1990, NUMBER 1 (SPRING 1990)

DISCRIMINATION TRAINING FOR PERSONS WITH DEVELOPMENTAL DISABILITIES: A COMPARISON OF THE TASKDEMONSTRATION

MODEL AND THE STANDARD PROMPTINGHIERARCHY

ALAN C. REPP, KATHRYN G. KARSH, AND MARK W. LENZ NORTHERN ILLINOISUNIVERSITY AND

EDUCATIONAL RESEARCHAND SERVICES CENTER, INC.

Acomparison wasmadebetweentwoprocedures for teachingpersons with severehandicaps: (a) thetask demonstrationmodel,which isbasedupon afading procedure and generalcase program-ming, and (b) thestandardprompting hierarchy, a least-to-mostintrusive promptingprocedure commonlyused to teach these individuals. Fivephaseswere used incomparingtheprocedures: pretesting, training, twogeneralization tests,anda6-monthmaintenance test.Eightstudents learned twodiscriminationtasksbyeachprocedure,with each taskinvolvingtwo- orthree-digitnumbers. Results showed that under the task demonstration model all 8 subjects had more unprompted correct responses(about 1.5 times as many) intraining,all 8subjectshadfewererrors(about0.6 times asmany) in training, all 8 subjectshad more correct responding inthegeneralization test withuntrained stimuli in the training room, 6 of8 subjects had more correct responding with untrainedstimuli in another room, a7thhadequivalent amounts,and 7 of8subjectshad more correctrespondingon a 6-month maintenance test. Thus,thetask demonstration model proved superior to thestandardprompting hierarchyin29of 32 testsofcorrectresponding. Resultsare

discussed intermsofimplicationsfor stimulus controltraining strategies.

DESCRIPTORS: mentally retardedstudents,fading,promptinghierarchy, taskdemonstration model, discriminationlearning

During the past 10 years, several curriculum programs have been developed to teach basic life skills to learners with severe mental retardation (e.g., Brown et al., 1979; Falvey,

1986;

Sailor&

Guess, 1983;Wehman,Renzaglia,&Bates, 1985).

Theseprograms haveallowed teacherstoorganize

and implementinstructionsothatindividualswith severe handicaps have the opportunity to acquire

functional skills for community living. Although

thesearegoodcurricula, educatorsandparentshave

been disappointed because these skills have not beenmaintained or generalizedacross stimuli and

settings

(Liberty,

1985; Voeltz & Evans, 1983). An analysisof functional skillsshows thatmost skills require correct discriminations among com-plex stimulus arrays; thus, the individual has to attend to multiple cues. Researchers have shown that persons with autism and mental retardation

ThisresearchwassupportedinpartbyGrantG008730035 fromthe U.S. Department of Education.

Please address correspondence and reprint requests to

Kathryn G.Karsh,Educational Research and ServicesCenter, 425 FiskAvenue,DeKalb, Illinois60115.

are often selective in their attention; that is, they often attendonly tosingle components ofa mul-tiple-component stimulus (Koegel & Wilhelm, 1973; Lovaas, Koegel, & Schreibman, 1980; Lo-vaas,

Schreibman,

Koegel,&Rehm, 1971).

There-fore, individualswithseverementalhandicapswho

do not

respond

to all the relevant components of a stimulus may make errors during training and generalization offunctional skills.

An instructional procedure characterized by an extrastimulus prompt to guide the individual to the right choice is the least-to-most intrusive promptinghierarchy (Snell, 1987;

Steinberg,

1988; Wolery, Bailey, & Sugai, 1988). It is commonly used to train persons with mental handicaps and is referred to as the standard (Mosk & Bucher, 1984) ortraditional(Schreibman,Charlop, & Koe-gel, 1982; Steege, Wacker, & McMahon, 1987) teaching method.Althoughutilizationdataonthis strategyare notavailable,itspopularityisillustrated

by an informal survey of 50 teachers of students with severe or moderate retardation. When asked howtheywouldteachasimplediscrimination, each 43

teacherdescribedsomevariationofaleast-to-most intrusive prompting hierarchy. This instructional procedureisattractive toteachers becauseit iseasy to implement andcan be used across a variety of tasks.

Although some researchers have reported the successofextrastimulusprompts(e.g., Brown,

Bel-lamy, Perlmutter, Sackowitz, & Sontag, 1972; Cuvo,Leaf,&Borakove, 1978;Gold,1972,1976; Koop, Martin,Yu,&Suthons, 1980), others have reported their ineffectiveness (Koegel &Rincover, 1976;Schreibman, 1975;Schreibmanetal., 1982).

Thefailure ofextrastimuluspromptsas ateaching strategy may be related to the selective stimulus control (Allen & Fuqua, 1985) exhibited by in-dividualswith severe handicaps. The introduction ofa single extraneous cue that is familiar to the student(i.e., afrequentlyusedprompt)may over-shadow thestudent'sattentiontothetraining

stim-ulus (Mackintosh, 1977). As a result, functional control may never be transferred from the extra-neous cue or prompt to the training stimulus (Schreibman, 1975; Schreibmanetal., 1982; Sid-man & Stoddard, 1966, 1967).

An alternative to an extrastimulus prompting systemisastimulusfadingprocedure inwhichthe training stimuliaresystematically changed so that

the individual can respond correctly without the

introduction ofextraneous cues (Terrace, 1963a, 1963b). Themanipulation of thetraining stimuli focuses the individual's attention on the essential

stimulus components necessary for the final

dis-crimination. Successful teaching procedures based

on this type ofstimulus manipulation have been

described bySchreibman(1975),Rincover(1978), Mosk and Bucher (1984), and Allen and Fuqua (1985). However, they have not been widely adoptedfordassroomuse.Onereasonmaybethat

someof theseprocedures canbetediousand time-consumingtoimplement and, as aresult, maynot

be the most parsimonious method for dassroom

instruction (Etzel & LeBlanc, 1979).

Analternative fadingprocedure,basedon with-in-stimulusmanipulations,has beendevelopedand

used inmanyteaching situations. Labeled the task

demonstration model or TDM (Deitz, Rose, &

Repp, 1986), it is based on a fading procedure

(Terrace, 1963a, 1963b; Sidman & Stoddard,

1966, 1967) in which the S+ and S- are si-multaneouslypresented while the S- becomes more like the S+ over trials. In addition to the fading procedure, irrelevant dimensions of S + and S-are presented so that the discrimination is made only according tothe relevant featuresof the S+. Based on general caseprogramming (Homer, Bel-lamy, & Colvin, 1984; Homer, McDonnell, & Bellamy, 1986), this procedure presents multiple examples ofS+ to ensure attention to the range

ofrelevant dimensions that characterizes the posi-tive stimuli. Because the irrelevant dimensions of S+ are varied across trials and are often found in

both S+ and S-, the learner cannot misinterpret themastherelevantdimensions. General case pro-gramming has been usedsuccessfully to teach per-sons with severe retardation how to use a soap dispenser(Pancsofar&Bates, 1985), purchase gro-ceries (McDonnell, Homer, & Williams, 1984), and cross streets (Homer, Jones, & Williams, 1985). Although we have been using TDM to teach persons with severe handicaps in classrooms and community settings, wehavemade noformal test

ofits efficiency. The purpose of this paper, then, is tocompare the task demonstration model with the standard prompting hierarchy (SPH), a pro-cedure that iswidelyimplementedbyteachers across avariety of conditions.

Critical features ofthe TDM indude multiple

examples ofS+ and S-, asamplingofthe irrel-evant dimensions, and a fading of the S- to be morelikethe S+ acrosstrials. Thecriticalfeatures of the SPH include a constant S + and S-, an extrastimulusprompting hierarchyprogressingfrom

least-to-most intrusive, modeling, and physical guidance. The test wasmadewith8subjects, each ofwhom learnedtwo number discriminations un-derthe TDM and two underthe SPH.

METHOD

Subjects

and Setting

Two females and 6 males with moderate or severe retardation served as subjects inthis exper-iment.Theyrangedinagefrom16.2to21.0years, and their I.Q. scores

(Stanford-Binet

or WAIS)

ranged from 30 to44. All werebeing educated in aself-contained school for individualswith devel-opmental disabilities. Each student had sufficient

attending and compliance skills to remain seated at a table for 20 min and to follow directions to look atandtouch the trainingstimuli. Noneof the students possessed significant visual, hearing, or motor impairments.

Experimental Design and General Procedures An alternating treatments design (Barlow & Hayes, 1979; Barlow& Hersen, 1984) was used to evaluate the effectiveness of the two teaching

procedures. Training conditions were

counterbal-anced acrosssubjectsandtasks,withatleast 1 day between treatment sessions. Tasks were

randomly

assignedtothetwo teaching

procedures

across the

five conditions (pretesting, training, two

general-ization tests, and a 6-month maintenance test).

Pretesting, training,andthefirst generalizationtest were conducted in the school's observation room.

The secondgeneralizationandthe maintenance tests wereconductedinthestudents' dassrooms.During

all phases, the trainer and

subject

sat

facing

each other separated by a desk, while an observer re-corded data. For20%ofthetrials,asecondobserver recorded data for

reliability

purposes.

STIMULUS MATERIA1S Training Stimuli: TDM

There were two setsof trainingstimulifor each two- or three-digit

number,

one representing the S+ stimuli andonerepresenting the S-. TheS+ was alwayspresented inits criterion format (shape and orientation),althoughthereweremultiple ex-amples of the S+ that differed with respect to dimensions irrelevanttothe

discrimination,

indud-ing size, script, color, texture,

background,

and

location of S+ on the table used for training. In theTDM,eachnumber appearedin various

com-binationsof 7 colors(black,white,gold, blue,red, orange,pencil), 15 sizes

(¼t6

in. to3/8 in.), 5 scripts

(handwritten, labels, newspaper, rub-on, printed stick-on), 3 textures (smooth, indented, elevated), 3

background

sizes (1.5 by 2.5 in., 3 by 5 in., 4

by

6in.),6

background

colors(pink, blue, yellow,

orange, green,white), and 3 locations on thetable (left, middle, or right third).

IntheTDM, the S- stimuliwerefaded during training so thattheybecame more and more similar to the S+ stimuli, until they differed only on the relevant dimensions. Initially, very different S-s were presented. These varied across the irrelevant dimensions and consisted ofsingle letters, single digits, four-letter nonsense words and four-digit numbers. Then, a set ofmoderatelydifferent S-s was presented. These consisted of two- or three-letter nonsensewords; two- orthree-digitnumbers judged by the experimenters to be moderately dif-ferent in shape; and two- or three-place letter/ number combinations, presented across the irrele-vantdimensions (e.g., when S + = 74,moderately different S-s wereH7,HI, 7W, etc., on different

backgrounds, withslightly differentsizes, etc.). Fi-nally, slightlydifferentS-s were presented. These

varied across the irrelevant dimensions and con-sisted ofnumbers similarto the S+ stimulus (e.g., when S+ = 345,slightly different S-s were 543, 346, 445, etc.). By the end of the last step, only the relevant dimensions (i.e., the shape and ori-entation ofthenumbers)distinguishedthe S+ and S-. The irrelevantdimensionswereoftenidentical onanysingle trial.

Training Stimuli: SPH

In this condition, the S- numerals induded

bothhigherand lower consecutivenumbersto S +

(e.g., 51 and 53 for 52) and a single reversal of

S+ (e.g., 25for52). All SPHmaterials(S+,S-) consisted ofthenumber handwritten witha black marker pen on white cards (3 by 5 in.). We

at-temptedtoequate the taskdifficulty ofTDM and SPH by selecting numerals that were similar in shape (topographicalfeatures),similar in sequential

order,and similarinquantityof"easy" and "hard" numericaldiscriminations (3 and 8, 1 and 7, 6 and

9).

PROCEDURES

Pretestingand Selection of Training Materials

Eachsubjectwas trained on four discrimination tasks of equal difficulty, each involving two- or

three-digitnumbers. Thesetaskswere chosenfrom individualized educational program (IEP) goals de-signed to teach these students to identify addresses, grocery prices, and so on. The pretestwas used to select two numbers to be taught by the TDM and two to be taught by the SPH for each person. Each

subjectwaspresented with 10 trialsforeach

num-berinwhich one S+ (e.g.,92) and oneS- (e.g., 91) were presented, and the subjectwas asked to "touch the number 92." There were no contin-gencies for correct or incorrect responses. If the

subjectpassed the test (> 50% correct) for a given

number,adifferentnumberwas tested;ifthe iden-tification test was failed (' 50%), the number was

selectedas an S+ to betaughtinanidentification

task.

From the pretest, four two-digit (92, 61, 49,

and 76) or three-digit (345, 736, 341, and 643) numberswere chosen as the final or criterion S+ stimuli for training and generalization. Although

thesubjects may havehad someexposure to these numerical sets, the pretest probes indicated that none hadsufficientexperience to help themidentify

the S+ at more than chance (50%) level. The criterion S- numerals were both the higher and lower consecutive numerals to each S+ (e.g., 91

and93 when S+ was 92) and reversals of the S+ (e.g., 29 when S+ was 92).

Training

Thesecondphase consistedoftraining each

sub-ject to

identify

four

numbers,

two ofwhich were taught byeach ofthetwo procedures. Correct re-sponses resulted in verbal praise (e.g.,

"Good,"

"That'sright") for all

subjects;

4 ofthe

subjects

also receivedpennies (whichwereusedasreinforcers in the students' vocational tasks) for correct re-sponding. This change was madebecausethefirst 4studentsbecame lessresponsiveas asession

(typ-ically 20 min) progressed.

Task demonstration model. During TDM, identification was taught inthreestepsthatvaried in the difficulty ofthediscrimination to be made. Initially, examples of the S- were characterized as very differentand varied from the S+ in the one relevant dimension(shape of the stimulus; e.g., 92 vs. G) as well asfouror fiveirrelevant dimensions

(script, color ofS+, colorof background, texture of S+, and location ontable). Whencriterion was met (correct responding on allofthe last five and 9 ofthe last 10 trials), moderately different S-s were introduced. These usually varied on two or three irrelevant dimensions as well as the S+. Fi-nally, slightly different S-s were introduced. In this set, the final discriminations were made

be-tween avery similar S+ and S- (e.g., 92 and 93) that had identicalirrelevant dimensions (e.g., color, size, background).

For all trials, thesame instruction was given to eachstudent, thatbeing "Touch the ." Atfirst,

the S+ was placeddose to the student while the S- was placed farther away. Within afew trials,

bothstimuli wereplaced equidistantfromthe stu-dent. Each correct trial was followed by a change in either S+ or S-.

Any incorrect ordelayed (10 s) response during TDM training was followed by a correction se-quence that beganwith averbalcommand ("No, touch theother one" or "No, touch the ") and physicalguidance to the correct stimulus. This was

followedby repeatedinstructionwith thematerials

left intact. If the subject responded correctly, the stimuli wereremoved and then returned to the same position. A correct response wasfollowedby another

trial withthe same stimulibutwith theirpositions

changed. Anothercorrectresponse wasfollowedby

two more trials, with the position of the stimuli

changed each time. Successfulresponses to the en-tirecorrection procedure resultedina return to the next trial in the normal training sequence. Three consecutive errors or three errorswith the sameset ofstimuliresultedinareturn totheprevious stim-ulus set onwhich the

subject

had responded suc-cessfully.

Standard prompting hierarchy. TheSPH is a

recommended teaching procedure (Snell, 1987;

Sternberg, 1988; Woleryetal., 1988)that uses a hierarchy of least-to-most intrusive prompts to teach thestudenttomakecorrectdiscriminations such as thosemadein TDM.

During training, theexperimenter simultaneous-ly placed one S + and one S- card next to each other on atable directlyinfront of the student and asked the student to "Touch the ." A correct

unprompted trial resultedinanothertrial, whereas anincorrectresponseresultedin the least intrusive prompt (pointingtothe S+)and repetition of the

instruction. Ifa correct response was notmade at this level ofprompting,successive levelswereused untilcorrectresponding occurred. The fill prompt

sequence was as follows: (a) instruction only, (b) instructionplus pointing, (c) instructionplus mod-eling (the experimenter touched S+), (d) instruc-tionplusphysicalprompt(theexperimentertouched

the back of the student's hand) and, finally, (e) instruction plus physical guidance (the experi-menterprovided the leastamountof direct manual guidance neededto obtaina correctresponse).

Generalization and Maintenance

Onthesameday thatastudentreached criterion during the training phase, twogeneralization tests were given. The first test was conducted in the training room and used examples ofthe S + and S- that differed fromthe numerals used intraining along irrelevant dimensions (e.g., script, color, size). Thistestwasrepeated laterinthedayinthe dass-room. Thegeneralizationtestinthedassroomwas repeated 6 months after training. Nofeedback (in-duding prompts)wasgiven duringthe 10 trials in either of the two generalization conditions or the maintenance test. Each trial ended after a correct response, an incorrect response, or 10 s in which noresponding occurred.

DATA COLLECTION Dependent Measures

Alldatawerecollected byanobserver whowas seated near the trainer and student. Dependent measuresindudedunpromptedcorrectresponses(a correctresponsemade within 10 s), prompted cor-rectresponses,anderrors. These data alsoprovided information on the number of training trials to

criterionfor each condition.

InterobserverAgreement on Dependent and Independent Measures

For 20% ofthe sessions, a second observer re-corded data on student responding. Percentage agreementonthe threecategoriesof thedependent

measuresvariedfrom97% to 100% and averaged

99%.

Tobe certain that both the TDMandSPHwere

being used correctly, data were also collected in

20%ofthe sessions on the implementation ofthe independent variables. For the SPH, the observer marked whether a prompt was given and which of thefourlevels wasused.For the TDM, theobserver

determined fidelity by recording whether the ap-propriate S+ and S- stimuli were used, whether the location of the items on the desk was correct,

andwhether theerror-correctionprocedure was

fol-lowed. Percentage agreement on the implementa-tionof SPH was 100%. For the TDM, percentage agreement on implementation ranged from 98%

to 100% and averaged 99%.

RESULTS

Figures 1 and 2 show the percentage of un-prompted correct responses during training under the twoteaching conditionsforthe4studentswho werepraised(Figure 1) and the4who werepraised andreceivedpennies (Figure 2). Because there were 8 subjects, there wereeight opportunities tomake

the comparison between the two teaching proce-dures. In each of the eight comparisons during training, the TDM resulted in more unprompted

correct responses (UCR). For the first 4

subjects

whowerepraisedfor unpromptedcorrect responses, themeansforthegroupunder theTDMandSPH were 77% and 50%, respectively. For the other 4

students, who received praise as well as pennies, thegroup means were 75% and 51%. The mean percentages across thefour discriminationslearned

by

each subject under the TDM and SPH were,

for

Subject

1, 68% and 45%; forSubject 2, 91%

and59%;for

Subject

3, 77% and 47%;forSubject

4, 71% and 48%; for Subject 5, 90% and 62%; forSubject 6, 67% and 52%; forSubject 7, 78% and 40%; and for Subject 8, 65% and 48% for TDM and SPH, respectively.

Figures 3 and 4 show the percentage oferrors

that occurred under these two procedures. Data from the training phase again showeddear differ-encesbetweenthe twoprocedures.Theoverall means forthefirst and second set offoursubjects for the

I

a

I

I

100 [Tanm TakDfohalono 50

0

F%/I

100 50 Ron | John

1

S

0-I

I

I

I

I

I

|Steve

I

Ic I5

f

t//i,

Figure 1. The percentage of unprompted correct re-sponsesby subjectsunderthetaskdemonstration model and the standard prompting hierarchy. Phases were baseline, training, generalization in training room, generalization in classroom, and maintenance.

TDMand SPH were 12%(TDM) and25%(SPH), and 16% (TDM) and 25% (SPH). Results from the individual

subjects

showed that every one of

theeight comparisonsfavoredthe TDM. Theerror percentages forthe 8

subjects during

trainingwere

15% and 27% for

Subject

1, 4% and 20% for

Subject

2, 11% and 27% for

Subject

3, 16%and

26% for Subject 4, 6% and 20% for

Subject

5, 23% and 24% for

Subject

6,

13% and 30% for

Subject 7, and 23% and 26% for

Subject

8 for TDM and SPH,

respectively.

These four figures

show that, during the training phase, the TDM producedahigherpercentage ofunprompted cor-rectresponses and a lowerpercentage oferrors.

The data were then analyzed for the two gen-eralizationconditions,onethatpresented untrained examplesin thetraining room(TG) and one that

presenteduntrained examples in the classroom (CG).

Duringbothgeneralizationphases,nofeedbackwas

D-avidl100= |Shaun|

50i

0

1%/

Figure 2. The percentage of unprompted correct re-sponsesby subjectswho were praised and received a penny for eachcorrectresponseduringtrainingunderthetask dem-onstrationmodel and the standardpromptinghierarchy.Phases werebaseline,training,generalizationintrainingroom, gen-eralizationinclassroom,and maintenance.

givenfollowingcorrect orincorrectresponding(i.e., no praise, pennies, or prompts). Thus there were onlytwocategoriesofresponding: unprompted cor-rectresponses and errors. Because these two num-bers must sum to 100% (unlike thetrainingdata

forwhich there were three types ofresponses: un-prompted correct, prompted correct, and errors), onlyonecategory isreported. The data fromFigures 1 and 2 show that in all eight cases, responding was superior in the training setting following the TDM condition. The overall means of correct

re-sponding across the 8 subjects for the TDM and SPH were 79% and 54%, respectively. Meansfor Subject 1 were40% and 35%; forSubject 2, 85% and50%;for Subject 3, 75% and40%;forSubject 4, 75% and 45%; for Subject 5, 85% and 60%;

forSubject 6, 75% and 55%; forSubject7, 100% and 65%; and for Subject 8, 90% and 85% for

/ ,/

4f

451 cjp rf

100ITwmy 1 STaukRuoshion n o1_

50-

50

'a 0 /

f-01

ji

<///

/

/

h o

L11100

50 50 n 0.

/

f/

Figure 3. The percentage oferrors by subjects taught

discriminationsunder the taskdemonstrationmodel andthe

standardpromptinghierarchy.Phases weretraining,

gener-alizationinthetrainingroom,generalizationintheclassroom, andmaintenance.

TDM andSPH, respectively. For the 10 general-ization trials per task delivered in the dassroom, TDMwasthe superior procedureinsixof theeight casesandequivalentinoneother.Theoverallmeans were 84% (TDM) and 62% (SPH), and the in-dividualmeanswere60% and 60% for Subject 1, 75% and 35% for Subject 2, 55% and 40% for Subject 3, 80% and 70% forSubject 4, 95% and 45% for Subject 5, 65% and 55% for Subject 6, 75% and 40% forSubject 7, and 80% and 90% forSubject 8 forTDM and SPH,respectively. At the end of 6 months, 10 maintenance trials for each taskwerepresentedtothesubjectsindividually

in their dassrooms. The mean percentage of un-promptedcorrecttrialsacrosssubjectswas74%for TDM and 58% for SPH. Individual means for Subject 1were43% and40%6; forSubject 2, 85% and73%; forSubject 3,78%and55%;;forSubject

4, 83% and 55%; for Subject 5, 75% and 53%; for Subject 6, 70% and 60%; forSubject 7, 75%

Alex

Iiohn 100 I M

II

100 I To* _-,hd 50 -50-r* 0 0L 1100_ 100 i 50

Figure4. Thepercentageoferrorsby subjectswhowere

praisedandreceivedapennyfor eachcorrectresponseduring

training.Phasesweretraining,generalizationin thetraining

room,generalizationintheclassroom, andmaintenance.

and

40%;

and for

Subject

8, 83% and 85% for TDM and

SPH,

respectively.

Ameasureofefficiency,the totalnumber of trials

tocriterionineachcondition,alsofavoredtheTDM

for most ofthe subjects. The average number of trials for theTDM conditionwas 102;for the SPH

condition,

137 trials. For 5 of the subjects, the

TDM resultedinfewer trials, with 2 of those

sub-jects requiring more than twice as many trials in the SPHconditiontoreach criterion. Trialsto

cri-terion for individual

subjects

were 142 and 194 for

Subject

1, 70 and 44 for Subject 2, 81 and 247 forSubject3, 129 and 142 for Subject4,

69

and 84 forSubject 5, 145 and 129 forSubject

6,

97 and 194 for Subject 7, and 81 and 61 for

Subject 8 for TDM and SPH, respectively.

DISCUSSION

Theeffects oftwo procedures on teaching four

discriminations were compared. The task

demon-OL

leI

Af.'f W.

/

,tf

/

"el Cf ef

stration model (Deitz et al., 1986) relied primarily on a fading sequence in which multipleexamples of S- were made successively more like multiple

examplesof S+. The standard prompting hierarchy wasbased on a commonly used procedure for teach-ing persons with retardation in which successively

more intrusive prompts were used following in-correct responding.

Althoughtheusualcontrol forstudies offading

is a series of unprompted trials in which correct responses arereinforced andincorrect ones are ex-tinguished (see Lancioni & Smeets, 1986), this procedure was notused in our study for two reasons. One is that prior research suggests that the control procedure isweakerthan the experimental one, so the test might be too easy and trivial. A second

reason is that wewantedtocompare two teaching procedures, one that is based on stimulus fading and is untested and one that is based on response prompting and adoptedbyteachers as apreferred method. Results were compared in the training condition, generalization in thetraining room and classroom, and maintenance. Of 32 comparisons

ofthe procedures(8 subjects X 4comparisons of

training,generalization, and maintenance), thetask

demonstration model proved superiorin 29. Thiscomparison of the task demonstration mod-el and the standard prompting hierarchy concurs with the available research on the effectiveness of

stimulusfading

procedures

whentheyare

compared

toextrastimulus prompting

procedures.

However, thetask demonstration modelconsistsofanumber

ofcomponents(i.e.,

fading

the S- tobemorelike the S+, multiple examples of S+ and S-, an error-correction procedure) not

duplicated

in the SPH. Which oftheseprocedural differences made thegreatestcontributiontothe successofthe TDM is notevidentfromthis investigation. In anearlier

study, Schreibman and Charlop (1981) reported fadingtheS+ tobemoreeffectivethan

fading

the S-. If the superiority of fading S + was due to stimulusnovelty and its effects on stimulus selection as thoseauthors have suggested, then the presen-tationof multiple examples ofS + over trials may haveproduced a similar novelty effect and increased thesalienceof S+ inthe TDM.Multiple examples

of S+ and S- have proven effective in teaching persons with severe handicaps, although a detailed analysis has not been developed to indicate how the positive and negative examples should be ar-ranged (Albin & Homer, 1988). In the TDM, the S - examples were faded to become increasingly like the S+ examples. Further research is needed to determine the combined effect of fading and multiple examples when compared to the separate fading and multiple-exemplar procedures.

Inthis study, the TDM wasbothmore effective and efficient in the majority of cases. Formal in-vestigationsofthe external validityofthe procedure

fordassroom use are necessary. Although we have used the procedure across a variety of tasks, no

formaltests havebeenmadeofitsapplication across

tasks, modalities, materials (two or three dimen-sional), or simultaneous versus successive discrim-inations. Although the initial implementation of

the TDM may be less efficient than standard prompting proceduresfromtheaspectofpreparing materials orcollecting multiple examples, we have

foundthat thisprocedurecanbe usedwith stimuli

from the natural environment and that S + ma-terials for one task can become S- materials for

othertasks.

Whether individuals will learn discriminations taught withtraditionalproceduresmore

effectively

followingexposure to the task demonstration mod-elshouldbe addressed. Autisticchildrenwereable torespondcorrectly to stimuliduringan extrastim-ulusprompting procedure

following

trainingwith a

fading

procedure

(Schreibman

etal., 1982). The TDM procedure may produce increased attention tomultiplecues or tocriticalfeatures

distinguishing

S+ from S- that maygeneralize to avariety of

teaching situations and complex discriminationsin

the naturalenvironment.

REFERENCES

Albin, R. W., & Homer, R. H. (1988). Generalization withprecision. In R. H. Homer,G. Dunlap, & R. L.

Koegel(Eds.), Generalization andmaintenance: Life-stylechanges in appliedsettings (pp. 99-120).

Balti-more: Paul H.Brookes.

stimulus control: A comparison oftwo procedures for teaching mentally retarded children torespond to

com-pound stimuli.Journal of Experimental Child Psy-chology, 39, 55-71.

Barlow, D. H., & Hayes, S. C. (1979). Alternating

treat-mentsdesign: One strategyforcomparingtheeffects of two treatments in asinglesubject. Journal of Applied BehaviorAnalysis, 12, 199-210.

Barlow, D. H., & Hersen, M. (1984). Singlecase exper-imentaldesign:Strategiesforstudying behaviorchange (2nd ed.). NewYork: Pergamon.

Brown, L., Bellamy, T., Perlmutter, L., Sackowitz, P., &

Sontag, E. (1972). Thedevelopment of quality, quan-tity,anddurabilityintheworkperformance ofretarded students in a public school prevocational workshop. TrainingSchoolBulletin, 68, 58-69.

Brown, L.,Branston, M. B.,Hamre-Nietupski,S.,Pumpian, I., Certo, N., &Gruenwald,L. (1979). A strategyfor developing chronologicalage-appropriateandfunctional curricularcontent for severely handicapped adolescents and young adults.Journal of Special Education, 13, 81-90.

Cuvo,A. J.,Leaf,R.B.,&Borakove,L. S. (1978). Teach-ingjanitorialskillstothementallyretarded:Acquisition, generalization,andmaintenance.JournalofApplied Be-haviorAnalysis, 11,345-355.

Deitz, D. E. D., Rose, E., & Repp, A. C. (1986). The taskdemonstration model for teaching severely hand-icappedpersons. DeKalb, IL: Educational Research & Services Center, Inc.

Etzel, B. C., &LeBlanc, J. M. (1979). The simplest treat-mentalternative: The lawofparsimonyappliedto choos-ing appropriateinstructional controland errorless-learn-ingproceduresforthedifficult-to-teach child. Journalof Autism and DevelopmentalDisorders, 9, 361-382. Falvey, M. A. (1986). Community-basedcurriculum:

In-structional strategiesfor students with severe handi-caps. Baltimore: Paul H. Brookes.

Gold,M. W. (1972). Stimulusfactorsinskill trainingof the retarded on acomplex assembly task: Acquisition, transfer, and retention. AmericanJournal ofMental

Deficiency, 76,517-526.

Gold,M. W. (1976). Taskanalysis ofacomplexassembly task bytheretarded blind. ExceptionalChildren, 43, 78-84.

Homer, R. H., Bellamy, G. T., & Colvin, G. T. (1984). Responding in the presence ofnontrained stimuli: An appliedanalysisofgeneralization. Journalofthe Asso-ciationfor Persons with Severe Handicaps, 9, 287-296.

Homer, R. H., Jones, D. R., &Williams, J. A. (1985).

Afunctional approach to teaching generalized street cross-ing. JournaloftheAssociationforPersons with Severe Handicaps, 10,7-20.

Homer, R. H.,McDonnell,J. J.,&Bellamy, G. T. (1986). Teachinggeneralized skills: General case instruction in simulation and community settings. In R. H. Horner, L. H. Meyer, & H. D. Fredericks (Eds.),Education of learners with severe handicaps: Exemplary service strategies (pp. 289-314). Baltimore: Paul H. Brookes.

Koegel, R. L.,&Rincover, A. (1976). Somedetrimental effects of usingextrastimulitoguidelearninginnormal andautisticchildren. Journalof Abnormal Child Psy-chology, 4, 59-71.

Koegel,R.L.,&Wilhelm,H. (1973). Selectiveresponding to the components ofmultiple visual cues by autistic children.Journalof Experimental ChildPsychology, 15, 442-453.

Koop,S., Martin,G.,Yu,D.,&Suthons,E. (1980). Com-parisonoftworeinforcementstrategies invocational-skill trainingofmentally retardedpersons.American Journal ofMentalDeficiency, 84,616-626.

Lancioni, G. E., &Smeets, P. M. (1986). Proceduresand parametersoferrorlessdiscrimination training with de-velopmentallyimpaired individuals.InN. R.Ellis&N. W. Bray (Eds.), International review ofresearch in

mentalretardation (Vol. 14, pp. 135-163). Orlando, FL:AcademicPress.

Liberty, K. A. (1985). Enhancinginstructionfor mainte-nance,generalization,andadaptation. InK. C.Lakin&

R. H. Bruininks (Eds.), Strategiesfor achieving com-munityintegrationofdevelopmentallydisabledcitizens (pp.29-71). Baltimore: Paul H.Brookes.

Lovaas, 0. I., Koegel, R. L., & Schreibman, L. (1980). Stimulusoverselectivityinautism: Areviewofresearch. Psychological Bulletin, 86, 1236-1254.

Lovaas, 0. I., Schreibman, L.,Koegel,R. L., &Rehm,R. (1971). Selectiverespondingbyautisticchildrento mul-tiplesensory input. Journalof Abnormal Psychology, 7,

211-222.

Mackintosh, N.J. (1977). Stimuluscontrol: Attentional factors. In W. K. Honig &J. E. R. Staddon (Eds.), Handbook ofoperant behavior (pp. 481-513). Engle-woodCliffs, NJ:Prentice-Hall.

McDonnell,J.J., Horner, R. H., &Williams,J.A. (1984). Comparisonofthreestrategies for teaching generalized grocerypurchasingto highschoolstudentswith severe handicaps. Journalofthe AssociationforPersonswith SevereHandicaps, 9, 123-133.

Mosk,M.D.,&Bucher,B. (1984). Promptingand stim-ulusshaping proceduresfor teachingvisual-motorskills toretardedchildren.JournalofApplied Behavior Anal-ysis, 17,23-34.

Pancsofar, E. L., & Bates, P. (1985). The impact ofthe acquisitionofsuccessive training exemplarson general-izationbystudents with severehandicaps.Journal ofthe Association for Persons with Severe Handicaps, 10, 50-63.

Rincover, A. (1978). Variables affecting stimulus-fading and discriminative responding in psychotic children. JournalofAbnormalPsychology, 87, 541-553. Sailor, W., & Guess, D. (1983). Severely handicapped

students: An instructional design. Boston: Houghton MifflinCo.

Schreibman,L. (1975). Effects of within-stimulus and ex-tra-stimulusprompting ondiscrimination learningin au-tisticchildren. Journalof Applied Behavior Analysis, 8,91-112.

Schreibman,L.,&Charlop,M.H. (1981). S+versus S-fadingin prompting procedures with autistic children.

JournalofExperimental Child Psychology, 31, 508-520.

Schreibman, L., Charlop, M. H., & Koegel, R. L. (1982). Teaching autisticchildren to use extra-stimulus prompts. Journalof Experimental Child Psychology, 33,

475-491.

Sidman, M., & Stoddard, L. I. (1966). Programming, perception,andlearningforretarded children. In N. R. Ellis(Ed.), Internationalreviewof research inmental retardation (Vol. 2, pp. 151-208). New York: Aca-demic Press.

Sidman,M., &Stoddard, L. I. (1967). Theeffectiveness of fadinginprogramming a simultaneous form discrim-inationfor retardedchildren.Journal of the Experimen-talAnalysis of Behavior, 10,3-15.

Snell, M. E.(Ed.). (1987). Systematicinstruction of per-sons withseverehandicaps (3rd ed.). Columbus, OH: CharlesE. Merrill.

Steege, M. W.,Wacker,D.P., &McMahon, C. M. (1987). Evaluationoftheeffectiveness andefficiency oftwo stim-ulus prompt strategies with severely handicapped stu-dents.JournalofAppliedBehaviorAnalysis,20, 293-299.

Stemnberg,L. (Ed.). (1988). Educating students with se-vere orprofound handicaps. Rockville, MD:Aspen.

Terrace, H. (1963a). Discrimination learning with and without errors. Journalofthe ExperimentalAnalysis of Behavior, 6, 1-27.

Terrace, H. (1963b). Errorless transfer of a discrimination across twocontinua.Journal of the Experimental Anal-ysisof Behavior, 6, 223-232.

Voeltz, L. M., & Evans,I. M. (1983). Educationalvalidity: Procedures to evaluate outcomes in programs for severely handicapped learners. Journal of the Association for Personswith SevereHandicaps, 8, 3-15.

Wehman, P., Renzaglia, A., & Bates, P. (1985). Func-tionallivingskills for moderately andseverely hand-icapped adults. Austin, TX: Pro-Ed.

Wolery, M., Bailey, D. B., & Sugai, G. M. (1988). Ef-fective teaching: Principlesand procedures of applied behavior analysis with exceptional students. Boston: Allynand Bacon, Inc.

ReceivedMarch 21, 1988

Initial editorial decisionJuly6, 1988

Revisions received October 14, 1988; January 10, 1989 Final acceptanceJuly 12, 1989