Grammatical Constructions in Typical Developing Children- Effects of Explicit Reinforcement, Automatic Reinforcement and Parity

8/10/2019 Grammatical Constructions in Typical Developing Children- Effects of Explicit Reinforcement, Automatic Reinforce

1/10

Grammatical Constructions in Typical Developing Children:

Effects of Explicit Reinforcement, Automatic Reinforcement

and Parity

Leni stvik, Svein Eikeseth, and Lars Klintwall,

Oslo and Akershus University College

This study replicated and extended Wright (2006) and Whitehurst, Ironsmith, and Goldfein (1974) byexamining whether preschool aged children would increase their use of passive grammatical voice ratherthan using the more age-appropriate active grammatical construction when the former was modeled byan adult. Results showed that 5 of the 6 participants began using the passive voice after this verbalbehavior had been modeled. For 3 of the participants, this change was large. The change occurred eventhough the adult model explicitly rewarded the participant with praise and stickers for using the activevoice, while providing no praise or stickers for using the passive form that was modeled. For 1

participant, the modeling procedure had no effect on use of the passive voice. These results indicate astrong automatic reinforcement effect of achieving parity with the grammatical structures used by adults,compared to the effects of explicit reinforcement by the adult. This might help to explain why childrenacquire grammatical structures prevalent in their language community apparently without explicitinstruction.

Key words: verbal behavior, modeling, automatic reinforcement, parity, children, imitation

Skinner (1957) defined language or verbalbehavior as an operant behavior that isreinforced through the mediation of other

persons (p. 2). This refers to a functionalrelationship between behavior of the speakerand conditions in the speakers social envi-ronment. For verbal behavior to emerge,operant verbal responses must be reinforced

by the verbal community (Skinner, 1957). Animportant question is how this reinforcementoccurs.

Researchers have observed that childrenseem to acquire many language skills appar-ently in the absence of explicit instruction or

explicit reinforcement, and this has beentaken to contradict Skinners analysis ofverbal behavior. Brown and Hanlon (1970)

presented data showing that although parentsoften give feedback on some aspects oftheir childrens speech, other aspects are sel-dom instructed. Specifically, parents tend to

provide positive and negative consequences

for the content of the statements (e.g.,naming objects and actions correctly) whileaccepting many errors in grammar and

pronunciation.Although this argument was weakened by

the work of Moerk (1983), who re-analyzedBrown and Hanlons (1970) data and showedthat their analysis overlooked many subtlecontingencies of reinforcement, Moerk wasnot able to refute Brown and Hanlonsgeneral conclusion. There are several exam-

ples of grammatical distinctions that cannotplausibly be explained by a history of explicitreinforcement. This is simply because most

adults are unaware of them. That is, adultsrespect certain regularities in word orderwithout being able to tact those regularities(Palmer, 1998).

Chomsky (1980) asserted that a special,innate linguistic device accounted for howgrammar and other important linguistic skillswere acquired in the absence of directinstruction. Although Chomskys theorycould explain how children are able to speakgrammatically without explicit reinforce-

ment, and proposed that this type of neuralmodule was established through the processof evolution, the theory suffers from a major

problem: It must address how contingencies ofsurvival can select grammatical distinctions.

This study is based on a thesis submitted by thefirst author to the Graduate Faculty of Akershus

University College in partial fulfillment of therequirements for the degree of master of science.Address correspondence to Svein Eikeseth,

Oslo and Akershus University College, P.O. Box4, St.Olavs plass, N-0130 Oslo, Norway. (e-mail:[email protected]).

The Analysis of Verbal Behavior 2012, 28, 7382

73


2/10

What precisely has been selected and whichevolutionary contingences might have selectedthem (Palmer, 1996).

In Skinners analysis, the apparent missing

source of reinforcement is automatic rein-forcement, arising from the child conformingto the behavior of a verbal model. Anautomatic reinforcer is related to the responsein such a way that it is produced directly

by the response (Catania, 2007; Novak &Pelaez, 2004; Vaughan & Michael, 1982).In other words, reinforcement that is notmediated by the action of another person.

This can be illustrated with an example. Agirl who hears and sees her big sister whistlewill very likely want to whistle as well. Shewill exhale air and try to shape her mouth thesame way as her big sister does. Initially,she will not be able to produce anythingresembling the whistling of her model, butgradually, after repeated attempts, the firstwhistling sounds appear, and the produc-ed sound is likely a powerful reinforcer. Ifthe girl continues to practice whistling,sometimes producing a good imitation andsometimes not, differential and automaticreinforcement will shape her behavior into

skilled whistling.This example illustrates how imitation and

automatic reinforcement might play animportant role in the acquisition of verbal

behavior. It must be emphasized that thistype of imitation is not an echoic verbal

behavior, because the verbal response is notproduced in immediate temporal relationshipto the verbal stimulus (Skinner, 1957, p. 164).Skinner discussed these issues in his bookVerbal Behavior(e.g., Skinner, 1957, pp. 58,

357), including the following:

Automatic reinforcement may shape thespeakers behavior. When, as a listener,a man acquires discriminative responsesto verbal forms, he may reinforcehimself for standard forms and extin-guish deviant behavior. Reinforcingsounds in the childs environment pro-vide for the automatic reinforcement ofvocal forms. . . . The child can thenreinforce himself automatically for theexecution of vocal patterns which are

later to become part of his verbalbehavior. At this stage the child resem-bles a parrot, which is also automaticallyreinforced when its vocal productionsmatch something heard in the environ-ment. A similar effect may lead to a

special manner of speaking or to parti-cular forms of response characteristic ofthe behavior of others. The effect isoften called identification, but we haveno need to appeal to a special process

here. The listener usually finds certainspeakers particularly reinforcing, eitherbecause what is said is reinforcing, orbecause the speakers are reinforcing inother ways. Parents, favorite employers,persons of prestige, and close friends areexamples. Since, for one reason oranother, it is often reinforcing to hearsuch people speak, it is automaticallyreinforcing to speakas they speakwitha particular intonation, mannerism, orfavorite vocabulary. Terms characteris-tic of the adult repertoire are likely to beused by children with special frequency

when first acquired. This is not echoicbehavior, because the borrowed re-sponse is not emitted in the propertemporal relation to the verbal stimulus.The borrowing occurs because of theautomatic self-reinforcement generatedby the speaker as a result of his earlierconditioning as a listener. (Skinner,1957, p. 164)

Palmer (1996) used the term parity todescribe how a speaker who is already acompetent listener can detect when he or she

confirms or deviates from the practice of theverbal community, and hence regulates his orher verbal behavior to match the modeled

behavior. Children usually are competentlisteners before they reach the same level asa speaker, and thus they are able to detecteven slight differences in verbal utterances(Horne & Lowe, 1996). This means thatchildren can tell whether their own utterancesconform or deviate from that of a model. Toachieve parity in ones verbal behavior might

be a strong generalized reinforcer becauseachieving parity is highly adaptive in mostsituations, and not conforming is often

punished by the verbal community (Smith,Michael, & Sundberg, 1996; Sundberg,Michael, Partington, & Sundberg, 1996).

The behavior analytic interpretation oflanguage acquisition is thus based on twotypes of reinforcement. First, verbal behavioris, in part, explicitly reinforced by membersof the verbal community. Second, verbal

behavior is, in part, automatically reinforcedwhen the speaker hears him/herself speakingin accordance with verbal conventions.

To test the parity hypothesis one coulddemonstrate that children can acquire novel

74 Leni stviket al.


3/10


4/10

Inclusion criteria were as follows: (a) chro-nological age between 3 and 6 years, and (b)

no reported history of developmental delays.The participants were recruited from thekindergarten they attended and were, accord-ing to the staff, typically developing. Parentsof all children in the kindergarten classreceived information about the experimentand parents of 12 of the 15 children in theclass gave their written consent for partici-

pation. The experiment was conducted in onesession lasting for approximately 30 minutes.All children with parental consent whoattended the kindergarten at the day of the

experiment participated. The participantsage and sex is shown in Table 1. If a

participant in any way expressed or indicatedthat he or she wanted to quit, the experimentwas terminated and the participant wasreturned to the other children with the toyand stickers earned.

Setting and Materials

The experiment was conducted in a room

at the kindergarten. The room was usedfor music lessons and contained a piano,a blackboard, a table, several chairs, and asofa. Placed on the table were stickers, a

plastic sheet (14 3 19 cm) with 20 dots onwhich the stickers could be placed, an mp3-recording device, and three sets of drawingson cards (10 3 15 cm). The first two sets ofdrawingsthetestset and the trainingsetincluded 20 drawings each (40 pictures intotal). Each set of 20 drawings consisted of

10 pairs. Every pair of drawings depicted twoanimals involved in an activity. In one of thedrawings in the pair, animal X was doingsomething to animal Y, and in the otherdrawing in the pair, animal Y was doing the

same thing to animal X (see Figure 1). Thethird set, the generalization set, consisted of17 pictures depicting animals involved in anactivity. The activities were the same as inthe training and test and training sets, but oneof the animals was substituted with anotherone (also taken from the test or training sets)(see Figure 2). The drawings in all sets werethe same as those used in Wright (2006),

except for the generalization set. For this set,new drawings were made based on thedescriptions from Wright.

Dependent Variable and Reliability

The dependent variable was the verbalbehavior of the children, coded into threecategories; passive voice, active voice, andundefined, as follows:

Passive voice was defined as any of the

following sentences: (a) A sentence in thepassive voice (i.e., X is being done some-thing to by Y, e.g., the elephant is being

pushed by the mouse) in which both subjectand object in the drawing are named correctly;

Table 1Participants Sex and Chronological Age

Participant Sex Age (year, months)

Eskil M 4,6Henny F 4,6Simon M 4,0Amanda F 4,3David M 3,5Frode M 5,5

Figure 1. A pair of pictures from the training set.Picture A would be described in the active voiceas The elephant is pulling the mouse. Inpassive voice A would be described as: Themouse is being pulled by the elephant. Picture Bwould be described in the active voice as Themouse is pulling the elephant and in the passivevoice as The elephant is being pulled by themouse.

76 Leni stviket al.


5/10

(b) a sentence that is structured in the passivevoice, but the speaker reverses the naming ofthe subject and object; (c) a sentence that isstructured in the passive voice, but the sameanimal is named as both subject and object; and(d) a sentence that is structured in the passivevoice, but in which the actor is omitted. If the

participant named the animal as a similar butdifferent species, such as naming a zebra as ahorse or naming a moose as a deer, this was

considered correct. This definition of thepassive voice is identical to the one used byWright et al. (2006).

Active voice was defined as any of thefollowing sentences: (a) A sentence contain-ing the active voice (e.g., the mouse is

pushing the elephant) in which both subjectand object in the drawing is named correctly;(b) a sentence that is structured in the activevoice, but the speaker reverses the naming ofthe subject and object; (c) a sentence that is

structured in the active voice, but the sameanimal is named as both subject and object;and (d) a sentence that is structured in theactive voice, but in which the actor isomitted.

Undefined Verbal Behavior. Any otherverbal behavior or lack thereof was scored asundefined. The two latter categories (Activeand Undefined) were not used by Wright

et al. (2006).Inter-rater agreement was calculated for allparticipants using the records of two inde-pendent observers. Point-by-point agreementwas calculated by dividing the number ofagreements by the sum of agreements anddisagreements and multiplying the outcome

by 100. The mean total inter-rater agreementwas 98.9% (range 98.6% to 100.0%).

Procedure

The experiment started when the partici-pant was seated in the sofa in front of thetable and across from the experimenter. The

participant was given a toy and askedwhether he or she would like to stay and

play a game and earn stickers. If the childpreferred to leave the experiment, he or shewould still get the toy. All participantsstayed, and they were shown the stickersthat could be earned during the experiment.A total of 20 stickers could be earned. Except

during the baseline phase, when no feedbackwas given, stickers and praise were deliveredcontingent upon use of the active voice todescribe the drawings. Praise or stickerswere never given when the participant usedthe modeled passive voice, or when the

participant produced an undefined response.Instead, whenever a participant used the

passive voice or produced an undefinedresponse, the experimenter said in a neutralvoice, Now lets look at the next drawing,

and went on to the next stimulus.Initially, the experimenter told the participant

that he or she would be shown some drawings,and that the participant and the experimenterwould take turns describing them. Throughoutall 6 phases, each participant was shown a totalof 74 pictures. The experimenter always usedthe passive voice to describe the stimuli. Atthe end of the experiment, participants who hadnot received all 20 stickers were given theremaining stickers.

Phase 1: Baseline using the test set. Whenthe experiment began, the participant wasgiven the following instructions: I wouldlike you to tell me about some drawings that Iwill show you. The experimenter showed

Figure 2. A pair of pictures from the generaliza-tion set. These drawings depict the same activityas those in Figure 1. The elephant in drawing Ahas been replaced with a panda bear, and themouse in drawing B has been replaced with asheep.

PARITY AND EXPLICIT REINFORCEMENT 77


6/10


7/10

accordance with the contingency describedabove. Procedural integrity was found to be97%, containing only one error: once a

participant was given praise following useof the passive voice.

RESULTS

One participant in the study left the exper-imental room after completing baseline, andhence was excluded from data analysis. Theremaining six children completed the experi-ment in its entirety. Percentage use of passivevoice across all phases for each participant isshown in Figure 3. As can be seen, no

participant used the passive voice duringbaseline.

Results for Eskil are shown in Figure 3and Table 2. During baseline, Eskil used the

passive voice on 0% of the trials (he used100% active voice). In the subsequenttraining phase, he used passive voice on80% of the trials (20% active voice). In thefirst test phase (Phase 3) the passive voicewas used in 53% of the trials (41% activevoice). In the second training phase (Phase 4)the passive voice was used on 100% of thetrials, and in the next test phase (Phase 5) the

passive voice was used for 59% of the trials(35% active voice), and in the final general-ization phase the participant used the passivevoice in 70% of the trials (30% active voice).The results for Simon and Frode were similar

to that of Eskil (see Figure 3 and Table 2),and hence, the verbal behavior of Eskil,Simon, and Frode appeared to be affected bythe modeling condition.

David, by contrast, failed to use thepassive voice throughout the experiment

(see Figure 3). As shown in Table 2, Davidused the active voice on 40% of the trialsduring baseline, and undefined respondingduring the remaining 60% of the baselinetrials. During the two subsequent training

phases (2 and 4), David used the active voiceon 80% of the trials, and during the finalgeneralization phase he used the active voiceon 100% of the trials. As can be seen inFigure 3 and Table 2 the results of Hennyand Amanda were similar to that of David,

though both Henny and Amanda began usingsome use of the passive voice. Hence, theverbal behavior of David appeared unaffect-ed by the modeling condition, whereas themodeling condition had some effect on theverbal behavior of Henny and Amanda.

DISCUSSION

This study replicated and extended Wright(2006) and Whitehurst et al. (1974) by exam-

ining whether preschool-aged children wouldincrease their use of the passive grammaticalvoice when it was modeled by an adult, ratherthan using the active grammatical construction,which was explicitly reinforced.

Figure 3. Percentage of responses in the passive voice across phases for each participant.



8/10

Results showed that 5 of the 6 participants

began using the passive voice after thisverbal behavior had been modeled. For 3 ofthe children this change was large and

persistent into the final generalization phase.The modeling procedure had a limited effecton the verbal behavior of 2 of the partici-

pants, who continued to use the active voiceon most of the trials though they sometimesalso used the passive voice. Finally, 1

participant never used the passive voice.During baseline he used 60% undefined

responding and 40% active voice, and duringthe final generalization phase he used 100%active voice. Hence, he increased the use ofthe active voice (rather than the modeled

passive voice) over the course of theexperiment.

In the final generalization phase of theexperiment, the participants were asked todescribe drawings in which the previousanimals and actions were presented in newcombinations. The drawings in the general-

ization set were different from the drawing inthe training and test sets as follows: Thetraining and test sets consisted of pairs ofanimal X doing something to animal Y, andthen with the roles reversed. The generaliza-tion set consisted of animals and actions fromthe earlier sets but with a new object animal.The object animals was also taken from thetraining and test sets, but had not beencombined with that agent animal or action.During the final generalization phase, 4 of the

participants generated new sentences in thepassive voice, suggesting that the increase orestablishment of the passive voice was amatter of learning an autoclitic or intraverbalframe, rather than simple imitation.

The main question of the study was to

investigate whether children could learn anovel verbal behavior with the manipulationof antecedents only, not consequences. Theincrease of the use of the passive voice isindicative that this was the case, but islimited by the possible explanation that thechildren already had this grammatical framein their repertoire and that the modeling onlyincreased the strength of that frame. Simply

put, it cannot be determined empirically fromthe present study whether the increase in

the use of the passive voice was a matterof acquisition of novel behavior, or only amatter of strengthening existing verbal be-havior. Either way, the results suggest thatthe verbal behavior of children is affected bythe verbal behavior of a speaker, presumablythrough automatic reinforcement and parity.

A limitation of this study was that it lackeda control group. Hence, maturation andreactivity of the probes may be confoundingvariables that cannot be ruled out empirical-

ly. However, it is unlikely that maturationcould account for the change observed in theparticipants verbal behavior, since the ex-periment was conducted in one session, andsince all participants failed to use the passivevoice during baseline despite varying in agefrom 3 years, 5 months to 5 years, 5 months.Also, it is unlikely that reactivity of testingcould account for the acquisition of the

passive voice since the participants wererewarded for using the active voice. Also,

Whitehurst et al. (1974) included a no-treatment control group in their study, andnone of the participants in the control groupused the passive voice at any point over thecourse of the experiment. This observed

Table 2Participants Use of Active Voice and Undefined Verbal Behavior Across Phases as

Percentage of Trials

ParticipantsActive voice/Undefined verbal behavior

Baseline Modeling Test Modeling Test Generalization

Eskil 100/0 20/0 41/6 0/0 35/6 30/0Henny 80/20 80/0 65/0 70/0 76/0 80/0Simon 100/0 60/0 82/0 20/0 18/0 10/0Amanda 80/20 100/0 94/0 80/0 6/0 100/0David 40/60 80/20 35/65 80/20 65/35 100/0Frode 80/20 30/10 82/6 10/0 65/12 40/0

80 Leni stviket al.


9/10

stability of the active voice occasioned thedesign of the present study. Alternatively,a single-case design, such as a multiple

baseline design, could have been used.

Interestingly, the 3 participants for whomthe modeling condition had the largest effectwere on average 6 months older than theother 3 children. It is possible that for thesechildren, the automatic reinforcement pro-duced by achieving parity had higher valencethan the praise and stickers given contingenton use of the active voice. For the youngerchildren, the opposite could have been thecase. However, it should be noted that noreinforcer assessment was conducted on the

stickers or the praise. It is possible, thoughunlikely, that neither stickers nor praisefunctioned as reinforcers for some of the

participating children. Another possibilityis that delivery of praise and reinforcerscontingent on turn-taking lead to satiation forthese stimuli as reinforcers.

Future studies could investigate the extentto which participants would use the passivevoice, after modeling, on novel drawings inwhich neither subjects nor activities had been

previously modeled. Future studies couldalso model other grammatical frames thanthe passive voice. This could be frames thatdo not exist or are extremely unusual, such asMurdered by Brutus, Caesar or BrutusCaesar murdered. Another possibility is tomodel an artificial pronunciation of a famil-iar word. If such artificial verbal responsescan be established in the verbal behavior ofchildren, this would suggest that certainaspects of language indeed can be learnedthrough automatic reinforcement and parity.

This study highlights the need to considerboth explicit social reinforcement as well asautomatic reinforcement when studying theacquisition of verbal behavior.

REFERENCES

Allen, S. E., & Crago, M. B. (1996). Earlypassive acquisition in Inuktitut.Journal ofChild Language, 23, 129155.

Berman, R. A. (1985). The acquisition ofHebrew. In D. Slobin (Ed.), The crosslin-guistic study of language acquisition(Vol. 1, pp. 255371). Hillsdale, NJ:Erlbaum.

Brown, R., & Hanlon, C. (1970). Derivation-al complexity and order of acquisitionin child speech. In J. R. Hayes (Ed.),Cognition and the development of lan-

guage(pp. 1153). New York: Wiley.Catania, A. C. (2007). Learning (4th ed.).New York: Sloan Publishing.

Chomsky, N. (1980). Rules and representa-tions. New York: Columbia University Press.

Demuth, K. (1989). Maturation and theacquisition of the Sesotho passive. Jour-nal of the Linguistic Society of America,65, 5680.

Horne P. J., & Lowe, C. F. (1996). Namingand other symbolic behavior. Journal of

the Experimental Analysis of Behavior,65, 185241.Marchman, V. A., Bates, E., Burkardt, A., &

Good, A. B. (1991). Functional con-straints of the acquisition of the passive:Toward a model of the competence to

perform. First Language, 11, 6592.Moerk, E. L. (1983). A behavior analysis

of controversial topics in first languageacquisition: Reinforcement, corrections,modeling, input frequencies, and the

three-term contingency. Journal of Psy-cholinguistic Research, 12, 129155.Novak, G., & Pelaez, M. (2004). Child and

adolescent development: A behavioralsystems approach. Thousand Oaks, CA:Sage Publications.

Palmer, D. C. (1996). Achieving parity: Therole of automatic reinforcement. Journalof the Experimental Analysis of Behavior,65, 183353.

Palmer, D. C. (1998). The speaker as listener:

The interpretation of structural regularitiesin verbal behavior. The Analysis of VerbalBehavior, 15, 316.

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

Smith, R., Michael, J., & Sundberg, C. A.(1996). Automatic reinforcement automat-ic punishment in infant vocal behavior.The

Analysis of Verbal Behavior, 13, 3948.Sundberg, M. L., Michael, J., Partington, J. W.,

& Sundberg, C. A. (1996). The role of

automatic reinforcement in early languageacquisition. The Analysis of Verbal Behav-ior,13, 2137.

Trosborg, A. (1982). Reversibility and theacquisition of complex syntactic structures



10/10

Documents

Grammatical Constructions in Typical Developing Children- Effects of Explicit Reinforcement, Automatic Reinforcement and Parity