The Analysis of Verbal Behavior 1991, 9, 19-28

The Effects of Specific Versus NonspecificReinforcement on Verbal Behavior

Steven J. BraamNew Medico Rehabilitation Center of Wisconsin

AND

Mark L. SundbergSundberg and Associates, Concord, CA

The current study is a systematic replication and extension of previous research on the differ-ences between specific (mand) and nonspecific (tact) reinforcement. The focus was on the rolethat these different consequences played in the acquisition of verbal behavior. Using both awithin-subject and a between-subjects design, the current researchers trained eight essentiallynonverbal individuals to tact a variety of foods under two different reinforcement conditions.The results showed no significant differences between the four matched-pairs in rates of acqui-sition, or in the resistance to extinction. However, subjects in the specific reinforcement condi-tion emitted more untrained mand-compliance responses, while subjects in the nonspecificgroup demonstrated increased generalization to multiply controlled mand conditions. Theresults supported previous findings which indicated that the two types of consequences wereequally effective in the acquisition of tacting, but each had unique features and implications forlanguage training with nonverbal populations.

Skinner (1957) proposed that a distinc-tion be made between the mand and thetact because these two operants involvedifferent types of controlling variables. Themand is evoked by establishing operations(EOs) and followed by specific conse-quences, while the tact is evoked by non-verbal discriminative stimuli (SDs) and fol-lowed by nonspecific, or generalizedreinforcement. The conceptual distinctionsSkinner made between these verbal oper-ants seem clear. However, there is very lit-tle experimental work that identifies thedifferences in the evocative or repertoire-

This research was part of a study conducted by thefirst author in partial fulfillment of the requirementsfor a Ph.D. degree at Western Michigan University.Portions of this paper were also presented at the 1990meeting of the Association of Behavior Analysis:International, Nashville, TN. The authors thankCassandra Braam, R. Wayne Fuqua, Jack Michael,Paul Mountjoy, and Thelma Urbick for their com-ments on an earlier version of the manuscript.Address correspondence and reprint requests toSteven J. Braam, New Medico Rehabilitation Center ofWisconsin, 1701 Sharp Road, Waterford, Wisconsin53185.

altering effects of these separate indepen-dent variables (Michael, 1982, 1983, 1988).The distinction between establishing oper-ations and discriminative stimuli, theantecedent side of these two verbal oper-ants, has received the most amount ofattention. Several empirical studies havenow demonstrated differences in theevocative effects of the EO versus the SD(e.g., Carroll & Hesse, 1987; Lamarre &Holland, 1985; Hall & Sundberg, 1987;Sigafoos, Reichle, Doss, Hall, & Pettitt,1990; Sundberg, San Juan, Dawdy, &Arguelles, 1990). This research has alwresulted in several practical applications.However, the consequences of the mand

and the tact have received less conceptualand experimental attention, hence, less isknown about their differential strengthen-ing effects. The first known attempt todirectly examine the effects of specific ver-sus nonspecific reinforcement was con-ducted by Saunders and Sailor (1982).These researchers, while studying recep-tive language (mand compliance), pre-

19

20 STEVEN J. BRAAM and MARK L. SUNDBERG

sented developmentally disabled (DD)children with toys that appeared to func-tion as reinforcers. Two toys (training stim-uli) were presented and the subjects wereasked to point to each one when named bythe experimenter. In the specific reinforce-ment condition the subjects' correctresponse was followed by an opportunityto play with the toys to which they hadpointed. In the nonspecific reinforcementcondition, subjects' correct responses werefollowed by an opportunity to play withtoys that were not training stimuli. Therewas also a variable reinforcement condi-tion where, following a correct response,subjects were given either specific or non-specific reinforcement. The results showedthat the percentage of correct responseswere greater in the specific reinforcementcondition than in either the nonspecific orthe variable reinforcement condition.Although Saunders and Sailor (1982)

studied the use of these types of reinforce-ment with mand-compliance behavior, theapplication of the techniques to verbalbehavior training has also been conducted.In reviewing the literature regarding tacttraining with DD individuals, it appearedthat there were two general teaching pro-cedures differentiated by the relationshipbetween discriminative and reinforcingstimuli. One procedure is typified by usingreinforcers bearing no specificity to the dis-criminative stimuli evoking the responses.For example, when shown a cup or a pic-ture of one and a learner said "cup," ateacher delivered a small and quickly con-sumable edible reinforcer, such as M&Mcandies (e.g., Faw, Reid, Schepis,Fitzgerald, & Felty, 1981). In a secondapproach to tact training, reinforcers spe-cific to the discriminative stimulus weregiven contingent on correct responses. Inthis approach, when shown a ball and alearner responded correctly (said "ball")s/he received the object to play with, or inthe case of edibles, the food to eat (Carr,Binkoff, Kologinsky, & Eddy, 1978). Inanother variation of this procedure alearner was shown a picture of an object orevent as the stimulus, and following a cor-rect response, the actual object or event

depicted by the picture was delivered tothe learner (Carr et al., 1978). All of theseprocedures resulted in increased tacting,however they have not been comparedwith each other.

In an attempt to further examine thesetwo types of consequences, Stafford,Sundberg, and Braam (1988) replicated thebasic procedure used by Saunders andSailor (1982), but focused on mands andtacts rather than mand-complianceresponses. A within-subject multielementdesign was used to study the acquisition oftwo separate complex signed mand/tactresponses with a developmentally disabledyouth. In both conditions stimuli whichalso functioned as reinforcers were shownto the subject and placed in the condition-respective container (a cup in one condi-tion and a bowl in the other). At the sametime the experimenter asked the youth,"What do you want?" Following a correctresponse (e.g., food-blue-cup-on-table) in thespecific reinforcement condition, the exper-imenter gave the youth the reinforcer inthe cup. In the nonspecific reinforcementcondition, following a correct response(e.g., soda-green-bowl-under-chair) the exper-imenter delivered a reinforcer differentfrom that in the bowl.

Stafford et al. (1988) found no differencein the percentage of correct respondingacross the increasing levels of difficultybetween the two conditions. However,they found a shorter response latency inthe specific reinforcement condition.Furthermore, the subject chose the specificreinforcement condition more frequentlywhen allowed to choose between trainingconditions. Although the data demon-strated that responses trained with specificreinforcement were not acquired morerapidly than responses trained with non-specific reinforcement, the differences inlatency and condition preference indicatedthat such training may affect verbal reper-toires in unique ways.The current study was designed to

further investigate these two types ofreinforcement procedures. Since theStafford et al. (1988) study was conductedwith a single subject it seemed important

SPECIFIC VERSUS NONSPECIFIC REINFORCEMENT 21

to replicate the procedure using severalsubjects. In addition, a different researchmethodology, along with three new depen-dent variables, was used. Using both awithin-subject and a between-subjectsdesign, eight essentially nonverbal individ-uals were trained to tact a variety of foodsunder the two different reinforcement con-ditions. Of particular interest were themeasures of maintenance of previouslyreinforced responding during extinction,the generalization of trained responses todifferent antecedent conditions, and theemergence of untrained mand-compliancerelations.

METHOD

Subjects

Eight young adults, ages 18 to 24 years(mean=22), served as subjects. Stan-dardized testing conducted with theStanford-Binet (Terman & Merrill, 1973)and the Vineland Social Maturity Scale(Doll, 1965) indicated their IQs were below25. They were all diagnosed SeverelyMentally Impaired. The subjects were allambulatory, toilet-trained, and self-feed-ing. They resided in the community, eitherwith their families or in group foster care,

and attended daytime special educationprogramming.These individuals were chosen to partici-

pate because of their extremely limitedrepertoires. Isolated examples of mands,tacts, and mand compliances (pointing tonamed objects) were observed. However,these behaviors were under poor stimuluscontrol, and occurred at a very low fre-quency rendering the subjects essentiallynonverbal. Subject characteristics are pre-sented in Table 1. Prior to the start of thisstudy the subjects received no systematiclanguage training either vocal or signed.

Setting and Apparatus

The study was conducted at a large spe-cial education center. All sessions wereconducted in a 4' X 6' partitioned area ofthe subjects' classroom. The experimenter(first author) worked daily with each of theeight subjects on an individual basis. Tothe experimenter's left was a table onwhich training stimuli were displayed ormanipulated. Also, to the experimenter'sleft was a small bookshelf used to storesubsequent training stimuli out of sightduring training or testing with anotherstimulus.

Table 1

Subject characteristics.

Subject Sex Age Mand Tact Mand Compliance Matched Pair/Condition

BL M 25 None None None 1 /Specific

JN M 21 None None None 1 /Nonspecific

JD M 19 None None None 2/Specific

FD M 18 Weak Weak Weak 2/Nonspecific

ML M 19 Weak Weak Weak 3/Specific

ME M 25 Weak Weak Weak 3/Nonspecific

KH M 24 Weak Weak Weak 4/Specific

MN F 25 Weak Weak Weak 4/Nonspecific

22 STEVEN J. BRAAM and MARK L. SUNDBERG

Procedures

A matched-pair design was used.Between-groups and between-subjects sta-tistical analyses of the number of tact train-ing trials to criterion were made using A-tests. In addition, within-subject andbetween-groups visual analyses weremade across behaviors (mand, tact, andmand compliance) in post-training probes.Assignment to treatment groups. Subjects

were matched according to their ages andfunctioning levels. This procedure yieldedfour matched-pairs of subjects. Using arandom numbers table, one subject of eachpair was assigned to an experimental con-dition. That subject's corresponding pairwas then automatically assigned to theopposite condition. For example, if the firstsubject of a pair was assigned to the spe-cific reinforcement condition, his/her pairwas automatically assigned to the nonspe-cific reinforcement condition.The next subject to be assigned was cho-

sen at random, but the condition was pre-determined by the previous pair's assign-ments. That is, if the first subject chosen atrandom went to specific and his pair wentto nonspecific, then the next subject chosenat random automatically went to the non-specific condition and his pair automati-cally would go to the specific condition.This procedure ensured the equal assign-ments of subjects to conditions.

Reinforcer sampling. Foods used for train-ing stimuli were demonstrated functionalreinforcers for the subjects in classroomactivities. To assess relative levels of depri-vation and satiation and also to provide abasis for selecting training stimuli 10 two-phase reinforcer sampling sessions wereconducted with each subject prior to thestart of training. For each subject, five ofthe reinforcer sampling sessions were con-ducted during the mornings and five wereconducted during the afternoons of 10 con-secutive school days. At the start of a rein-forcer sampling session, each of the 10 fooditems were individually presented in a ran-dom order. During this first phase of sam-pling the experimenter placed a piece offood on a 13 cm paper plate directly infront of the subject. If hesitant, the subject

was encouraged to pick up the food fromthe plate and eat it. Afterwards, the nextitem was presented in exactly the sameway. This procedure continued for all ofthe 10 items.Immediately following the above proce-

dure, the second phase of reinforcer sam-pling began. Ten items identical to thosepreviously presented and eaten weresimultaneously placed in front of the sub-ject. Each item was on an identical paperplate and arranged in an arc so that eachwas relatively equidistant from the sub-ject's dominant hand. Since some of thesubjects had gross and fine motor move-ment limitations, this arrangement was anattempt to ensure that an equivalentresponse effort would be required tochoose each item.The experimenter then directed (by

pointing) the subject to look at each itemand touch the corresponding plate (observ-ing response). Any attempts to pick up thefood were discouraged. After an observingresponse was made to all 10 stimuli, theexperimenter allowed the subject to chooseand eat any one item s/he wished. Oncethe chosen item was eaten, the procedurebegan again and continued until all thefood was eaten. The numerical order of theselections was recorded, yielding a hierar-chy for that session. This information wasused to choose training stimuli, asexplained in the subsequent section.

Selection of training stimuli. Following the10 reinforcer sampling sessions, the num-bers corresponding to each subject's selec-tion order for each reinforcer weresummed across the 10 test sessions (e.g.,peach = 1+3+2+1+6+2+4+3+2+1=25). Thesesummed scores were averaged and rank-ordered, yielding each subject'ssatiation/deprivation level for each rein-forcer. The ranks of all subjects' reinforcerswere summed according to reinforcers andaveraged across subjects, yielding a rank-order measure of satiation/deprivationlevels of reinforcers for all subjects.The experimenter selected training stim-

uli and reinforcers (these were identical forspecific reinforcement subjects) in a bal-anced fashion from the rank-order list.

SPECIFIC VERSUS NONSPECIFIC REINFORCEMENT 23

Initially, one-half of the nonspecific rein-forcement subjects were trained withcookie (ranked first) reinforced by popcorn(ranked ninth) and peach (ranked seventh)reinforced by pudding (ranked third). Theother two subjects were-trained with cereal(ranked tenth) reinforced by banana(ranked second) and pop (ranked fourth)reinforced by bean (ranked eighth). Whencriteria were met, paired training stimuliand reinforcers were switched between thenonspecific reinforcement subjects. One-half of the specific reinforcement subjectswere trained with identical stimuli andreinforcers ranked first (cookie) and sev-enth (peach) and one-half with thoseranked fourth (pop) and tenth (cereal).When criteria were met, the training stim-uli were switched.

Pretesting. After reinforcer sampling andassignment to groups, pretesting was con-ducted to control for untrained acquisitionof the target responses prior to training. Athree-trial pretest occurred prior to thestart of the first training session for eachstimulus. A pretest trial was identical to atraining trial, as described in the followingsection, but was not systematically rein-forced. Responses were recorded as corrector incorrect during the pretest.

Training sessions. The experimenterplaced a piece of food on a paper plate on acolored mat. He then prompted the subjectto look at the food by pointing at it. Afterwaiting for the subject to look at the food,he said/signed, "What is this?" Thus, thetraining stimulus had four components: (1)a nonverbal stimulus (food), (2) an observ-ing prompt, (3) a verbal stimulus ("What isthis?"), and (4) the colored mat. A correctresponse was a signed tact correspondingto the nonverbal training stimulus andoccurring within five seconds of the experi-menter saying/signing, "What is this?".The experimenter reinforced (praised andgave the subjects food) following correctunprompted signed responses (tacts). Afterthree consecutive tacts, training ended forthat stimulus in that session.A graduated prompt procedure was

instituted for incorrect or no responding.Modeling the correct response was the first

level of prompting. First, following anincorrect response or no response withinfive seconds, the experimenter re-pre-sented the training stimulus and modeledthe correct signed response. He reinforcedcorrect responses. At the second level ofprompting following an incorrect or noresponse, the experimenter re-presentedthe four-component training stimulus andimmediately took the subject's hands andhelped him/her form the correct signedtact. As before, the experimenter reinforcedcorrect responses.Two tacts were consecutively trained

daily with each subject. After the daily cri-terion was met for the first training stimu-lus, training with the second stimulus wasconducted in exactly the same manner.Daily training continued until the subjectgave three consecutive tacts or 10 trainingtrials occurred (the end-of-session crite-rion). The end-of-training criterion for eachtarget response was three consecutive ses-sions at or above 90% unprompted correctaccuracy. For practical purposes, if thetraining criterion was met for one responseprior to the other, training continued forboth until the criterion was met for bothresponses, or the end-of-training limit (25sessions) occurred. This procedure alloweda trained response to be minimally main-tained while fraining of the secondresponse continued.Reinforcement procedures differed

between conditions. In the specificreinforcement condition following acorrect tact (prompted or unprompted) theexperimenter gave the subject foodidentical (specific) to the nonverbal train-ing stimulus. For example, the experi-menter gave subjects pieces of peacheswhen real pieces of peaches were used astraining stimuli. In the nonspecificreinforcement condition, the experimentergave subjects food that was different(nonspecific) from the nonverbal trainingstimulus. For example, if subjects correctlytacted peaches, the experimenter gavethem pudding. Nonverbal stimuli usedduring the nonspecific reinforcementcondition remained consistently pairedthroughout training. For example, the

24 STEVEN J. BRAAM and MARK L. SUNDBERG

correct response "pop" always resulted indelivery of a bean. Additionally, a differentcolored paper mat was placed underneatheach nonverbal training stimulus to clearlydistinguish between training stimulichanges.

Post-training probe sessions. Probe sessionswere used to measure the resistance toextinction of the previously reinforcedresponses. In addition, probes were used toassess generalized responding in a multiplycontrolled mand condition [a combinationof EOs (mand), nonverbal stimuli (tact),and verbal stimuli (intraverbal)]. Finally, aprobe was conducted to measure mandcompliance. Sessions were conducted everyfive to seven days after the criterion for aparticular stimulus was met. The experi-menter said/signed one of the following:(a) "What is that?" (tact stimulus); (b)"What do you want?" (multiply contolledmand); and (c) "Show me the ,"_____(mand-compliance stimulus). The experi-menter simultaneously presented the train-ing stimuli (pieces of food) with each of theverbal probes. Additionally, as before, theexperimenter placed the training stimuli ondifferent colored paper mats to distinguishclearly between probe stimulus changes.Probe trials were not reinforced and werepresented in a random order.

Data collection and interobserver agreement.The experimenter collected data on correctsigned tacts corresponding to the non-verbal training stimulus and occurringwithin five seconds of the verbal stimulus,"What is that?" For example, a correct tactfor the training and probe stimuli of cornwas the sign "corn." The experimenter alsocollected data on correct mands and mandcompliances during probe sessions. A cor-rect mand following the experimenter'spresentation of corn and saying/signing,"What do you want?" was the subject sign-ing "corn." A correct mand compliance forthe experimenter's presentation of corn andsaying/signing, "Show me the corn," wasthe subject touching or pointing at the corn.A second observer independently

recorded the subjects' responses during33% of pretest, training, and probe ses-sions. Interobserver agreement was calcu-

lated according to the formula: [Number ofAgreements/(Number of Agreements +Disagreements)] X 100. Agreement was100% for pretest sessions, 96% for trainingsessions, and 100% for probe sessions.

RESULTS

Number of Learned Verbal Responses

Figure 1 shows that specific reinforce-ment subjects learned 13 out of 16 (81%)tacts to criterion. The nonspecific subjectslearned 12 out of 16 (75%) tacts to criterion.No significant difference in the databetween groups was found using an A-test.Figure 1 also shows that the pairs wereclosely matched with Pair 1 performing thelowest and Pair 4 the highest.

Tacts Acquired100- *Ua * Specific

0E Nonspecific

o 60

4040 1 1||1 z

0

1 2 3 4

Matched Pairs Means

Fig. 1. The percent of the tacts acquired to criterionfor the matched pairs. The group means are also pre-sented.

Trials to Criterion

Figure 2 shows that there was very lit-tle difference in the trials to criterionbetween the two conditions. An A-testsupported this observation showing nostatistically significant differences inthree out of four pairs. An A-test alsoshowed that there was no significant dif-ferences between the trials to criterion foracquisition for any of the four trainingstimuli. These data again show that thepairs were well-matched with Pairs 1 and2 generally requiring more training thanPairs 3 and 4.

SPECIFIC VERSUS NONSPECIFIC REINFORCEMENT 25

Trials to Criterion300

* Specific* Nonspecific

F 200

0A .100 l -1

1 2 3 4

Matched Pairs Means

Fig. 2. The trials to criterion for the matched pairs.The group means are also presented.

Tact Probes (Resistance to Extinction)

All subjects demonstrated resistance toextinction for some, but not all of thetrained responses. Figure 3 shows that spe-cific reinforcement subjects, as a group,averaged similar amounts of correct tacts(52%) compared to nonspecific reinforce-ment subjects (50%). Subjects 3 and 4 main-tained responding closest to the trained cri-teria. These data further demonstrated theclose matching of pairs, as well as the dif-ferences in performance between pairs.

Tact Probe (Resistance to Extinctic100

* SpecificD Nonspecific

80

o60

s 40

e, 20

1 2 3 4

Matched Pairs MeansFig. 3. The percent correct during resistance to

extinction conditions for the matched pairs. Thegroup means are also presented.

Multiply Controlled Mand GeneralizationProbes

Limited generalized responding wasdemonstrated by all subjects during the

multiply controlled mand probes as shownin Figure 4. However, nonspecific rein-forcement subjects averaged a greater per-centage of correct responses (38%) duringthese probe sessions than did specific rein-forcement subjects (28%). In addition, Pairs1 and 2, who consistently performed lowerthan Pairs 3 and 4 in the acquisition andmaintenance of the tacts, performed betterunder these probe conditions.

Multiply Controlled Mand100

U Specific* Nonspecific

60

X* 20

01 2 3 4

Matched Pairs MeansFig. 4. The percent of correct responses during the

multiply controlled mand condition for the matchedpairs. The group means are also presented.

Mand Compliance Probes

Figure 5 shows the emergence ofuntrained mand-compliance responses forall subjects. As a group, specific reinforce-ment subjects averaged a greater percent-age of correct mand compliances during

Mand Compliance Probes100 - Specific

Nonspecific80-

O 60

eL 20

01 2 3 4

Matched Pairs Means

Fig. 5. The percent of correct responses during themand compliance condition for the matched pairs.The group means are also presented.

26 STEVEN J. BRAAM and MARK L. SUNDBERG

probes (55%) than did nonspecific rein-forcement subjects (21%). Subjects in Pairs3 and 4 who acquired and maintained thetact responses better than Subjects in Pairs1 and 2, also acquired more mand-compli-ance relations. In addition, Pair 1 averagedonly 3% on this probe, yet performed thebest of all groups on the multiply con-trolled mand probe.

DISCUSSIONThe current study found no significant

differences between the four matched-pairs in rates of acquisition, or in the resis-tance to extinction, for tacts learned underspecific reinforcement conditions versustacts learned under nonspecific reinforce-ment conditions. However, there was amixed, but orderly difference in general-ization and in the emergence of untrainedmand-compliance relations. These resultssupport those of Stafford et al. (1988) whoalso found no difference in tact acquisition,but did find a difference in measures oflatency and choice. These results, takentogether, seem to indicate that the twotypes of consequences were equally effec-tive in the acquisition of tacting. Yet spe-cific reinforcement resulted in quickerresponding, seemed preferred by the sub-jects, and resulted in the emergence ofuntrained mand compliances. Nonspecificreinforcement, while less preferred,seemed to facilitate more generalization ofverbal responses to multiply controlledstimuli.

This study extended the methodologicalprocedures used by Stafford et al. (1988) bymeasuring the resistance to extinction ofreinforced responses, the generalization oftrained responses to different antecedentcontrolling stimuli (multiply controlledmand) and testing for mand compliance.Generalized responding during probes,although unexpected, was demonstratedby both groups. Previous work in languagetraining with DD individuals suggestedthat during early training little transfer oflearning occurs between different verbaland nonverbal repertoires. Yet, as thelearner's verbal repertoire expanded and amore extensive history of reinforcement for

responding under different stimulus con-ditions was established, transfer betweenrepertoires occurred. The current subjectsclearly did not have functional verbalrepertoires at the outset of this study, yetthey demonstrated generalized respond-ing. These data also support the positionthat severely developmentally disabledindividuals are capable of acquiring anincreasingly complex sign language reper-toire, provided the appropriate training isgiven.One of the most interesting findings of

the current study was the increased occur-rence of mand compliances followingspecific reinforcement tact training.Subjects trained with specific reinforce-ment performed significantly better thansubjects trained with nonspecific reinforce-ment during the mand-compliance probes(55% vs. 21%, respectively). These resultsindicate that tact training with specificreinforcement may encourage the emer-gence of untrained mand compliances.Although not directly trained, componentsof this new behavior were taught and prac-ticed throughout the study. The relationbetween the nonverbal stimulus (e.g.,peach), and the verbal (signed) responsewas strengthened during specific reinforce-ment training trials, as was the touchingresponse when the subject reached for,touched, and received the object in thiscondition (e.g., peach). In addition, therelation between reaching for and touchingthe nonverbal stimulus was strengthenedduring Reinforcer sampling trials. During themand-compliance probes, the verbalstimulus (now emitted by the experi-menter) combined with the new verbalstimulus ("Touch the ... .), and the novelarray of nonverbal stimuli (the food items)to evoke the correct mand-complianceresponse.The only facilitative effect observed with

the nonspecific subjects was that theydemonstrated slightly better generalizedresponding to the multiply controlledmand stimuli. In these probes subjectstrained with nonspecific reinforcementaveraged 38% correct, while subjectstrained with specific reinforcement aver-

SPECIFIC VERSUS NONSPECIFIC REINFORCEMENT 27

aged 28% correct. However, this pattern ofincreased correct responding in the non-specific condition may have been due tothe fact that the nonspecific reinforcementsubjects never received the probe stimulias consequences in training (although theyhad been exposed to them in reinforcersampling). This contrived state of depriva-tion and being asked, "What do youwant?" in the presence of the now avail-able food-a novelty effect-might havebeen responsible for some of the increasedresponding.Another type of generalized responding

also occurred for subjects in both condi-tions. As correct tacts were emitted andreinforced more frequently in a trainingsession, the same response was sometimesemitted (although incorrectly) to othertraining stimuli. For example, one subjectwho correctly tacted a peach and was rein-forced, later signed "peach" in response tocookie. Similar patterns of responding (fordifferent stimuli and responses) weredemonstrated by each subject. Althoughconsidered incorrect tacts, this type ofresponding suggested reinforcement carry-over effects and verbal stimulus control ofa thematic nature. It was possible for thesubjects to emit different tacts such as"dog" or "cup" or no response at all.However, a verbal response from the samethematic category (both peaches and cook-ies are food) was given instead. This pat-tern is similar to that reported by Braamand Poling (1983), who found that as train-ing progressed and correct intraverbalsincreased, more incorrect (thematicallyunrelated) intraverbals were given inresponse to untrained verbal stimuli dur-ing intraverbal generalization probes. Theresults of Braam and Poling (1983) and theresults of the current study suggest thatlearning might be facilitated by concur-rently training groups of thematicallyrelated responses in context and across dif-ferent stimulus conditions (e.g., train"toys" as mands, tacts, and intraverbals).This is a language training strategy thatmerits further investigation.

It still is unclear as to why specific rein-forcement did not differentially affect the

rate of acquisition or resistance to extinc-tion, but did affect latency, preference, andthe emergence of untrained relations.Especially given that specific reinforce-ment seems intuitively more effective asindicated by Carr et al.'s, (1978) recom-mended use in addition to the countlessnumber of parents, grandparents, teachers,etc. who use this technique. It may be thatthe more rapid responding, preference forspecific reinforcement, and the transfer tountrained conditions are enough to make itmore desirable.A possible explanation for the finding

that there was no difference in the tactacquisition between the two different con-ditions may have been because the estab-lishing operations (the relative depriva-tion/satiation levels for the foods) werebalanced for both conditions (see Selectionof training stimuli). That is, both specificand nonspecific subjects received more andless preferred foods as reinforcers. Thisbalancing of stimuli, intended to controlfor confounding effects, may have negatedthe effects of the establishing operations inthe specific reinforcement condition. Skin-ner (1957) noted that the consequences thatcontrol the mand are directly related to theconditions of deprivation or aversive stim-ulation, and that in the mand "theresponse...comes to 'specify' its characteris-tic consequence" (p. 83). With these condi-tions being balanced in the current experi-ment all forms of consequences wererelatively equally effective in strengthen-ing tacts.Perhaps by using the most preferred

foods in the specific reinforcement con-dition a different rate of acquisition mayhave been seen since the establishing oper-ation for these foods was strongest. Yet, itshould be noted that Stafford et al. (1988)did not use all consumables, but did usethe subject's strongest forms of reinforce-ment (e.g., ball, tickle, food), and it wasfound that these reinforcers were alsoequally effective in strengthening tacts.However, in both studies it was not possi-ble to control for the subject's access toreinforcers outside of the school setting, soit was possible that their satiation/depri-

28 STEVEN J. BRAAM and MARK L. SUNDBERG

vation levels varied over time. These vari-ables need to be examined in futureresearch where different stimuli (not onlyfood), and the momentary value of theestablishing operation can be contrived.For example, conducting tact training forthe response "pop" with specific reinforce-ment following the eating of a salty pretzelmay increase the momentary value of popas a reinforcer and may speed up training.Similar results might also be expectedwhen training "umbrella" as a tact whilegetting wet standing in the rain or in ashower.

In conclusion, the results supportSkinner's (1957) distinction between themand and the tact by demonstrating thatdifferent repertoire-altering effects are pro-duced by the independent variables thatdefine those two verbal operants. Theresults also suggest that both nonspecificand specific reinforcement should bearranged whenever possible to facilitategeneralized responding and the transfer oflearning to mand compliances, respec-tively. In addition, isolated tact trainingmay have limited effectiveness in theestablishment of verbal and nonverbalrepertoires. The concurrent training of thesame sign or word under conditions withdifferent controlling variables may have amore dramatic impact on the establishmentof verbal behavior.

REFERENCESBraam, S. J., & Poling, A. (1983). Development of

intraverbal behavior in mentally retarded individu-als through transfer of stimulus control procedures:Classification of verbal responses. Applied Researchin Mental Retardation, 4, 279-302.

Carr, E. G., Binkoff, J. A., Kologinsky, E., & Eddy, M.(1978). Acquisition of sign language by autistic chil-dren. Journal of Applied Behavior Analysis, 11, 489-502.

Carroll, R. J., & Hesse, B. E. (1987). The effects of alter-nating mand and tact training on the acquisition oftacts. The Analysis of Verbal Behavior, 5, 55-65.

Doll, E. (1965). Vineland Social Maturity Scale. CirclePines, MN: American Guidance Service.

Faw, G., Reid, D., Schepis, M., Fitzgerald, J., & Welty,P. (1981). Involving institutional staff in the devel-opment and maintenance of sign language skillswith profoundly retarded persons. Journal of AppliedBehavior Analysis, 14, 411-424.

Hall, G., & Sundberg, M. L. (1987). Teaching mandsby manipulating conditioned establishing opera-tions. The Analysis of Verbal Behavior, 5, 41-53.

Lamarre, J., & Holland, J. G. (1985). The functionalindependence of mands and tacts. Journal of theExperimental Analysis of Behavior, 43, 5-19.

Michael, J.L. (1982). Distinguishing between discrimi-native and motivational functions of stimuli. Journalof the Experimental Analysis of Behavior, 37, 149-155.

Michael, J.L. (1983). Evocative and repertoire-alteringeffects of an environmental event. The Analysis ofVerbal Behavior, 2, 21-23.

Michael, J.L. (1988). Establishing operations and themand. The Analysis of Verbal Behavior, 6, 3-9.

Saunders, R. R., & Sailor, W. (1979). A comparison ofthree strategies of reinforcement on two-choiceleaming problems with severely retarded children.AAESPH Review, 4, 323-333.

Sigafoos, J., Reichle, J., Doss, S., Hall, K., & Pettitt, L.(1990). "Spontaneous" transfer of stimulus controlfrom tact to mand contingencies. Research inDevelopmental Disabilities, 11, 165-176.

Skinner, B. F. (1957). Verbal behavior. EnglewoodCliffs, NJ: Prentice-Hall.

Stafford, M. W., Sundberg, M. L., & Braam, S. J.(1988). A preliminary investigation of the conse-quences that define the mand and the tact. TheAnalysis of Verbal Behavior, 6, 61-71.

Sundberg, M. L., San Juan, B., Dawdy, M., &Arguelles, M. (1990). The acquisition of tacts,mands, and intraverbals by individuals with trau-matic brain injury. The Analysis of Verbal Behavior, 8,83-99.

Terman, L., & Merrill, M. (1973). Stanford-BinetIntelligence Scale. Boston: Houghton-Mifflin.