nonverbal behaviorcorrelatedwiththe shaped verbal behavior ofcatania/abacnj/1990 avb nonvb... ·...

13
The Analysis of Verbal Behavior 1990, 8,43-55 Nonverbal Behavior Correlated With the Shaped Verbal Behavior of Children A. Charles Catania University of Maryland Baltimore County and University College of North Wales AND C. Fergus Lowe and Pauline Home University College of North Wales Children under 6 years old pressed on response windows behind which stimuli appeared (star or tree). Presses occasionally lit lamps arranged in a column; a present was delivered when all lamps were lit. A random-ratio schedule in the presence of star alternated with a random- interval schedule in the presence of tree. These contingencies usually did not produce respec- tive high and low response rates in the presence of star and tree, but the shaping of verbal behavior (e.g., "press a lot without stopping" or "press and wait") was sometimes accompa- nied by corresponding changes in response rate. Verbal shaping was accomplished between schedule components during verbal interactions between the child and a hand-puppet, Garfield the Cat, and used social consequences such as enthusiastic reactions to what the child had said as well as concrete consequences such as delivery of extra presents. Variables that may constrain the shaping of verbal behavior in children seem to include the vocabulary avail- able to the child and the functional properties of that vocabulary; the correlation between rates of pressing and what the child says about them may depend upon such variables. When behavior is determined by its con- sequences, it is called contingency-shaped. For example, the rate of a pigeon's pecks on a key is determined by the schedule according to which its pecks produce food; that rate is higher when food deliveries produced by pecks depend on response number than when they depend on elapsed time (ratio versus interval sched- ules). Human behavior, however, is often rule-governed rather than contingency- shaped: to the extent that it is determined This research was made possible by a Senior Research Fellowship from the Fulbright Foundation and a James McKeen Cattell sabbatical award to the first author. We thank Gaerwyn Williams and the UCNW technical staff, the several UCNW students who assisted (including Tamsin Cowen, Huw Williams, and Lillemor Crowther), the teachers of the Cae Top School in Upper Bangor and especially Mr. John McBride, Headmaster, and, last but of course not least, the children who participated in the various procedures reported here. For reprints, write the first author at the Department of Psychology, UMBC, 5401 Wilkens Avenue, Baltimore, MD 21228 USA. by verbal behavior, it is determined only indirectly by its consequences (Skinner, 1969). Rule-governed behavior itself is presumably maintained by its conse- quences, and these consequences are typi- cally socially mediated. To the extent that they produce a higher-order class of behavior called rule-following, verbal antecedents of the behavior may override its nonverbal and nonsocial consequences (cf. Catania, Shimoff, & Matthews, 1989). These properties of rule-governed behavior have been examined experimen- tally with adults (e.g., Catania, Matthews, & Shimoff, 1982; Hayes, Brownstein, Zettle, Rosenfarb, & Korn, 1986; Matthews, Catania, & Shimoff, 1985; Shimoff, Matthews, & Catania, 1986). In one study, a student's presses on left and right but- tons occasionally produced points later exchangeable for money. Lights above the buttons lit alternately, indicating which button was operative. A number-based 43

Upload: others

Post on 28-Mar-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

The Analysis of Verbal Behavior 1990, 8,43-55

Nonverbal Behavior Correlated With the ShapedVerbal Behavior of Children

A. Charles CataniaUniversity of Maryland Baltimore Countyand University College of North Wales

AND

C. Fergus Lowe and Pauline HomeUniversity College of North Wales

Children under 6 years old pressed on response windows behind which stimuli appeared (staror tree). Presses occasionally lit lamps arranged in a column; a present was delivered when alllamps were lit. A random-ratio schedule in the presence of star alternated with a random-interval schedule in the presence of tree. These contingencies usually did not produce respec-tive high and low response rates in the presence of star and tree, but the shaping of verbalbehavior (e.g., "press a lot without stopping" or "press and wait") was sometimes accompa-nied by corresponding changes in response rate. Verbal shaping was accomplished betweenschedule components during verbal interactions between the child and a hand-puppet,Garfield the Cat, and used social consequences such as enthusiastic reactions to what the childhad said as well as concrete consequences such as delivery of extra presents. Variables thatmay constrain the shaping of verbal behavior in children seem to include the vocabulary avail-able to the child and the functional properties of that vocabulary; the correlation between ratesof pressing and what the child says about them may depend upon such variables.

When behavior is determined by its con-sequences, it is called contingency-shaped.For example, the rate of a pigeon's peckson a key is determined by the scheduleaccording to which its pecks produce food;that rate is higher when food deliveriesproduced by pecks depend on responsenumber than when they depend onelapsed time (ratio versus interval sched-ules). Human behavior, however, is oftenrule-governed rather than contingency-shaped: to the extent that it is determined

This research was made possible by a SeniorResearch Fellowship from the Fulbright Foundationand a James McKeen Cattell sabbatical award to thefirst author. We thank Gaerwyn Williams and theUCNW technical staff, the several UCNW studentswho assisted (including Tamsin Cowen, HuwWilliams, and Lillemor Crowther), the teachers of theCae Top School in Upper Bangor and especially Mr.John McBride, Headmaster, and, last but of course notleast, the children who participated in the variousprocedures reported here. For reprints, write the firstauthor at the Department of Psychology, UMBC, 5401Wilkens Avenue, Baltimore, MD 21228 USA.

by verbal behavior, it is determined onlyindirectly by its consequences (Skinner,1969). Rule-governed behavior itself ispresumably maintained by its conse-quences, and these consequences are typi-cally socially mediated. To the extent thatthey produce a higher-order class ofbehavior called rule-following, verbalantecedents of the behavior may overrideits nonverbal and nonsocial consequences(cf. Catania, Shimoff, & Matthews, 1989).These properties of rule-governed

behavior have been examined experimen-tally with adults (e.g., Catania, Matthews,& Shimoff, 1982; Hayes, Brownstein, Zettle,Rosenfarb, & Korn, 1986; Matthews,Catania, & Shimoff, 1985; Shimoff,Matthews, & Catania, 1986). In one study,a student's presses on left and right but-tons occasionally produced points laterexchangeable for money. Lights above thebuttons lit alternately, indicating whichbutton was operative. A number-based

43

Page 2: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

44 A. CHARLES CATANIAet al.

schedule operated for the left button (ran-dom ratio or RR: the last of a random num-ber of presses produced a point) and atime-based schedule for the right button(random interval or RI: the first press aftera random time since the last point deliveryproduced a point). Between periods ofresponding, the student completed sen-tences such as "The way to earn pointswith the left button is to...." Pointsawarded for sentence completions wereused to shape verbal behavior, i.e., theywere awarded for successively closerapproximations to particular statementsabout performance (cf. Greenspoon, 1955).The nonverbal performance, button

pressing, was typically determined not bythe contingencies arranged for pressing butrather by the student's verbal behavior.The shaping of statements that pointsdepended on slow left pressing and fastright pressing produced correspondingpressing rates, slow left and fast right, eventhough these schedules respectively pro-duce high and low rates in nonhuman con-tingency-shaped performances.The correlation between shaped verbal

behavior and nonverbal responding inadults presumably has its antecedents inthe contingencies that operate on the ver-bal and nonverbal behavior of children,

so 5-05M

40 O STARU TME

0

so80 _5-04M

400L

co

w 0

Ero0 5-03F

40

00 20 40 60 80 100

COMPONENTS

Fig. 1. Effects of low-rate contingencies during starand high-rate contingencies during tree on the multi-ple-schedule response rates of three children. Thechildren are identified by age in years and months (Y-MM) and by gender (F or M). Over roughly 2 to 3hours of multiple-schedule components, these ratecontingencies did not produce differential responserates.

and the present research was designed toexplore with children the effects of proce-dures analogous to those used with adults.The transition from contingency-shapedperformances that resemble those of non-human organisms to the rule-governedperformances that are characteristic ofhuman adults seems usually to occur sometime between 2 and 6 years of age (Bentall,Lowe &, Beasty, 1985; Lowe, Beasty, &Bentall, 1983). The differential respondingreliably obtained with nonhuman organ-isms (such as the higher rates maintainedby ratio than by interval schedules) islikely to be absent in the behavior of chil-dren more than 5 years old.

Figure 1 provides an example. Star andtree, presented by a computer monitorbehind response windows, were the twostimuli of a multiple schedule arranged fora child's presses. The presses occasionallylit lamps arranged in a column; a presentwas delivered when all lamps were lit.Alternating components lasted 60 s each.In the first (star), an RI 10-s schedulearranged consequences for satisfying alow-rate contingency; in the second (tree),it did so for satisfying a high-rate contin-gency (see Catania, Horne, & Lowe, 1989,for details).Data from three children are shown

(each child is coded by age in years andmonths, Y-MM, and by gender, F or M; 5-05M is the child whose data are reported inCatania, Horne, & Lowe, 1989). Unfilledcircles show rates from the low-rate com-ponent (star); filled squares show ratesfrom the high-rate component (tree). Thelow-rate and high-rate contingencies wereintroduced after 3 or 4 sessions of multipleRI RI schedules. The plan was to increasethe stringency of the rate requirementsgradually over sessions so as to produce arate during the high-rate component twoto four times higher than that in the low-rate component. Over successive sessions,no separation of response rates occurred.Those sessions included many periodsduring which the low-rate and high-ratecontingencies were either or both satisfied,but such periods were not consistently fol-lowed by continuations of the perfor-

Page 3: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

SHAPED VERBAL BEHAVIOR IN CHILDREN 45

mances that had satisfied those contingen-cies. In subsequent sessions for each child,however, verbal interventions rapidly pro-duced differential response rates in the twocomponents.Demonstrating reduced sensitivity to

schedule contingencies with age leavesopen the question of the basis for thatreduced sensitivity. It can be argued thatolder children have longer histories of con-tact with ranges of contingencies and thatthese histories interact with effects ofschedules (cf. Wanchisen, Tatham, &Mooney, 1989), or that the effectiveness ofreinforcers varies with age and that rein-forcers arranged for the behavior of olderchildren are not comparable to those usedwith infants or nonhumans (cf. Perone,Galizio, & Baron, 1988), or that verbalbehavior involves social contingencies forrule-following that interfere with the directeffects of nonverbal contingencies (cf.Catania, Matthews, & Shimoff, 1990).Performances that are insensitive to

schedule differences, such as those ofFigure 1, might be changed to sensitiveones by providing more exposure to theschedules, by arranging different histories,by varying schedule parameters, byswitching to different reinforcers, or bymanipulating verbal behavior. But demon-strating an effect of any one of these vari-ables does not in itself establish that vari-able as the source of the insensitivity. Forexample, a performance that originated asan instance of rule-following might never-theless be susceptible to variations inexperimental history, schedule parametersor reinforcer magnitudes.Whatever the source of the insensitivity

to contingencies, rule-governed behavior isdemonstrated when such performancescan be modified by manipulating verbalbehavior. The present account is con-cerned primarily with the effects of shapedverbal behavior, and describes attempts toreplicate with children the shaping of ver-bal behavior and the correlated changes innonverbal performance that have beenobtained with adults. It should thereforebe regarded as an exploratory study thatidentifies some procedural prerequisites

and pitfalls in the maintenance of the non-verbal behavior of children and in theshaping of their verbal behavior.

METHODSubjectsChildren were selected for the study on

the basis of age and times of availabilityfrom classes of the Cae Top elementaryschool in Bangor, North Wales. As judgedby their teachers, all were fluent in Englishat levels appropriate to their ages; somewere bilingual Welsh and English speak-ers. The continued participation of eachchild depended on uncontrollable factorssuch as apparatus problems or reliability ofattendance (especially owing to illness orfamily holidays). The 7 girls and 7 boyswho served in various stages of the experi-ment ranged in age from 4-02 to 5-11; eachchild for whom data are presented is iden-tified by an arbitrary 2-letter code followedby age in years and months at the start ofparticipation (Y-MM) and gender (F or M).No data from children for whom verbal

.... _.....

e:i-B L M

K>

Fig 2 A child pressing a window on the experimen-tal panel. A computer monitor behind the tworesponse windows presents stimuli. The five bottomcolumn lamps on the right panel have been lit; a pre-sent will be placed on the shelf near the top of thecenter panel when the child has lit the remainingthree. Garfield the Cat, a hand puppet, looks on fromthe left panel

Page 4: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

46 A. CHARLES CATANIAet al.

shaping was successful have beenexcluded, except for one case in which ver-bal shaping was followed by sessions thatincluded explicit instructions.

Apparatus and settingFigure 2 shows a child seated before the

main experimental panel, which wasclamped to a desk along one wall of ateachers' area (see Catania, Horne, &Lowe, 1989, for additional details). Ametal plate on it contained two responsewindows, each 5-cm square (a metal coveron the right masked three other windows).Stimuli could be presented in the windowsby a computer monitor located behind thepanel. The monitor was connected to acomputer that arranged experimental con-tingencies. A microphone fastened at thelower left allowed sessions to be recordedon audio cassette. Presents earned by thechild were placed on a shelf located behinda Plexiglas-covered opening in the upperpart of the panel.A column of eight lamps behind translu-

cent colored inserts (red, blue, yellow, andwhite) was mounted on the right extensionof the panel. The left extension screenedthe experimenter(s) and apparatus fromthe child, and included a curtained open-ing through which, as shown, a hand pup-pet (Garfield the Cat) occasionally emergedto interact with the child.

ProcedureInitial sessions. Sessions were conducted

on a semi-regular basis during morningand afternoon school hours and varyingwith other scheduled school activities andoccasional absences. The youngest chil-dren, from a preschool class, were avail-able only during morning school hours.On the day of the first session, childrenwere escorted individually from theirclassrooms to the experimental area by aUCNW student who assisted in the proce-dures. The general procedure arranged forall children is described here for a repre-sentative girl.Upon arrival in the experimental area,

the child was given a scrapbook and washelped to write her name in it. Next she

was shown sheets of decals with a broadselection of pictures, and then a "treasurechest," a box containing a variety of smalltoys, drawing materials, etc. She wasasked if she would like to play a game inwhich she could win pictures for her scrap-book and a chance to pick out somethingfrom the treasure chest. Upon her assentshe was shown to the stool facing theexperimental panel, where Garfield the Catemerged from the curtained opening in theleft panel (cf. Figure 2).

Garfield introduced himself, asked thechild's name, and explained that she couldget presents by lighting up all of the lampson the right panel. The lamps were then litone by one, starting at the bottom. As eachnew lamp was added, it and the previouslylit ones blinked five times at a rate of 5 persecond before remaining continuously on;each blink was accompanied by a com-puter beep the pitch of which increasedwith the height of the lit lamps in the col-umn. This sequence of events as lit lamps

l RR

* RI

Fig. 3. Multiple-schedule stimuli as presented by thecomputer monitor. Two components alternated.Presses produced consequences according to a ran-dom-ratio (RR) schedule during component 1 (star,shown in top window) and according to a random-interval (RI) schedule during component 2 (tree,shown in bottom window).

Page 5: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

SHAPED VERBAL BEHAVIOR IN CHILDREN 47

were added to the column remained ineffect throughout the experiment.When all lamps of the column were lit,

one of the small boxes that served as pre-sents was inserted on the upper shelf. Thelamps blinked until shortly after the pre-sent was placed and then went out.Garfield called the present to the child'sattention, and then a star appeared in theupper of the two available response win-dows. Garfield, telling the child to watch,pressed that window and the bottom lampturned on. Garfield then said, "Now youtry." When she pressed, the second lampturned on. Each of her next presses litanother lamp until the entire column wasagain lit, at which point Garfieldannounced the delivery of another presentand withdrew. At this point, the multiple-schedule procedure was instituted.Multiple schedule. Figure 3 shows the

stimuli correlated with the two multiple-schedule components: during component1, a star was presented on the top window(bottom window dark) and a random-ratio(RR) schedule operated; during component2, a tree was presented on the bottom win-dow (top window dark) and a random-interval (RI) schedule operated. Com-ponents alternated; each session beganwith component 1. With the change inlocation, whether the child was attendingto the visual stimuli could be determinedeven when response rates in the two com-ponents were roughly equal (note, how-ever, that following of stimulus locationdemonstrated control by the window thatwas lit but not necessarily by the stimulusthat appeared in that window).Each component lasted 60 s, excluding

fixed periods during which the lampsblinked as new ones were added to the col-umn and variable periods when presentswere delivered (the latter, under experi-menter control, were to allow for Garfield'sentrances and exits and to insure that thechild attended to the delivery of each newpresent). During these periods and 5-speriods between multiple-schedule compo-nents, the stimuli behind the response win-dows were off, presses on the response

windows had no consequences, and sched-ules did not operate.Multiple-schedule sessions typically

lasted 4 to 6 pairs of components (10 to 20min), though occasional sessions wereshorter because of interruptions for vari-ous reasons (e.g., the child asking to leavefor the bathroom, or a session extendinginto a time when the child had to be else-where); occasional sessions wereextended for one or two additional pairswhen a child wanted to continue (chil-dren often addressed the experimenterbehind the panel, and/or Garfield even inhis absence). At the end of each session,the child selected one decal for her scrap-book for each present she had earned andalso chose an item from the treasurechest.During the first session for a given child,

an RR 5 schedule operated in component 1and an RI 5-s schedule in component 2.These values were typically both increasedfrom 5 to 8 in the second session and from8 to 10 in the third and all subsequent ses-sions, except that the lower value wasmaintained in a component for one ormore additional sessions if response rateswere not well maintained in that compo-nent.The RR schedule made each press eligi-

ble to produce a consequence with a prob-ability equal to the reciprocal of the RRvalue. In the RI schedule a setup corre-sponded to the eligibility of the nextresponse to produce a consequence. Onceper second with a probability given by thereciprocal of the RI value in s (e.g., p = .10for RI 10 s), 1 was added to availablesetups. If at least one setup was available,the next press on the appropriate windowlit the next lamp in the column andreduced available setups by 1. Within ses-sions, setups accumulated at the end of oneRI component were saved until the start ofthe next RI component. This arrangementdiffers from those in which RI timing stopsafter a single setup in two ways: latenciesfrom scheduling to production of a conse-quence do not accumulate, thus reducingthe difference between rates of obtainedand scheduled consequences; and two or

Page 6: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

48 A. CHARLES CATANIAet al.

more responses in a row can produce con-sequences if more than one setup has accu-mulated. Overall, actual rates were closeto the 6 lamps per 60-s component speci-fied by the RI 10-s schedule.

The shaping of verbal behavior. The deci-sion to shape consistent with or in opposi-tion to the contingencies was made inde-pendently of performance, and the start ofverbal shaping was based on the numberof sessions for which a given child wouldcontinue to be available for the experimentrather than upon any criteria for stabilityof baseline performance (the final sessionof the experiment, with Child RT 5-02M,was conducted on the next-to-last day ofschool preceding the summer vacation).At the start of verbal shaping, Garfield

began to appear during an extended inter-component interval that followed eachcomponent pair. During that time, heasked the child questions about playing thegame that were designed to lead her to saythat the tree worked best when pressedslowly and the star worked best whenpressed fast (e.g., Garfield: "I'd like tolearn how to play this game. Can you tellme how the tree works? Should I pressand wait or should I press a lot withoutstopping." Child, after a pause: "Press andwait." Garfield, excitedly: "Oh, Thatsounds like a good idea. That's what I'lltry when I play the game."). Thereafter,attempts were made to shape verbal elabo-rations that incorporated the stimuli andthat included both components (e.g., Child:"When there's a tree you press and wait,and when there's a star you press a lotwithout stopping"). The duration of theinterval varied with such factors as theprogress of verbal shaping and the child'spreference for interacting with Garfield orplaying the game.

This procedure might be regarded as acombination of verbal prompting and/or avariety of verbal shaping in which the rein-forcing consequence is provided by the lis-tener's behavior (Greenspoon, 1955; cf.Catania et al., 1982). To the extent thatGarfield's comments often included state-ments about properties of performance, asin the above example, the procedure might

also be regarded as instructional (cf.Bentall & Lowe, 1987). It is also reasonableto assume that the situation may haveincluded implicit demand characteristics;hand puppets as well as adults may exertsuch effects on the behavior of children(e.g., Moore & Frye, 1986). All of the chil-dren interacted with Garfield but knewthat he was a hand puppet operated by anexperimenter behind the partition (in oneinstance, reminiscent in its effect of thescene in which Toto pulled the curtainaway from the Wizard of Oz, a childpeeked around the partition and said,"Hello, Professor").Shaping procedures varied in the conse-

quences arranged for the child's verbalbehavior. In the earliest sessions of verbalshaping, it was assumed that features ofGarfield's response to the child's utterance(e.g., Garfield, enthusiastically: "Welldone! That's a good guess!") would pro-vide effective reinforcers, as in ordinarysocial interactions. In most later instancesof verbal shaping, presents of the same sortas those provided for multiple-scheduleresponding were added (e.g., Garfield:"That was such a good guess that I'mgoing to give you an extra present!").Unless otherwise noted, this procedurewas used with all of the children for whomdata are presented.

In what seemed to be the most effectiveprocedure, though used with only onechild (RT, Figure 6), a satisfactory guesshad the same consequences as were sched-uled for presses in the multiple schedule:turning on an additional light on the col-umn. When this first happened, Garfieldpointed this out to the child: "Look, whenyou said that another light came on. Whydon't you try saying it again and see whathappens?" Other aspects of shaping arediscussed in conjunction with the presenta-tion of data.During verbal shaping, Garfield

refrained from commenting on the relationbetween the child's guesses and the child'sresponse rates. At the end of a givenchild's experimental participation, Garfieldsometimes had an opportunity to askabout the correspondence (or lack thereof)

Page 7: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

SHAPED VERBAL BEHAVIOR IN CHILDREN 49

between guesses and rates; those verbalreports, which could not be obtained sys-tematically, were variable but usually con-sistent with the child's verbal behaviorduring sessions.When the experimenter was satisfied

that appropriate verbal behavior had beenshaped, Garfield continued to appear dur-ing extended intercomponent intervals thatfollowed each pair of components. Duringthese conditions, Garfield asked reminderquestions (e.g., "I don't remember whatyou told me about the star and the tree.Can you tell me again?"), and differentialconsequences were maintained for thechild's verbal responses.As is necessarily the case in a shaping

procedure, judgments of the appropriate-ness of the shaped verbal behavior weremade at the time consequences werearranged; taped sessions of verbal shapingwere reviewed only after the experimentwas concluded. It was not practical tocount social and conversational reinforcers,and deliveries of presents or lit lamps var-ied not only with the progress of verbalshaping but also with available time, thechild's rate of talking, and other uncontrol-lable factors. Thus, quantitative measures

of the progress of shaping, as in Catania etal. (1982), were not feasible.

RESULTSGiven the exploratory nature of these

procedures, details changed as a conse-quence of the interaction of the childrenand the experimenters, but variabilityacross children makes it impractical to pre-sent the data chronologically. Figure 4contrasts a case of successful verbal shap-ing (LY 5-1OF) with one in which verbalshaping was unsuccessful (KE 5-1iF); inboth cases, however, verbal behavior wasconsistent with performance.The shaping of LY's verbal behavior was

initiated after eight sessions during whichno systematic RR and RI rate differencehad emerged. Within the first session ofshaping "You press the star quickly" wasestablished as a response to "What's thebest way to make the star work?," and"You press the tree slowly" as a responseto "What's the best way to make the treework?" Corresponding rates of pressingdeveloped within the same session; com-parable verbal behavior was maintainedover the next four sessions, and wasaccompanied by higher rates during starthan during tree throughout those ses-

80 ,. .

70 -100 KE 5-1hF

6080 -

50

Z 60-40

CD 0 40

20 -0- STAR-RR -0-- STAR-RR-U-TREE-RI I2 U-TREE-RI

10

00 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70

COMPONENT COMPONENT

Fig. 4. Response rates during multiple-schedule components for two children. A random-ratio schedule operatedduring one stimulus (STAR-RR) and a random-interval schedule during the other (TREE-RI). Mean response ratesover sessions are plotted against components (e.g., LY's first session included 3 presentations of each component,and data from that session are plotted at component 2). Verbal shaping began at the dashed vertical line; the direc-tion of verbal shaping, shown by the arrows (star, fast; tree, slow), was consistent with typical RR and RI effects incontingency-shaped performances. Data from LY illustrate a maintained effect of verbal shaping on response rate,whereas those from KE illustrate performance in a case of unsuccessful verbal shaping with a highly verbal child.

Page 8: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

50 A. CHARLES CATANIAet al.

100 . 60 . .

70

s060

40

CO 4B030w

2020

--- TE-I-- STAR4ARR0TREE-RI ~~~~~~10--TREE-RI0 -,0

0 10 20 30 40 s0 60 70 s0 0 10 20 30 40 s0 60 70 s0COMPONENT COMPONENT

60 . . -y-., .. . . . . 5 I *

HY 5-09M | sos0

--STAR-RR-f-TREE-RI

40 1 60

z30-

COMW 20

cc 20 -~~~~~~~~~~~~~~~~~0--STAR-ARR

10

0 10 20 30 40 s0 60 70 s0 C0 10 20 30 40 so 60 70

COMPONENT COMPONENT

Fig. 5. Response rates during multiple-schedule components for four children, with details as in Figure 4. For bothBN and JE, the direction of verbal shaping, shown by the arrows, was opposed to typical effects of RR and RI incontingency-shaped performances. For JE, slow and fast referred to speed rather than rate of hand movement; atransient effect of verbal shaping was obtained with a change of vocabulary (star, press and wait; tree, press a lotwithout stopping). For both HY and CG, response rates in later sessions were consistent with schedule contingen-cies; HY responded to yes-no questions but rarely initiated other verbal behavior, and with CG no verbal behaviorwas established through verbal shaping.

sions. The RR and RI rate differences in thefirst, second and fourth sessions of verbalshaping for LY were greater than anyobserved during baseline sessions acrossall children in the study (the largest base-line differences occurred in BN's sixthbaseline session and RT's fifth; see Figures5 and 6).For KE, verbal shaping began after ten

sessions. Her characteristic response toquestions about either star or tree was "Ilike to press them both the same," and herresponse rates were consistent with thatverbal behavior. It was occasionally possi-ble to get her to say something aboutpressing the star fast and the tree slow, butsuch verbal behavior was short-lived and

often followed by a correcting statementsuch as "I say slow but I press fast" (chil-dren often talked to Garfield during com-ponents, even though Garfield, havingwithdrawn behind the curtained window,was not present). Response rates increasedin both components over sessions, but dif-ferential response rates were not obtained.

Figure 5 presents data from four otherchildren. With BN (5-liM), verbal shapingwas attempted in opposition to the sched-uled contingencies. It began with socialconsequences only; these were supple-mented by extra presents beginning in thefourth session of verbal shaping. Much ofhis verbal behavior early during verbalshaping consisted of agreement with

Page 9: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

SHAPED VERBAL BEHAVIOR IN CHILDREN 51

Garfield's statements (e.g., responding"Yes" to "Does the star work better if youpress and wait?"). As shown in Figure 5, atransient rate difference occurred in thesecond session of verbal shaping, duringwhich relevant verbal behavior seemed tohave been established; the magnitude ofthis difference was larger than that in thesixth baseline session. Thereafter, how-ever, rates during star and tree wereroughly equal and verbal behavior wasvariable; in particular, BN's verbalresponses sometimes reversed the relationbetween response rates and stimuli.The data for JE (5-07M) illustrate another

case in which shaping was attempted inopposition to the scheduled contingencies.The first two sessions of verbal shapingwere not accompanied by correspondingchanges in response rate, but topographiesof pressing in the presence of star and treewere considerably different. At the begin-ning of the third session, Garfield asked JEto show him how to press fast or slow, andit became clear that these terms corre-sponded to different speeds of hand move-ment toward and away from the responsewindows, even though rates of pressingwere roughly equal. For JE, as for manychildren at this age or younger, fast and

120 . T -. r-

RT 5-02M100

--STAR-RR-U-TREE-RI

so

20

00 10 20 30 40 50 60

COMPONENTFig. 6. Response rates during multiple-schedule com-ponents for RT, illustrating the effects of a reversal ofverbal behavior. At the first dashed line, shapingestablished verbal behavior consistent with typicaleffects of RR and RI contingencies (star, press a lotwithout stopping; tree, press and wait). After a ratedifference emerged, the direction of verbal shapingwas reversed (second dashed line); as the verbalbehavior changed, it was accompanied by a reversalof response rates. Details as in Figure 4.

slow applied to rate of movement throughspace and not to event frequencies. In thethird session of verbal shaping, Garfieldestablished "press and wait" and "press alot without stopping" as substitutes for"fast" and "slow," and a correspondingdifference in rate emerged. This differencediminished over the remaining three ses-sions, during which JE often failed to lookat the stimuli while pressing and devel-oped a "sing-song" patter (e.g., "with thestar you press and wait, with the tree youpress a lot") that was maintained duringboth stimuli but was unrelated to rates ofpressing.

For HY (5-09M), responding was estab-lished very slowly; because of lowresponse rates during the RR component(star), this schedule was not increased toRR 10 until session 5 (component 25).Verbal shaping, consistent with contingen-cies, was attempted beginning with session12, but virtually all of HY's verbal behaviorconsisted of yes or no responses toGarfield's questions; questions for whichyes or no responses were inappropriatewere typically unanswered. Nevertheless,a rate difference emerged, but it had beenevident prior to verbal shaping and mightbest be regarded as behavior under the dif-ferential control of the RR and RI contin-gencies.The only other case in which such

behavior was observed was with theyoungest child studied, CG (4-02M). Forthis child, higher rates on the RR than onthe RI schedule emerged by session 10, andover the remaining seven sessions ratesmaintained by the RR schedule (star) wereroughly 50% higher than those maintainedby the RI schedule (tree): 51 and 33responses per min, respectively. Rateswere variable, however, and included sub-stantial rates of dark-window and inter-component pressing. Attempts at verbalshaping were aimed not only at statementsabout differential responding in the pres-ence of each stimulus, but also at state-ments about looking at the stimuli whilepressing. This shaping too was unsuccess-ful, however. The differential RR and RI

Page 10: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

52 A. CHARLES CATANIAet al.

rates are therefore presumably instances ofcontingency-shaped responding.

Figure 6 shows data obtained when thelighting of column lamps was substitutedfor social consequences and extra presentsin shaping the verbal behavior of RT (5-02M). In this instance, the shaping wassufficiently effective that it was possible toarrange a reversal. First, verbal behaviorthat was consistent with contingencies wasshaped. This shaping was accompanied bya corresponding separation of responserates; the RR and RI rate difference in thesecond session of shaping (more than 50responses/min) was the largest obtainedfor any child in any session throughout thestudy. A reversal of the shaping, so thatthe verbal behavior was opposed to thedifferential effects of RR and RI contingen-cies, began two sessions later. The reversalof the verbal behavior was accompaniedby a gradual reversal of response rates.Data have been presented for 7 of the 14

participants in the experiment. Theremaining seven ranged in age from 4-04to 5-11. For various reasons (see Method),sessions were discontinued prior to verbalshaping for five of those. No systematicdifferences in RR and RI response ratesemerged in any of those cases, and theirdata are not relevant to conclusions aboutthe relation between nonverbal behaviorand shaped verbal behavior (in a compara-ble situation with nonhuman subjects, suchinstances would probably not even enterinto a data presentation; for example, thereport of a study of schedule effects on thekey-pecking of pigeons would not ordinar-ily mention a pigeon who was excludedfrom the experiment because of a problemin the shaping of its key pecks).Of the remaining two, in the case of ZA

(5-11F), Garfield switched to direct instruc-tions after two or three sessions of unsuc-cessful verbal shaping (e.g., Garfield: "Trypressing the star a lot without stopping").This intervention produced differentialresponse rates but confounded the effectsof the direct instructions with those of ver-bal shaping (for studies of the effects ofinstructions on schedule performances, seeBaron & Galizio, 1983; Hayes, Brownstein,

Haas, & Greenway, 1986; LeFrancois,Chase, & Joyce, 1988). Finally, in the caseof NA (5-08F), verbal shaping was discon-tinued when it became clear that the datawould be uninterpretable because of sub-stantial variability in response rates. Thevariability came about because NA fre-quently paused for long periods of time tocount and recount lights and presents dur-ing multiple-schedule components.

In summary, verbal shaping correlatedwith corresponding changes in nonverbalresponding was demonstrated in at leastthree cases: LY, JE (transiently with respectto response rate, but also with a change intopography that followed a change in ver-bal behavior) and RT (including a rever-sal). Although verbal shaping was unsuc-cessful with KE, her verbal behavior wasconsistent with her performance, and theambiguous verbal shaping for BN wasaccompanied by a transient rate difference.In the remaining two cases, HY and CG,verbal shaping was unsuccessful and per-formance was characteristic of contin-gency-shaped rather than rule-governedbehavior. Furthermore, some of the RRand RI rate differences that followed verbalshaping were larger than those obtained inany baseline sessions, and there were noinstances of rate differences that wereopposed to rather than consistent with ver-bal behavior.

DISCUSSIONThe present experiments should be taken

mainly as demonstrations. They show thatsome varieties of verbal behavior can beshaped with 4-yr-old to 6-yr-old children,and that such verbal behavior is sometimescorrelated with changes in correspondingnonverbal behavior. The verbal behaviorwas produced by shaping before the corre-sponding nonverbal behavior emerged,thereby establishing that the direction ofcontrol was from verbal behavior to non-verbal. Furthermore, verbal behavior wassometimes accompanied by differentialresponse rates that were opposed, even ifonly transiently, to the effects of RR and RIcontingencies. In such cases, it is appropri-ate to refer to the nonverbal behavior as

Page 11: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

SHAPED VERBAL BEHAVIOR IN CHILDREN 53

rule-governed. The results were variable,but that variability is what should beexpected if the ages at which adult correla-tions between verbal and nonverbal behav-ior begin to develop are within the range ofages included in this study.

It is tempting to conclude from thesefindings that it is more difficult to shapeverbal behavior in opposition to RR and RIcontingencies than consistent with them, orthat the effectiveness of verbal shapingvaries depending on whether social or con-crete consequences were used, or that dif-ferent outcomes consistently depend onage, gender, or other individual differ-ences. It would be inappropriate to try tosummarize these results in such terms,however, because the verbal shaping pro-cedure and, perhaps more important, theexperimenter's skill in verbal shaping,evolved over the group of children withwhom it was used and were perhaps asimportant as other sources of variability inthese procedures. The evolving proce-dures are relevant to the art of verbal shap-ing, but there is no a priori reason toassume that the observed correlationsbetween verbal and nonverbal behaviorshould be evaluated differently dependingon the efficiency with which shaping estab-lished the verbal behavior.One assumption implicit in the demon-

strations is that sensitivity to the RR and RIcontingencies would have occurred withnonhuman subjects. But cross-species com-parisons involve problems of equating thepotency of reinforcers, the effectiveness ofstimuli, the integrity of response classes,and so on. A comparison across children ismore appropriate than comparisons acrossspecies, and it is therefore worth notingthat the demonstrations include both casesthat seem to qualify as examples of contin-gency-shaped behavior and those thatseem to qualify as examples of rule-gov-erned behavior.

Insensitivity to the RR and RI contingen-cies might also be attributed to the limitedpotency as reinforcers of the consequencesarranged for the nonverbal behavior. Aswith comparisons across species, problemsare raised by comparisons across proce-

dures, ages, and so on. Yet the conse-quences that maintained pressing in theseprocedures were the same as those used inthe successful shaping of verbal behavior,and to that extent were functional and notmerely nominal reinforcers. *This is consis-tent with the rule of thumb that verbalbehavior is typically contingency-shapedin adult human behavior even thoughmuch nonverbal behavior is rule-governed(perhaps because human verbal communi-ties have not established extensive vocabu-laries concerned with the determinants ofverbal behavior; cf. Catania, Shimoff, &Matthews, 1989; Skinner, 1957).

In the vocabulary of verbal behavior(Skinner, 1957), the verbal behavior thatwas shaped in these procedures might becharacterized as intraverbal (though it mayalso have included substantial autocliticcomponents). To the extent that a wordsuch as "tree" occasions "press and wait,"the latter utterance might occur once "tree"is emitted as a tact during the relevantmultiple-schedule component. But "tree"as part of an intraverbal sequence and"tree" as a tact are members of differentoperant classes, and the ways in which dif-ferent verbal classes with similar topogra-phies can come to share their functions arejust beginning to be elaborated (cf.Stafford, Sundberg, & Braam, 1988;Watkins, Pack-Teixteira, & Howard, 1989).Even if this aspect of the performance is

resolved, it remains to demonstrate howthe utterance "press and wait" might thenoccasion the corresponding nonverbalbehavior. In other words, there is stillmuch to learn about the contingencies thatbring about correspondences between say-ing and doing (Risley & Hart, 1968). Thecontingencies that may shape such corre-spondences are necessarily more complexthan those that operate separately on thesaying and on the doing (cf. Baer, Detrich,& Weninger, 1988; Matthews, Shimoff, &Catania, 1987). Furthermore, relationsbetween verbal and nonverbal behaviorinvolve both directions of control: fromverbal to nonverbal, as in some of the pre-sent cases, and vice versa, as when oneaccurately describes one's own behavior.

Page 12: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

54 A. CHARLES CATANIAet al.

Say-do correspondences share this symme-try with equivalence classes (Sidman,Wynne, Maguire, & Barnes, 1989), and itmay be appropriate to speculate that thesetwo types of classes are functionally aswell as formally related.

In any case, the present demonstrationsset the stage for more precise characteriza-tions of the verbal behavior of children andthe sharper definition of verbal shapingand other experimental details. With thestandardization of such procedures, themonitoring of nonverbal performancessuch as those in the present multipleschedules might provide a tool for analyz-ing how the functional properties of verbalbehavior and its correlations with nonver-bal behavior develop.The demonstration of contingency-

shaped or rule-governed multiple RR RIperformances by children of different ages,with or without verbal shaping, may haveapplied relevance. For example, considerhow obtaining one rather than the otherperformance from a nonverbal institution-alized child might influence the prognosisfor establishing verbal behavior through alanguage program for that child. The dis-tinction between rule-governed versuscontingency-shaped may also be useful indiagnosis and perhaps even in treatment,to the extent that performance differencesturn out to be correlated with behavioraldeficits (e.g., aphasia) or other diagnosticcategories (e.g., autism).

REFERENCESBaer, R. A., Detrich, R., & Weninger, J. M. (1988). On

the functional role of the verbalization in corre-spondence training procedures. Journal of AppliedBehavior Analysis, 21, 345-356.

Baron, A., & Galizio, M. (1983). Instructional controlof human operant behavior. Psychological Record, 33,495-520.

Bentall, R. P., & Lowe, C. F. (1987). The role of verbalbehavior in human learning: III. Instructionaleffects in children. Journal of the ExperimentalAnalysis of Behavior, 47, 177-190.

Bentall, R. P., Lowe, C. F., & Beasty, A. (1985). Therole of verbal behavior in human learning: II.Developmental differences. Journal of theExperimental Analysis of Behavior, 43, 165-181.

Catania, A. C., Horne, P., & Lowe, C. F. (1989).Transfer of function across members of an equiva-lence class. The Analysis of Verbal Behavior, 7, 99-110.

Catania, A. C., Matthews, B. A., & Shimoff, E. (1982).Instructed versus shaped human verbal behavior:Interactions with nonverbal responding. Journal ofthe Experimental Analysis of Behavior, 38, 233-248.

Catania, A. C., Matthews, B. A, & Shimoff, E. (1990).Properties of rule-governed behaviour and theirimplications. In D. E. Blackman & H. Lejeune (Eds.)Behaviour analysis in theory and practice: Contributionsand controversies (pp. 215-230). Brighton, UK:Erlbaum.

Catania, A. C., & Shimoff, E., Matthews, B. A. (1989).An experimental analysis of rule-governedbehaviour. In S. C. Hayes (Ed.), Rule-governedbehavior: Cognition, contingencies and instructionalcontrol (pp. 119-150). New York: Plenum.

Greenspoon, J. (1955). The reinforcing effect of twospoken sounds on the frequency of two responses.American Journal of Psychology, 68, 409416.

Hayes, S. C., Brownstein, A. J., Haas, J. R., &Greenway, D. E. (1986). Instructions, multipleschedules, and extinction: Distinguishing rule-gov-emed from schedule-controlled behavior. Journal ofthe Experimental Analysis of Behavior, 46, 137-147.

Hayes, S. C., Brownstein, A. J., Zettle, R. D.,Rosenfarb, I., & Korn, Z. (1986). Rule-governedbehavior and sensitivity to changing consequencesof responding. Journal of the Experimental Analysis ofBehavior, 45, 237-256.

LeFrancois, J. R., Chase, P. N., & Joyce, J. H. (1988).The effects of a variety of instructions on humanfixed-interval performance. Journal of theExperimental Analysis of Behavior, 49, 383-393.

Lowe, C. F., Beasty, A., & Bentall, R. P. (1983). Therole of verbal behavior in human learning: Infantperformance on fixed-interval schedules. Journal ofthe Experimental Analysis of Behavior, 39, 157-164.

Matthews, B. A., Catania, A. C., & Shimoff, E. (1985).Effects of uninstructed verbal behavior on nonver-bal responding: Contingency descriptions versusperformance descriptions. Journal of the ExperimentalAnalysis of Behavior, 43, 155-164.

Matthews, B. A., Shimoff, E., & Catania, A. C. (1987).Saying and doing: A contingency-space analysis.Journal of Applied Behavior Analysis, 20, 69-74.

Moore, C., & Frye, D. (1986). The effect of experi-menter's intention on the child's understanding ofconservation. Cognition, 22, 283-298.

Perone, M., Galizio, M., & Baron, A. (1988). The rele-vance of animal-based principles in the laboratorystudy of human operant conditioning. In G. C. L.Davey & C. Cullen (Eds.), Human operant condition-ing and behaviour modification (pp. 59-85). Chichester,England: Wiley.

Risley, T. R., & Hart, B. (1968). Developing correspon-dence between the nonverbal and verbal behaviorof preschool children. Journal of Applied BehaviorAnalysis, 1, 267-281.

Shimoff, E., Matthews, B. A., & Catania, A. C. (1986).Human operant performance: sensitivity and pseu-dosensitivity to contingencies. Journal of the Experi-mental Analysis of Behavior, 46, 149-157.

Sidman, M., Wynne, C. K., Maguire, R. W., & Barnes,T. (1989). Functional classes and equivalence rela-tions. Journal of the Experimental Analysis of Behavior,52, 261-274.

Skinner, B. F. (1957). Verbal behavior. New York:Appleton-Century-Crofts.

Skinner, B. F. (1969). An operant analysis of problemsolving. In B. F. Skinner, Contingencies of reinforcement(pp. 133-157). New York: Appleton-Century-Crofts.

Page 13: Nonverbal BehaviorCorrelatedWiththe Shaped Verbal Behavior ofcatania/ABACNJ/1990 AVB nonvb... · 2018-05-20 · by verbal behavior, it is determined only indirectly by its consequences

SHAPED VERBAL BEHAVIOR IN CHILDREN 55

Stafford, M. W., Sundberg, M. L., & Braam, S.J. (1988). A preliminary investigation of theconsequences that define the mand and thetact. The Analysis of Verbal Behavior, 8, 61-71.

Wanchisen, B. A., Tatham, T. A., & Mooney, S. E.(1989). Variable-ratio conditioning history produces

high- and low-rate fixed-interval performance inrats. Journal of the Experimental Analysis of Behavior,52,167-179.

Watkins, C. L., Pack-Teixteira, L., & Howard, J. S.(1989). Teaching intraverbal behavior to severelyretarded children. The Analysis of Verbal Behavior, 7,69-81.