disabilities, 21, 61–73.

Miguel, C. F., Carr, J. E., & Michael, J. (2002).The effects of a stimulus-stimulus pairingprocedure on the vocal behavior of chil-dren diagnosed with autism. The Analysisof Verbal Behavior, 18, 3–13.

Normand, M. P., & Knoll, M. L. (2006). Theeffects of a stimulus-stimulus pairing pro-cedure on the unprompted vocalizations ofa young child diagnosed with autism. TheAnalysis of Verbal Behavior, 22, 81–85.

Skinner B. F., (1957). Verbal Behavior.Englewood Cliffs, NJ: Prentice Hall.

Smith, R., Michael, J., & Sundberg, M. L. (1996).Automatic reinforcement and automaticpunishment in infant vocal behavior. TheAnalysis of Verbal Behavior, 13, 39–48.

Sundberg, C. A., Michael, J., Partington, J. W.,& Sundberg, M. L. (1996). The role of auto-matic reinforcement in early language ac-quisition. The Analysis of Verbal Behavior,13, 21–37.

Sundberg, M. L., & Partington, J.W. (1998).Teaching language to children with au-tism or other developmental disabilities.Pleasant Hill, CA: Behavior Analysts, Inc.

Vaughan, M. E., & Michael, J. (1982). Auto-matic reinforcement: An important but ig-nored concept. Behaviorism, 10, 217–227.

Yoon, S., & Bennett, G. M. (2000). Effects of astimulus-stimulus pairing procedure on con-ditioning vocal sounds as reinforcers. TheAnalysis of Verbal Behavior 17, 75–88.