11
The Analysis of Verbal Behavior 1988, 6, 61-71 A Preliminary Investigation of the Consequences That Define the Mand and the Tact Mark W. Stafford, Mark L. Sundberg, and Steven J. Braam Goodwill Industries, Muskegon, MI, Sundberg & Associates, Concord, CA, & Western Michigan University Skinner (1957) proposed that the mand and the tact differed with respect to their unique antecedents and consequences. The present study examined the specific reinforcement charac- teristic of the mand, and the nonspecific reinforcement characteristic of the tact. A severely mentally impaired individual who used sign language served as subject. A concurrent-chain with latency measures and choice was used. The results showed that specific reinforcement produced stronger verbal behavior than nonspecific reinforcement, but only when response strength was measured in terms of latency and choice. These data lend support to Skinner's assertion that the mand and the tact are different operants. These results also have practical significance in that they may lead to more effective work with individuals who have speech and language impairments. The distinction between the mand and the tact is an important aspect of Skinner's (1957) analysis of verbal behavior. These two verbal operants are defined by variables not present in the other types of verbal behavior, and as a result, have often been a source of difficulty for those trying to understand and use Skin- ner's classification system. In the mand relation the form of the response is con- trolled by establishing operations (Michael, 1982; 1988) and specific consequences. In the tact relation the form of the response is con- trolled by nonverbal discriminative stimuli and generalized conditioned reinforcement. Research on these different types of control has been virtually nonexistent since Verbal Behavior was published thirty-one years ago. However, a thematic line of research on the mand and the tact has recently begun to appear in the behavioral literature. Portions of this paper were presented at the 4th Annual Meeting of the Midwestern Association for Behavior Analysis, Chicago, IL, May, 1978. The authors would like to thank Gerald L. Shook, former coordinator of the Kalamazoo Valley Multihandi- cap Center (now Croyden Ave. School), for his admini- strative support for this research. Appreciation is also extended to Cassandra Braam, David Ray, Mary Rior- dan, and Thomas Rueber for their assistance in conduct- ing this study, and to Joe Lancaster, Jane Howard, and Jack Michael for their helpful comments on earlier ver- sions of the manuscript. Reprint requests should be sent to Mark L. Sundberg, P.O. Box 21492, Concord, CA 94521. In one of these studies, Lamarre and Holland (1985) showed that subjects who learned to first mand using the phrases, On the right and On the left, did not demonstrate collateral acquisition of the tacts. And those who learned to first tact the locations were not automatically able to mand for them. These authors conclude that the mand and the tact are functionally independent at the time of acquisition. In a related study, Hall and Sundberg (1987) examined the distinction between establishing operations and nonverbal discriminative stimuli. Their results, like those of Lamarre and Holland, showed that verbal behavior acquired under the control of nonverbal discriminative stimuli did not automatically transfer to control by estab- lishing operations. Subjects showed no tend- ency to mand for objects, even though they had a successful tact history with respect to them. Special training was required to estab- lish the mand repertoire. These data, along with those of Lamarre and Holland, have begun to clarify the distinction between the mand and the tact, while providing support for Skinner's (1957) assertion that they are separate operants. However, this line of research is clearly in its preliminary stages. The current study was a further investiga- tion of the mand and the tact, but with the focus on consequences. Skinner (1957) states that in the mand "The response... .comes to 'specify' its characteristic consequence" 61

Preliminary Investigationofthe Consequences ThatDefinethe

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Preliminary Investigationofthe Consequences ThatDefinethe

The Analysis of Verbal Behavior 1988, 6, 61-71

A Preliminary Investigation of the ConsequencesThat Define the Mand and the Tact

Mark W. Stafford, Mark L. Sundberg, and Steven J. BraamGoodwill Industries, Muskegon, MI, Sundberg & Associates,

Concord, CA, & Western Michigan UniversitySkinner (1957) proposed that the mand and the tact differed with respect to their uniqueantecedents and consequences. The present study examined the specific reinforcement charac-teristic of the mand, and the nonspecific reinforcement characteristic of the tact. A severely mentallyimpaired individual who used sign language served as subject. A concurrent-chain with latencymeasures and choice was used. The results showed that specific reinforcement produced strongerverbal behavior than nonspecific reinforcement, but only when response strength was measuredin terms of latency and choice. These data lend support to Skinner's assertion that the mand andthe tact are different operants. These results also have practical significance in that they may leadto more effective work with individuals who have speech and language impairments.

The distinction between the mand and thetact is an important aspect of Skinner's (1957)analysis of verbal behavior. These two verbaloperants are defined by variables not presentin the other types of verbal behavior, and asa result, have often been a source of difficultyfor those trying to understand and use Skin-ner's classification system. In the mandrelation the form of the response is con-trolled by establishing operations (Michael,1982; 1988) and specific consequences. In thetact relation the form of the response is con-trolled by nonverbal discriminative stimuliand generalized conditioned reinforcement.Research on these different types of controlhas been virtually nonexistent since VerbalBehavior was published thirty-one years ago.However, a thematic line of research on themand and the tact has recently begun toappear in the behavioral literature.

Portions of this paper were presented at the 4thAnnual Meeting of the Midwestern Association forBehavior Analysis, Chicago, IL, May, 1978.The authors would like to thank Gerald L. Shook,

former coordinator of the Kalamazoo Valley Multihandi-cap Center (now Croyden Ave. School), for his admini-strative support for this research. Appreciation is alsoextended to Cassandra Braam, David Ray, Mary Rior-dan, and Thomas Rueber for their assistance in conduct-ing this study, and to Joe Lancaster, Jane Howard, andJack Michael for their helpful comments on earlier ver-sions of the manuscript. Reprint requests should be sentto Mark L. Sundberg, P.O. Box 21492, Concord, CA94521.

In one of these studies, Lamarre andHolland (1985) showed that subjects wholearned to first mand using the phrases, Onthe right and On the left, did not demonstratecollateral acquisition of the tacts. And thosewho learned to first tact the locations werenot automatically able to mand for them.These authors conclude that the mand andthe tact are functionally independent at thetime of acquisition.In a related study, Hall and Sundberg

(1987) examined the distinction betweenestablishing operations and nonverbaldiscriminative stimuli. Their results, likethose of Lamarre and Holland, showed thatverbal behavior acquired under the controlof nonverbal discriminative stimuli did notautomatically transfer to control by estab-lishing operations. Subjects showed no tend-ency to mand for objects, even though theyhad a successful tact history with respect tothem. Special training was required to estab-lish the mand repertoire. These data, alongwith those of Lamarre and Holland, havebegun to clarify the distinction between themand and the tact, while providing supportfor Skinner's (1957) assertion that they areseparate operants. However, this line ofresearch is clearly in its preliminary stages.The current study was a further investiga-

tion of the mand and the tact, but with thefocus on consequences. Skinner (1957) statesthat in the mand "The response... .comes to'specify' its characteristic consequence"

61

Page 2: Preliminary Investigationofthe Consequences ThatDefinethe

62 MARK W. STAFFORD et al.

(p. 83). The verbal response "I want food,"for example, specifies to the listener that foodwould be an effective form of reinforcement.The receipt of food can be identified as"specific reinforcement." The tact is notcharacteristically followed by any single typeof reinforcement, rather by various forms ofgeneralized conditioned reinforcement. Thetendency to say "It's raining." for example,when one sees rain (discriminative stimulus)is not reinforced by the receipt of rain, but bysome form of generalized conditioned rein-forcement such as "Thank you." or "That'sright, it is raining."There were no studies in the literature

which directly examined these differenttypes of consequences in relation to verbalbehavior, but Saunders and Sailor (1979)have provided some direction with theirresearch on receptive discriminations. Theyused a form of generalized reinforcementidentified as "nonspecific reinforcement"and contrasted it with specific reinforce-ment. In their study toys which appeared tofunction as reinforcement were assignednonsense names and presented in pairs toinstitutionalized retarded subjects. Theywere asked to point to one of the toys givenits nonsense name. Correct responses wereconsequated in the following manner:

1. In the specific reinforcement conditioncorrect responses were followed by anopportunity to play with the toy to which thechild pointed.2. In the nonspecific reinforcement conditioncorrect responses were followed by theopportunity to play with a toy offered by theexperimenter, which was not part of thetraining pair.3. In the variable reinforcement condition thechild was offered, on correct pointing occa-sions, either the toy to which he or shepointed, or the toy which was not one of thepair being trained, in a randomized order.

The results of this study demonstrated thatthe percent of correct receptive discrimina-tions was higher under the specific conditionthan it was under either of the other twoconditions. Based on these results, it seemsplausible to speculate that the consequencesof the mand may produce stronger behaviorthan those of the tact. However, there are nodata to support this extension. The purpose

of the current research was to examine theeffects of specific and nonspecific reinforce-ment on expressive verbal behavior, whenestablishing operations, nonverbal stimuli,and response requirements were equated.There were several features of this study

which differed from that of Saunders andSailor (1979). First, the research wasexploratory, and several procedural changesoccurred during the experiment which werean attempt to "follow" the behavioralchanges of the subject (cf. Skinner, 1956;Day, 1975). Also, the study differed in thefollowing ways: (a) Sign language was usedrather than pointing at objects, (b) a complexchain of verbal responses was requiredrather than a single response, (c) a multiele-ment design was used rather than a reversaldesign, and (d) latency and choice were ex-plored as dependent variables.

METHOD

SubjectAn eleven year old speech and language

impaired male with severe retardation servedas subject. He attended the KalamazooValley Multihandicap Center and was com-pletely nonverbal prior to beginning a signlanguage training program at the Center. Hehad a long history of self-stimulation, anadaily aggressive and destructive behaviors(e.g., hitting himself and others, scratching,pinching, kicking, spitting, screaming). Thecurrent study began after approximately twoyears in the sign program. The subject hadacquired over 100 single-component signedmands and tacts, and approximately 10 intra-verbal responses. Also, his rate of nega-tive behaviors decreased by about 80%.Although his vocal behavior had improvedduring the two years, the subject used signlanguage as his main form of verbal behavior,and all his responses in the study weresigned.

Materials and SettingThe materials in the experiment consisted

of a blue plastic 8 ounce cup, a green plasticsalad bowl, a small table approximately 30cm high, a small chair approximately 30 cmat the seat and 56 cm at the back, and twocardboard boxes. The boxes were the samesize (58 cm high), but differed in both color

Page 3: Preliminary Investigationofthe Consequences ThatDefinethe

THE CONSEQUENCES DEFINING THE MAND AND THE TACT 63

and pattern so that clearly distinguishablestimuli were associated with each condition.One box was covered with red-and-yellowone-inch paper stripes, while the other wascovered with white paper, and painted withpurple dots approximately 5 cm in diameter.Both boxes had a base measurement of 70 cmx 78 cm. They were cut diagonally along thesides leaving the back of the box at 58 cm,and reducing the front of the box to 12 cm.A stopwatch and data sheet on a clipboard

were used to record the data. A randomnumbers table was consulted prior to theexperiment to order the trials and presenta-tion of reinforcers. This information wasprinted on the data sheets, and was used todetermine which condition to present, andwhat reinforcer to place in the cup or thebowl during the experiment.The stimulus materials were separated into

two sets and placed in the boxes. Onestimulus array consisted of the blue cup on thetable in the striped box, and the other arrayconsisted of the green bowl under the chair inthe dotted box. All the stimuli were clearlyvisable, and the subject could tact all theitems in the chain-but not as a chain. Thetwo sets of stimuli were randomly assigned(using a random numbers table) to two dif-ferent conditions. The striped box and itscontents were assigned to the mand condi-tion, which involved specific reinforcement,and the dotted box and its contents wereassigned to the tact condition which involv-ed nonspecific reinforcement (the mand con-dition was more correctly both a mand andtact condition since there were alwaysnonverbal stimuli present along withestablishing operations and specificreinforcement).Five other objects or actions which had

consistently functioned as reinforcement inthe past were selected for use as conse-quences. These were soda, water, food(usually a potato chip), a ball, and a tickle (aphotograph of the experimenter tickling thesubject was used as the stimulus and actualtickling was used as the consequences). Thesubject could correctly tact, and mand for, allthese items.The sessions were conducted in a 3.3 m x

6.6 m x 2.4m room at the Center. In Phase 1a cardboard divider, approximately 1.5 mhigh, was used to reduce the room size to 3.3m x 3.3 m. The divider was removed during

three sessions in Phase 2. The two boxeswere placed along opposite walls, and the ex-perimenter and the subject stood facing eachother in the center of the room, directly bet-ween the two boxes. The mand box was onthe experimenter's left and the tact box wason his right. The reinforcement items wereplaced on the floor behind the experimenter,and the reliability observer sat in a chair inone corner of the room.

Design and Procedure

A within-subject multielement design wasused. Trials consisted of presenting the sub-ject with one of the two stimulus arrays, andconsequating correct responses with eitherspecific (mand) or nonspecific (tact) rein-forcement. Later aspects of the study usedlatency, and a concurrent-chain with forced-trials and choice-trials (Nevin & Mandell,1978), as dependent variables.

Baseline. The contingencies were exactlythe same for the mand and tact duringbaseline conditions. There were 3 sessions of10 randomized trials on each component ofthe concurrent-chain. The experimenter heldup one of the five reinforcement items,placed it in either the cup or the bowl (alldetermined by the order pre-printed on thedata sheets), and said and simultaneouslysigned, "Look here." The subject was askedto attend to the appropriate box by steppingin front of it, and orienting his head towardthe set of stimuli in that box. The experi-menter then said/signed, "Where?" A cor-rect response was scored if the subject signedat least two-components of the stimulusarray. In both conditions correct responseswere followed by verbal praise (e.g., Right!),but not by the receipt of any of the itemsplaced in the cup or the bowl.

Phase 1: Shaping the five-component chains.Each session consisted of an equal numberof trials in both the mand and tact condi-tions. During the two- and three-componenttraining there were a total of 20 trials per ses-sion, and during four- and five-componenttraining there were 10 trials per session.The training began with reinforcing a two-

component chain consisting of tacting (1) thereinforcement item, and (2) the cup or the

Page 4: Preliminary Investigationofthe Consequences ThatDefinethe

64 MARK W. STAFFORD et al.

bowl, in any order, within 10 seconds. Theprocedure was the same as in baseline,except the correct response was modeledonce for each condition, and a correctresponse in the mand condition was conse-quated with the reinforcement item placedin the cup. A correct response in the tactcondition was consequated with one of theother forms of reinforcement (presented inthe random order pre-printed on the datasheets), none of which were in the bowl.Incorrect responses, which were usually

only single-component responses, were fol-lowed by a correction procedure consistingof the experimenter stating the correctresponse, prompting the subject to imitatethe response, consequating correct imita-tions with verbal praise, and presenting theoriginal trial again. Correct responses on thistrial were followed with only verbal praise.When correct responses occurred on 90%

or more of the trials for two consecutive ses-sions, another component was added to theresponse chain. Each time a new stimuluswas added to the chain the correct responsewas modeled once for each condition. Thethird stimulus was table for the mand condi-tion and chair for the tact condition. Follow-ing two consecutive sessions of meeting theresponse requirement a fourth componentwas added to the chain. The prepositions onand under were used for the mand and thetact conditions, respectively (a correct mandresponse was now food-cup-on-table). Finally,a fifth stimulus was added following suc-cessful completion of the four-componentresponses. The adjectives blue and green wereused for the mand and tact conditions,respectively (a correct mand response wasnow food-blue-cup-on-table).Due to the exploratory nature of this

research, several procedural changes weremade during the study in an attempt tofollow the behavioral changes of the subject.Starting with the four-component trials, thefirst change occurred with the addition oflatency as a dependent variable. Timingbegan with the stimulus, "Where?" andended with the completion of the lastresponse in the chain. This measure wasadded to further quantify the differencesbetween the two sets of contingencies. Also,the number of trials per session was reducedfrom 20 to 10 per session to avoid satiation.Phase 2: Choice trials. The second phase of

the study (which had not been planned atthe outset of the study) was conducted tofurther measure the preference the subjectseemed to be showing for one contingencyover the other. Following the completion ofthe five-component trials, a choice conditionwas added to the concurrent schedule. Afoam block was placed on the floor betweenthe experimenter and the subject. Theexperimenter held up the reinforcement itemwhich was next on the list and said andsigned to the subject, "Throw the block inthe box," (four trials later this stimulus wasreduced to simply pointing to the block onthe floor). Following the block-toss to a box,the experimenter placed the reinforcer in thebox selected by the subject, and the pro-cedure from Phase 1 began.There were four different types of trials in

Phase 2. First, four sessions of five choice-trials were conducted. Second, stimuli fromthe boxes were exchanged and, choice-trialswere conducted. Third, 10 forced-choicetrials were added to the conditions justdescribed. And finally, the subject wasallowed to select a reinforcer to use on eachtrial prior to the choice-trials. Each of theseconditions will be described further below.In the first four sessions of choice-trials

there were four different types of choice-trials. In the first choice session, the boxeswere in the same position as they were dur-ing training. In the second session, the posi-tions were reversed. In the third session,they were placed in various positions in theroom without the divider. And in the forthsession, they were placed in various posi-tions in the larger room, including turningthe boxes backwards, and placing the mandbox behind the divider.During the second condition of Phase 2, in

an attempt to determine what variables werecontrolling the subject's behavior, five choice-trials were conducted with two of the stimulifrom the boxes exchanged. The cup and bowlwere exchanged, because they seemed to bethe most salient feature of the array (i.e., thecup was placed in the tact box and the bowlin the mand box). The two boxes were placedin various positions in the larger room andfive open-choice trials were conducted.The results from this condition controlled

the third condition which consisted of con-ducting 10 forced-trials (5 on each condition)

Page 5: Preliminary Investigationofthe Consequences ThatDefinethe

THE CONSEQUENCES DEFINING THE MAND AND THE TACT 65

before the choice-trials (Nevin & Mandell,1978). Prior to the first session of scheduledforced-trials, two warm-up trials (one in eachcondition), in which the correct responsewas modeled, were conducted. The samefoam block was used, and the boxes wereplaced in various positions in the room withthe divider back in place. The forced-trialswere conducted in the same manner as thechoice-trials, except the subject was toldwhere to throw the block. After the blockwas thrown, the procedure was the same asin Phase 1. After 10 interspersed forced-trialson the specific as well as nonspecific conse-quences, 5 choice-trials occurred.The results from these conditions con-

trolled the final condition, which con-sisted of an attempt to control for themomentary establishing operations duringeach trial. This condition consisted of choice-trials only, and the subject selected the formof reinforcement to be used on each trial. Theexperimenter pointed to the reinforcers andsaid, "Pick one." The subject then picked upone of the items and handed it to the experi-menter. The experimenter held up the rein-forcement as before, placed the block on thefloor in front of the subject, and pointed atthe block. After the subject threw the blockinto a box, the procedure was the same as inPhase 1.

ReliabilityReliability data were collected on correct

and incorrect responses, latency ofresponses,and choice. Reliability measures were takenon approximately 75% of the trials in theexperiment. Agreements, divided by agree-ments plus disagreements, times 100 wasused to compute reliability on percent correctand choice measures. In baseline the meanreliability was 86%, in Phase 1 the meanreliability was 99%, and in Phase 2 thereliability was 100%. Reliability on latencymeasures was calculated by weighing means.Table 1 shows that the difference betweenobservers varied from 0.0 to 0.7 seconds. Thenumber of differences at each tenth of a sec-ond were counted and multiplied by the cor-responding number of tenths of a seconddifference between observers. The resultingnumbers were then summed and divided bythe total number of reliability observations.This resulted in a mean difference of 0.19seconds (237.121=0.19 s).

Table 1.Reliability on latency measures.

Differences in tenths O .1 .2 .3 .4 .5 .6 .7 _ofsecond

# different at each 23 32 31 13 12 7 2 1 121tenthI#tenthsx#II 3262I39 48 35 12I7Idiffe ren cesiiiiO3__ _7

RESULTS

Baseline. Figure 1 shows that during thethree baseline sessions the subject failed totact more than one stimulus item from thearray on 92% of the trials. The subject typ-ically tacted the cup or the bowl, or the rein-forcement items correctly, but rarely theprepositions or adjectives, and never thetable or chair. The higher performance on themand condition was probably due to socialpraise, since specific and nonspecific rein-forcement were not used.

Phase 1. The subject acquired both twocomponent responses in three sessions (seeFigure 1). There was almost no differencebetween the two conditions. During the firstsession of intervention the subject was cor-rect in the mand condition on the last trial ofthe day. For the first time he came in contactwith the contingency of receiving a reinforcerout of either the cup or bowl. The reinforce-ment item on that trial was food, and ithappened to be placed in the cup, thusspecific reinforcement was used. However,the rapid acquisition of both response chainsin the next two sessions indicated that receiv-ing any of the reinforcers from the cup or thebowl was a stronger form of reinforcementthan verbal praise alone.The three-component responses were

acquired in two sessions, and again correctresponses and trials to acquisition wereequal (see Figure 1). However, the subjectseemed to behave differently during the twoconditions. He responded more slowly inthe tact condition, and occasionally emittednegative behaviors (e.g., screams, andattempts to pinch the experimenter) when hedid not receive the item in the bowl. Thesebehaviors led the experimenters to includelatency to completion of the response as adependent variable. Also, it appeared thatthe subject began to satiate toward the endof the sessions, so at this point the numberof trials was reduced to 10 per session.

Page 6: Preliminary Investigationofthe Consequences ThatDefinethe

66 MARK W. STAFFORD et al.

Baseline Two- Three- Four- Five-Component Component Component Component

Chain Chain Chain Chain

100

90I

0f 3

20 o

z~~

70

260 I-I0

10 \

,TACT

I ~ ~~~~~~~~~~~ .

2 3 4 5 6 7 8 9 10 11 12 13 14 1 16 17

SESSIONSFig. 1. The percent of correct responses for mand and tact chains during baseline and two- through five-component training.

Figure 1 also shows that during the four-component training, when the prepositionswere added, it took seven sessions to meetcriterion on both operants (but sessions werenow only 10 trials long). The criterion wasmet in the mand condition in only two ses-sions, thus showing the first differencebetween the two conditions in terms of per-cent correct and trials to acquisition. How-ever, mand performance deteriorated some-what before performance on the tact metcriterion, and as a result, the average percentcorrect was equal for both mand and tactconditions.There was, however, a consistent differ-

ence in the response latency. Figure 2 showsthe mean latency in seconds during the four-component and five-component training,

and a later condition. In the four-componenttraining, the mean latency for the mand con-dition was 2.5 seconds (range 1.5-3.2), andfor the tact condition it was 3.0 seconds(range 2.1-3.6). Thus, it now appeared thatthere was some measurable differencebetween specific and nonspecific reinforce-ment.In the five-component training, both

response chains were acquired in two ses-sions (see Figure 1). The percent of correctresponses, and the trials to acquisition wereequal. However, there was a relatively largerlatency differential than that observed dur-ing the four-component training. The meanlatency for the mand condition was 3.5 sec-onds (range 2.7-4.2), and for the tact condi-tion it was 4.7 seconds (range 3.6-5.7).

Page 7: Preliminary Investigationofthe Consequences ThatDefinethe

THE CONSEQUENCES DEFINING THE MAND AND THE TACT 67

Four-Component Five-Component Forced-TrialsChain Chain Stimuli Exchanged

6.0

5.5

5.0I

4.5I

C4.0zo -*--TACT

Z03.0s.0 'A

3 2' 6:

2.0 / I

\/~~ ~~~~~~~~~I I

1.51 MAND

1.0

0.5

9 10 11 12 13 14 15 16 17 23 24 25 26 27 28

Fig. 2. Latency in seconds for mand and tact chains during four- and five-component training, and during the forced-trial procedures with the stimuli exchanged.

Figure 2 also contains latency data for theforced-trials, which will be describedshortly.)

Phase 2. The subject's quicker respondingin the mand conditions controlled the exper-imenters' decision to continue studying thetwo verbal operants with a series of choice-trials. These trials were conducted in orderto determine if there was a preference forspecific reinforcement. Figure 3 shows thatduring the first four choice sessions themand box was chosen 100% of the trials(panel 1), and the five-component responsewas always correct (panel 2). The boxes were

placed in several different locations, andeven made difficult to get to, but the subjectcontinued to always chose the mand box.When the mand box was placed out of sight,behind the divider, he threw the foam blockover the divider. Since all the choices werefor the mand condition, there were no tacttrials.To determine if the strength of the behavior

could be reduced, and the control producedby specific reinforcement transferred to a dif-ferent stimulus array, the experimentersexchanged the cup and the bowl. These par-ticular stimuli were exchanged because oftheir involvement in all phases of the study,

Page 8: Preliminary Investigationofthe Consequences ThatDefinethe

68 MARK W. STAFFORD et al.

and the reinforcers were placed in them. Insession 21 (Figure 3), when the cup and thebowl were exchanged, the subject chose themand box four out of five times (80%). How-ever, he responded incorrectly on the firsttwo mand choices. The correct response inthe mand box was now (SR/Sr) green-bowl-on-table, but the subject rapidly signed "(SR/Sr)blue-cup-on-table." The errors resulted inwithholding reinforcement and the subjectchose the tactboxon the third trial. However,he againmade anerroron thefive-componentresponse, did not receive reinforcement, andchanged-over to the mandbox. He then emit-ted a correct response on the next two mandtrials and received reinforcement.The subject's incorrect tacting during this

session controlled the experimenters' deci-sion to add forced-trials to establish the newfive-component response chains (i.e., green-bowl-on-table rather than blue-cup-on-table),and facilitate the transfer of stimulus control.The forced-trials also allowed for latency datato again be collected and compared. Figure3 shows that following the first set of forced-trials, choice for the mand box dropped to0%. It appeared that the subject still pre-ferred that the reinforcer be placed in thecup. But over the next five sessions (numbers24-28), choice for the bowl (which waspreviously associated with the tact condi-tion), and the mand condition increased to60%, with 100% correct responses for thosemand choices (the mean percent of correctresponses for the tact choices was 90%).Thus, stimulus control transferred, in part,to the bowl and its accompanying array,because of its new correlation with specificreinforcement.In addition, when the stimuli were

exchanged, the latency differences prev-iously seen between the two conditionsdiminished during the forced-choice condi-tion (Figure 2, sessions 23-28). The speed ofresponding became about equal for the twoconditions. The mean latency during themand condition was 2.7 seconds (range1.7-4.2), and during the tact condition it was2.6 seconds (range 1.8-4.3). Thus, the dif-ferential control acquired by the two types ofconsequences was lost when the stimuli thathad been correlated with those conse-quences were reversed.The final procedural change occurred

when it was noted in session 28 that the sub-

ject consistently chose the mand conditionwhen a consumable was the item shownprior to the trials (water, food, or soda), andthe tact condition when the item was a non-consumable (tickle or ball). Apparently, thesubject became sensitive to the contingencythat there was a .75 probability of obtaininga consumable under these conditions. Inorder to control for the momentary estab-lishing operations, the subject was allowedto select the reinforcement item to be usedon each trial. As a result of this change, thesubject showed a stronger preference for themandcondition. Choice for the mand condi-tion increased to a mean of 85% with 85%correct responses. During sessions 30 and 32on three occasions errors on the five-component responses occurred following amand choice, which resulted in withholdingreinforcement. The subject always thenchose the tact box on the subsequent trial.When this choice did not produce theselected reinforcement, the next choice wasalways back to the mand condition.

DISCUSSION

The results of the study indicate thatspecific reinforcement produces strongerverbal behavior than nonspecific reinforce-ment, but only when response strength wasmeasured in terms of latency and choice.Specific reinforcement did not always pro-duce a higher percent of correct responses,or result in fewer trials to acquisition. Withthe exception of the results of the four-com-ponent training, both conditions weretypically equal. It was clear however, that thesubject preferred specific reinforcement andresponded quicker when it was available, buthe did not consistently learn the multipleresponses any faster under those conditions.These results may seem contrary to those

of Saunders and Sailor (1979), but there aresome possible explanations for the differ-ences. During baseline the only consequenceused was praise. The transition from praiseto obtaining any one of five stronger formsof reinforcement may have initiallydecreased any significant differencesbetween the two types of consequences.Regardless of which form of reinforcement(specific or nonspecific) the subject receivedin intervention, it was more effective thanpraise alone. Also, the individual tact

Page 9: Preliminary Investigationofthe Consequences ThatDefinethe

THE CONSEQUENCES DEFINING THE MAND AND THE TACT

Ohoi*Tra Nomi F. rT suWA -x weio comm qpldRUinSlkd

l

.,

l*@ @ @ ~~~~~~~~~~~18 19 20 21 22 23 24 25 26 27 26 29 30 31 32

0 0 0

Panel 2

16 19 20 21 22 23 24 25 27 28 29 30 31 32

SESSIONS

Fig. 3. Percent of mand choices in panel 1 and percent of correct responses for those mand chains in panel 2 duringchoice-trials, and choice-trials a) with the stimuli exchanged b) preceded by forced-trials with the stimuli exchanged,and c) the reinforcement selected by the subject each trial with the stimuli exchanged.

69

IOOT

901

LU- 700zC.) 60.a

CL20

I-40,ZLu() 30,

I. 201

101

100

0 9

Z N

M 700IL 60

USwa:

30

z20UJ

10

------------

I

I

Page 10: Preliminary Investigationofthe Consequences ThatDefinethe

70 MARK W. STAFFORD et al.

responses were already strong in the sub-ject's repertoire. Combining knownresponses may have been a simpler task thanlearning new nonsense names for objects. Infact, recent research using response topo-graphies not in the subject's repertoire hasshown that specific reinforcement does pro-duce more correct responses, and a fasterrate of acquisition (Carroll & Hesse, 1987).The results of the four-component trainingin the current study, which involved preposi-tions and the most errors, may also indicatethat differences could be seen when moredifficult tasks are involved.Based on the results of Saunders and

Sailor (1979) the experimenters expected dif-ferentiation in the acquisition of the tworesponses. Even though there was no dif-ference, casual observation indicated thatthere was something different about the waythe subject responded during the two con-ditions. He seemed "less interested" in thetact condition and "more interested" in themand condition. Therefore, latency andchoice were added as dependent variables inan attempt to quantify what appeared to bean obvious difference in the two conditions.With these additional measures, the

results of the current study support thegeneral findings of Saunders and Sailor(1979). The results also extend this line ofresearch into the area of verbal behavior, andspecifically to the distinction between themand and tact. This research is importantbecause there are several practical applica-tions of the mand-tact distinction in languageinstruction for the developmentally disabled.In procedures for teaching language tononverbal individuals, for example, specificreinforcement has been an effective tool forgenerating successful verbal behavior whenother methods have failed (Sundberg, 1987).The specific consequences characteristic of

the mand may improve language instructionbecause they directly benefit the speaker.This type of consequence may produce more"interest" or motivation for learning verbalskills because of the clear relation betweenwhat functions as reinforcement and obtain-ing that reinforcement. These can be usefulindependent variables and should be incor-porated into daily language instruction pro-cedures. Establishing operations and specificconsequences can also be used to teach othertypes of verbal operants. Carroll and Hesse

(1987) demonstrated that these contingenciescould improve tact training. And Sundberg(1987) has proposed that they could be usedto teach echoic, intraverbal and textualbehavior as well.Another use of specific reinforcement in

language training is that it may help developcontrol by the establishing operation for thatreinforcement. According to Skinner's analy-sis, if you consistently reinforce a responsewith a specific consequence that responsewill more likely come under the control ofthe relevant establishing operation (Skinner,1957, p. 35). If an establishing operation fora specific consequence is strong and appro-priate verbal behavior is not available, inap-propriate behavior such as aggression,pushing, or hitting may occur. If thesebehaviors are followed by specific reinforce-ment, then the next time the establishingoperation is strong they will likely occuragain. These behaviors may erroneously beviewed simply as negative behavior thatmust be punished, rather than verbalbehaviors which occur because of a defectivemand repertoire.Skinner's analysis of verbal behavior has

much to offer the fields of developmentaldisabilities and speech pathology. Profes-sionals in these areas typically emphasizeonly receptive and tact training in work withlanguage delayed individuals, while neglect-ing the mand (as well as some of the otherverbal operants). The blame for verbalfailures is often placed on the student'sgeneral lack of "cognitive" ability, ratherthan the contingencies and the nature of thetraining program. It may be interesting tonote that the subject in this study had beenlabeled "severely mentally impaired," andrarely emitted more than single-componentresponses. However, he acquired two five-component responses in only 17 sessions.Thus, the current study not only demon-strated some potential differences betweentwo types of consequences, but it was a suc-cessful demonstration of procedures forteaching sentence construction. Also, theseresults may support the analysis that thecauses of the subject's intellectual and socialbehaviors were due to verbal and environ-mental deficits, rather than cognitive deficits.This research was preliminary, and

attempted to provide some of the basicgroundwork for further research in this area.

Page 11: Preliminary Investigationofthe Consequences ThatDefinethe

THE CONSEQUENCES DEFINING THE MAND AND THE TACT 71

The effects of specific and nonspecific rein-forcement on verbal behavior have receivedvery little experimental attention and needto be further explored. There are severalpossible extensions of this study. First, sincethere was only one subject, the procedureneeds to be replicated with additional sub-jects. The effects of a praise and/or an extinc-tion condition interspersed with the othertwo conditions could be interesting, as couldfurther experimentation with latency andchoice as dependent variables. It may also beproductive to conduct the procedure usingspeech, and typical children. The use ofspecific consequences in the acquisition ofother verbal operants should also be furtherexamined, as should its role in bringingbehavior under the control of establishingoperations.The research methodology employed in

the study may also be of use in future verbalbehavior studies. The concurrent-chain pro-cedure allows for within-subject com-parisons which are important to verbalbehavior research because of the diverse ver-bal history of each person. This procedurecould be effective for studying the other ver-bal operants as well. The current researchwas exploratory in the sense that conditionsand dependent variables were changed in anattempt to follow events that occurred dur-ing the study. This approach to research wasinspired by Skinner's (1956) unformulatedprinciples of scientific method (e.g., "Whenyou run into something interesting, dropeverything else and study it."), and Day's(1975) "radical methodology," where thesubject's behavior becomes an independentvariable, and the experimenter's behavior adependent variable.

In conclusion, specific and nonspecificreinforcement seem to have differentbehavioral effects. The current study showsthese effects in terms of shorter latencies andpreference for specific reinforcement. Thesedata lend support to Skinner's assertion thatthe mand and the tact are different operants,but represent only the beginning of this lineof research. These results also have practicalsignificance in that they may lead to moreeffective work with individuals who havespeech and language impairments.

REFERENCES

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

Day W. F. (1975). Contemporary behaviorism and theconcept of intention. In J. K. Cole and W. J. Arnold(Eds.) Nebraska Symposium on Motivation (pp. 65-131),Lincoln: University of Nebraska Press.

Hall, G., & Sundberg, M. L. (1987). Teaching mands bymanipulating conditioned establishing operations. TheAnalysis of Verbal Behavior, 5, 41-53.

Lamarre, J., & Holland, J. G. (1985). The functionalindependence of mands and tacts. journal ofthe Experi-mental Analysis of Behavior, 43, 5-19.

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

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

Nevin, J. A., & Mandell, C. (1978). Conditioned rein-forcement and choice. ournal of the ExperimentalAnalysis of Behavior, 29, 135-148.

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

Skinner, B. F. (1956). A case history in scientific method.American Psychologist, 11, 221-233.

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

Sundberg, M. L. (1987). Teaching language to thedevelopmentally disabled: A course manual. PrinceGeorge, B. C.: College of New Caledonia Press.