JOURNAL OF APPLIED BEHAVIOR ANALYSIS

SYSTEMATIC REINFORCEMENT: ACADEMICPERFORMANCE OF UNDERACHIEVING STUDENTS'

BRUCE A. CHADWICK AND ROBERT C. DAY

WASHINGTON STATE UNIVERSITY

The effect of contingent tangible and social reinforcement on academic performance wasinvestigated in an experimental classroom of 25 selected underachieving students. Measureswere taken of both teacher and child behavior during a baseline and two experimentaltreatment periods. During Treatlnent I, a point system with tangible backup reinforcerswas combined with contingent social reinforcers dispensed by the teaching staff to assessthe effects on three measures of academic performance (i.e., per cent of time at work, workoutput per minute, and accuracy). During Treatment IL, the contingencies for the tangiblereinforcers were terminated while social reinforcement was continued to see if the positiveeffects of Treatment I on academic performance would persist. The results show that withcombined tangible and social reinforcers, students' work time, rate of output per hour, andaccuracy in all activities substantially increased. After termination of the tangible rein-forcers, the students maintained their high rates of ouput per hour and accuracy for theremaining period of the study while the total amount of time at work returned to thebaseline level.

A review of more than 30 previous studies,using some form of systematic reinforcementon two or more relatively "normal" (i.e., notemotionally disturbed) students in a fairlylarge classroom, reveals that only seven havedealt in some way with the task of elevatingacademic performance. Three studies usedvarious types of social reinforcement such asteacher attention, approval, free-play time,and field trips to increase the per cent of timespent in appropriate study behavior. Wasik,Senn, Welch, and Cooper (1969), using teacherattention and approval, substantially raisedthe per cent of appropriate classroom re-sponses of two disruptive, second-grade, mi-nority group girls from disadvantaged circum-stances in a clatss of 20 students. Bushell,Wrobal, and Michaelis (1968) and Hall,

'The authors wish to thank Mrs. Aurora Marichalar,Mr. Roger Candanosa, and Mr. Eugene Robles for theirassistance in gaining the trust and cooperation of theMexican-American migrant community; Mr. John Mer-rick and Mr. Brent Roper for their assistance in direct-ing the project; and Father Joseph Skehan for the useof St. Joseph's Catholic School. This research was sup-ported in part by the National Institute of MentalHealth grant MH14497-01, in part by Project 1934,Agricultural Experiment Station, Washington StateUniversity and is scientific paper number 3624. Re-prints may be obtained from Bruce A. Chadwick orRobert C. Day, Department of Sociology, WashingtonState University, Pullman, Washington, 99163.

Panyan, Rabon, and Broden (1968) producedsubstantial increases in study time with largergroups of 12 and 30 elementary students, re-spectively.Four studies dealt with very aggressive, dis-

ruptive, or withdrawn students of varied agesand backgrounds and used token systems andtangible backup reinforcers to achieve an ef-fect on rates of academic responses such asreading, arithmetic, etc. For example, Wolf,Giles, and Hall (1968) compared 15 sixthgrade underachievers (i.e., averaged 2 yr belowtheir norm for reading) living under urbanpoverty conditions, to a matched controlgroup, on academic gains in reading, grades,and achievement test scores. The experimentalgroup made significant gains in report cardgrades during the year (p < 0.005) and inStanford Achievement Test scores (p < 0.01),as well as in their reading, math, etc. Martin,Burksholder, Rosenthal, Tharp, and Thorne(1968) tested the differential effects of variedreinforcement systems with nine highly dis-ruptive, underachieving adolescents rangingin age from 13 to 18 yr and substantially in-creased their per cent of time actually at work.One criticism of the studies cited above is

that they have dealt with and reported onglobal percentages of time spent in appropri-ate behavior, grades, and achievement level,but in the process they have neglected to get

311

1971, 4, 311-319 NUMBER 4 (WINTER 1971)

BRUCE A. CHADWICK and ROBERT C. DAY

closer to more basic questions of academic per-formance such as: (1) Do reinforcement proce-dures raise the amount of students' time spentworking, without accelerating efficiency (i.e.,problems solved per hour) or accuracy (i.e.,per cent correct answers), or do they raise allthree in varying degrees? With these questionsin mind, the specific objective of the presentstudy was to test the effects of systematic rein-forcement on the following measures of aca-demic performance:

(1) Total work time (per cent of assignedclassroom time student engaged in aca-demic activities).

(2) Efficiency of academic activities (numberof problems solved or units of work com-pleted per unit of time).

(3) Accuracy of academic work (per cent ofcorrect answers).

METHODS

SubjectsElementary school teachers from three school

districts were asked during the last weeks ofthe regular school year to identify children intheir classes who were clearly underachievingfor various reasons and thus would benefitfrom a summer education program. From thelist of nominations, 30 minority children(blacks and Mexican-Americans) between theages of 8 and 12 yr were selected. Because ofmigration of their family to other towns, fiveMexican-American children dropped out atvarious stages of the study, leaving 25 subjectson which complete data were collected. Theracial composition of the group included 11blacks and 14 Mexican-Americans. The aver-age age of the subjects was 10.1 yr and in-cluded 19 males and 11 females. The educa-tional achievement of the parents was quitelow: the average education of mothers was 7.0yr and that of fathers 7.2 yr. The average fam-ily income was $3800 per year. Surprisingly,40%0 of the children came from homes brokenby separation, divorce, or death, which mayhave contributed to the group's educationalproblems.The average achievement level of the group

on the California Achievement Test fell 1.5 yrbelow the norm for the appropriate age. Theoverall mean grade point average for the

preceding year was 1.47 (D), which also indi-cates significant underachievement.

The SettingThe experimental class was conducted in a

modern, bright, air-conditioned school build-ing. Observers watched from a large booth ad-joining the classroom through a one-way mir-ror. Several microphones were located on theceiling of the classroom so that observers couldhear the verbal responses of the students andteachers, including whispered phrases. Classeswere conducted from 9 a.m. to 3 p.m., five daysa week, and included two 20-min recesses and1 hr for lunch. To insure regular attendancein school, transportation was provided forstudents.

Experimental ConditionsThe study lasted for 11 weeks, during the

summer, which were divided into three peri-ods: a three-week Baseline; a six-week experi-mental period labelled Treatment I, and fi-nally, a second two-week experimental period,labelled Treatment II.During the initial three-week Baseline (15

school days) the classroom was conducted in aconventional manner by a certified femaleteacher who directed the class and two maleteaching aides. One of the aides was a certifiedteacher with several years experience with mi-nority students and the other was a graduatestudent. The teacher and aides were instructedto respond to the children with commands,praise, disapproval, threats, removal, and pad-dling, as they saw fit, using the "normal class-room" as a model. The children receivedlunch, candy, toys, etc. from the school "store",without contingencies, at a rate comparable tothat planned for the experimental periods.The students' rates of response (work, effi-ciency, and accuracy) to various types ofacademic tasks were recorded. The rates ofapproving, disapproving, and instructional re-sponses for the teacher and aides were alsocontinuously recorded.

Treatment IThe reinforcement program, involving a

combination of social reinforcers from theteacher and aides and tangible backup rein-forcers for points earned in classwork was putinto effect during the six weeks of TreatmentI. A point system, nearly identical to that used

312

SYSTEMATIC REINFORCEMENT

by Wolf, Giles, and Hall (1968), was chosen ascurrency to "pay" students for work done in-stead of tokens due to ease of handling andcontrol over losses, theft, trading, etc. Eachchild was given a booklet with his name on it,containing three sets of pages of differing col-ors, each page containing 50 squares. When achild earned points, the teacher or other staffchecked the appropriate number of squares inthe booklet. Two types of behavior were rein-forced with points, academic work (i.e., read-ing, arithmetic workbooks, etc.) and appropri-ate personal and social classroom behaviorrelated to academic performance (i.e., remain-ing seated, raising hand for teacher's attention,working alone without talking or strikingneighbors). When an assignment was com-pleted it was collected, graded by the staff, andpoints were assigned on a schedule using aweighted formula emphasizing accuracy ofwork (weight = 67%) and speed (weight =33%). This procedure permitted relativelyrapid feedback to the students concerningtheir academic performance.

Points were passed out contingently for thefollowing types of personal and social behav-iors as they occurred in the classroom: (1) 20points for getting down to work quickly with-out disturbing others; (2) 20 points for work-ing continuously for 10 to 20 min; (3) 10points for working while a fight or loud noiseswere occurring nearby; (4) five points for rais-ing the hand for attention; (5) five points forobtaining permission before leaving one'sdesk. With minor misbehaviors, such as talk-ing to neighbors or calling loudly for theteacher, the child was first redirected into ap-propriate responses, given a warning of theconsequences if inappropriate conduct per-sisted and finally if misbehavior continued, 10to 50 points, depending on the seriousness ofthe behavior, were withdrawn from his book.

In a few instances when a child becamearoused over a fight or verbal insults and con-tinued to disrupt the class even after two setsof points were extracted from his book, he wasremoved from the situation without dramaticsand placed in social isolation in the "timeoutroom". Release from this room could beachieved only by completing a math assign-ment with no point given. Plans had beenmade to take students home if they refused tofinish the assignment in the timeout room butthis contingency was never required.

To redeem their points the students wereoffered the following backup reinforcers:

(A) Green points: lunch. The first 25 pointsearned each morning were placed on agreen page to pay for lunch. During the11-week study none of the students everfailed to earn lunch. After lunch wasearned, the students were able to choosewhich of the following two categories tohave their points recorded in.

(B) Yellow points: school store items. Theseitems from the store included candy,gum, toys, school supplies, goldfish, afew clothes, jewelry, makeup kits, base-balls, games of all kinds, etc.

(C) Red points: field trips. These weeklytrips occurred on Friday afternoon andincluded fishing, boating, swimmingparties, a trip to a farm, bank, hospital,and pet shop.

At the conclusion of the school day, the stu-dents' names were randomly drawn from abox and, in this order, went to the school storeto redeem their points with tangible reinforc-ers. In order to maximize the pairing of teacherapproval with material reinforcers, the teacherwas assigned to the store and as she exchangedtoys for points she expressed her approval tothe child and urged continued academic per-formance.

Finally, the teacher and aides systematicallyapplied social reinforcers to the students in theform of warm praise and support for study be-havior and for the appropriate social behaviorfor which points were earned and/or disap-proval for misbehavior. Staff social reinforcerswere counted by an observer using the catego-ries described later. The three components ofthe dependent variable, academic perform-ance, were measured as during Treatment I,and the resultant rates were compared to thoseobtained during the Baseline to assess the ef-fects of the experimental treatment.

Treatment IITo assess the potential long-range effects of

the reinforcement system on rates of academicbehavior, the point system was terminated dur-ing the last two weeks while the rate of teacher-mediated social reinforcement was continued.The children received candy, toys, lunch, etc.without contingencies as in the Baseline, whilethe systematic use of social reinforcement ap-

313

BRUCE A. CHADWICK and ROBERT C. DAY

plied during Treatment I, was continued un-

changed. This procedure was instituted to findout whether social reinforcement alone couldmaintain the academic performance of prob-lem children who had been treated with sixweeks of contingent tangible reinforcers, per-

sonally distributed by the teacher.

Measurement of Variables1. Per cent of time at work. Two electric

clocks, controlled by switches mounted on a

panel, were assigned to each student. Whenthe teacher gave a particular student an assign-ment, his "total-time" clock was turned on

and remained on until the student completedthe assignment or the teacher asked for it.When the student was actually engaged inworking on the task, his "work-time" clockwas turned on. Any interruptions in work suchas talking to peers, fighting, running aroundthe room, sleeping, etc., resulted in the workclock being turned off while the total-time re-

mained in motion. The most difficult timingdecision was when the student stopped work-ing to stare out of the window as if he were

thinking about the task. A 1-min period was

allowed for the child to return to the task be-fore the clock was stopped. If it became obvi-ous during this 1-min period that the studentwas responding to outside stimuli, and notworking on the task, then the work clock was

turned off. Five observers were assigned to theclocking task so that each had five or six chil-dren to watch. Observers were trained and pe-

riodically (one day each week) checked for re-

liability. The level of reliability varied from74 percent to 97 percent with the averagingbeing 86 percent. These timing proceduresallowed the computation of the percent oftime that the student actually worked on any

given assignment (work time/total time).2. Rate of work output. The second measure

of academic performance used was the rate ofwork output per unit of time. The numbers ofvarious types of exercises and workbook prob-lems completed per session were counted andthe total divided by the number of minutes re-

quired. The goal here was to find out whetheracademic underachievers not only would worklonger under contingent reinforcement, butwould actually complete more reading, spell-ing, arithmetic, etc. exercises per unit of timeas well. The number of units of each activitywere recorded for each student for each day.

The nature of the unit varied considerablyfrom simple math problems mimeographed bythe teacher or contained in the math work-book, to reading a paragraph and answeringquestions about the reading, to the number ofwords written during a creative writing assign-ment. The crucial point is that the unit foreach activity (i.e., math workbook, readingcomprehension, etc.) remained constant dur-ing the entire summer program. The difficultyof the exercises were gradually increased, overtime, to control for increased performance dueto the practice effect and maximize the learn-ing progress of each student. The total num-ber of problems completed on each activitywas recorded and was then divided by thenumber of minutes the student worked on theactivity that day.

3. Accuracy of work output. The third mea-sure of academic performance employed wasthe accuracy of work completed (i.e., the percent of problems attempted that were donecorrectly). It became obvious that an increasedrate of work done under the reinforcementprogram would be meaningless if most ofthe additional output was inaccuratelyexecuted.

4. The California Achievement Test. Thestudents were administered the appropriateversion of the California Achievement Test(CAT) during the first week of school. An al-ternative form of the same test was adminis-tered during the last week of school to assesswhat differences occurred during the summerprogram.

5. Teaching staff behavior. One observer wasassigned the task of categorizing and record-ing the behavior of the teacher and aides,(using a modified version of the system re-ported by Thomas, Becker, and Armstrong,1968), into three general classes: (1) Approvingbehavior; (2) Disapproving behavior; and (3)Instructional behavior. Approving and disap-proving behaviors were recorded only whenthey followed discriminable child behaviorsfalling into appropriate or disruptive catego-ries. Approving behaviors were directed at in-dividuals, not the group as a whole, and in-cluded verbal support, praise and affectionmixed with physical contact (i.e., pats on theback, etc.) and smiles. Disapproving behaviorssimilarly incorporated negative remarks, scold-ing, sarcasm, and threats to individuals of lossof privileges, points, etc. Negative, disapprov-

314

SYSTEMA TIC REINFORCEMENT

ing physical contacts were counted when achild was forcefully guided or hauled back tohis chair, out of a fight, or out of the room.Disapproving facial gestures were not re-corded. Finally, three forms of instructionalbehavior were recorded: requests, commands,and rules stated to individuals, small groups,or the whole class. Simultaneous measures ofteaching staff behavior were made by a secondobserver one day each week throughout thestudy. Reliability-computed by dividing thesmaller count by the larger count for eachcategory at the end of the observation-aver-aged .92 for all categories combined andranged between .83 and .99 on individual ob-servations of each behavior.

RESULTS AND DISCUSSIONS

A cademic PerformanceThe first indicator of the overall effect of the

reinforcement program on academic perform-ance is the per cent of total available time thatstudents actually worked (i.e., "per cent ofwork time"), and is portrayed graphically inFigure 1. During the Baseline, the averageamount of total time spent working was 39%.As. shown in Figure 1, the reinforcement pro-gram used during Treatment I produced animmediate upward shift in academic activitystarting with the first day, resulting in a 57%/Oaverage.Terminating the point system at the begin-

ning of Treatment II produced a drop in aca-demic activity to 42%, which approaches theoriginal Baseline rate.An examination of Figure 1 reveals a trend

of decreasing work time during the Baselineand also during Treatment I. Trend analysiswas performed to ascertain if these two trendswere significant.2 The Baseline period pro-duced a highly significant (F = 89.3; p < 0.001)downward trend, which indicates that the rateof work time had not reached a stable meanlevel. This finding suggests that if the "normalclassroom" procedures had been continuedlonger, the level of work time would havedropped significantly lower than the 39% ob-tained during the three-week Baseline. Treat-ment I also evidenced a significant (F = 20.2;

2For a discussion of the trend analysis techniquesused in the paper, see: Lewis, 1960, pp. 334-420; Hays,1963, pp. 539-566; and Bruning and Kintz, 1968, pp.123-149.

p < 0.001) decreasing trend, although it wasnot as rapid as that in the Baseline. A visualexamination of Figure 1 for Treatment I sug-gests that about the middle of the period, thework time leveled off. To test this observation,trend analysis was performed on the final 15days (15 through 29) of Treatment I; the re-sults (F = 0.06) are not significant. This find-ing indicates that for the last three weeks ofTreatment I, work time varied around a con-sistent mean level. Finally, the data in Treat-ment II do not evidence a significant trend ofany kind. The results in Figure 1 and Table 1thus provide strong support for the positivehypothesized effect of the combined reinforce-ment system on the amount of student worktime but they also indicate that a six-week re-inforcement period was not sufficient to obtaina continual effect after the tangible reinforcerswere terminated.The second measure of academic perform-

ance, rate of work output (i.e., the total num-ber of problems completed divided by thenumber of minutes at work), was included inthe study to find out whether the reinforce-ment system could raise not only the length oftime worked but also the actual number ofproblems and exercises completed per unitof time. The results in Figure 1 and Table 1indicate that for all seven types of academic ac-tivity, the rate of problem completion in-creased significantly during Treatment I. Therange of percentage increase in output, com-paring Treatment I to the Baseline, variesfrom 83% (math worksheets) to 410% (SRAReading Laboratory exercises). It is clear thatthe sample of underachieving students acceler-ated output to an impressive degree duringexposure to the tangible and social reinforce-ment system.There appears in Figure 1 (middle), a trend

of increasing output during Treatment I whichis significant (F = 77.8; p < 0.001). This meansthat during the reinforcement program, thelevel of student efficiency in the seven aca-demic subjects was significantly increasing.The data for Treatment II did not produceany significant trends which supports thecontention that after the termination of thetangible reinforcers, efficiency continued at alevel comparable to that obtained at the endof Treatment I, although the increasing trendin Treatment I did not continue in Treat-ment II.

315

BRUCE A. CHADWICK and ROBERT C. DAY

X- 57%

I . a.II15 I

I I a.. . I .... I .. I

29 115TREATMENT I

8

Z

6

M4w

-J

co0 2

0r0

100I--ULU 80

0 F-U Z 60w

V) 0r-

Uj W 40_ CL

° 20l-

0

15 1 15 29 1 9

BASELINE TREATMENT I TREATMENT

1 15 1 15 29 1 9

BASELINE TREATMENT I TREATMENT3

OBSERVATION DAYS

Fig. 1. (Top) Per cent of time at work: the number of minutes spent actually working on assignments dividedby the total time assigned to the task. (Middle) Rate of work output: the number of exercises and problems donedivided by the number of minutes. (Bottom) Accuracy of output: the number of problems and exercises done cor-

rectly divided by the total attempted.

316

100

80

Z 60

0~ 40

20

I--

LU

I~-

LI0 I II

BASELINE

x- 42%

I . .a I

9TREATMENT X

X a 39%

SYSTEMATIC REINFORCEMENT

Table 1

Rate of Work Output in Seven Academic Subjects: Comparison of Means, SignificanceTests, and Explained Variation for the Baseline and Treatment I Periods

Baseline versus Treatment I

Mean ProblemsAttempted/Min.

Treatment PercentAcademic Activity Baseline I Increase t P

SRA ReadingWorkbook 0.97 2.90 189 4.24 0.001 0.25

SRA Spelling Lab. 0.90 2.34 160 6.20 0.001 0.42SRA Reading Lab. 0.28 1.43 410 8.30 0.001 0.56Merrill Workbook 1.37 2.70 97 5.25 0.001 0.34Spelling Workbook 0.80 1.72 115 5.65 0.001 0.37Math Workbook 2.01 6.19 208 6.15 0.001 0.41Math Worksheets 2.65 4.84 83 6.94 0.001 0.49

The third measure of academic performance over Baseline rates on all seven types of aca-was accuracy of problem solving. The shifts in demic activity.accuracy during the experimental periods for Comparing the effects of the experimentaleach academic subject are reported in Table condition minus the tangible reinforcement2 and illustrated in Figure 1 (bottom). The (Treatment II) with the total reinforcementSRA Spelling Laboratory is omitted from this program (Treatment I) shows that the stu-analysis because the students corrected their dents maintained their high level of accuracyown work on this task and cheating and inac- in all of the activities except the SRA Readingcuracies were quite possible. Performance on Workbook. The accuracy levels varied slightlyspelling tests was collected as a substitute. In between the two Treatment periods, but withthese spelling tests, the teacher pronounced the exception of SRA Reading Workbook, allthe word to be spelled and used it in a sen- of the differences were nonsignificant.tence. The children then spelled the word. Trend analysis of the data concerning accu-Since the rate of words spelled was controlled racy of work for the three experimental condi-by the teacher, this activity was not analyzed tions revealed only one significant trend. Theor reported in Table 2. Baseline evidenced a significant downwardAs manifested in Table 2, the combined re- trend (F = 45.9; p < 0.001) in level of accu-

inforcement program in Treatment I was cor- racy. Again, it is not possible to ascertain howrelated with an increase in student accuracy much lower the average level of accuracy

Table 2

Accuracy of Work in Seven Academic Subjects: Comparisons of Means, Significance Tests,and Explained Variation for the Baseline and Treatment I Periods.

Baseline versus Treatment I Period

Mean Percent:No. Correct/Total

Treatment PercentAcademic Activity Baseline I Increase t P

SRA ReadingWorkbook 32.4 62.2 92 4.90 0.001 0.29

SRA Reading Lab. 30.9 64.9 110 7.38 0.001 0.51Merrill Workbook 53.3 68.2 28 3.65 0.001 0.18Spelling Workbook 52.0 68.0 31 4.86 0.001 0.30Math Workbook 51.6 72.9 41 4.91 0.001 0.30Math Worksheets 58.7 75.1 28 4.88 0.001 0.29Spelling Tests 46.7 71.9 54 6.12 0.001 0.39

317

BRUCE A. CHADWICK and ROBERT C. DAY

would have dropped if the Baseline periodwould have been longer, but it is safe to as-sume that it would have been lower than the50% recorded for the three-week Baseline pe-riod. This finding provides additional supportfor the significance of the improvement in ac-curacy observed during the reinforcementprogram.The results concerning the three indicators

of performance, work time, rate of output andaccuracy, together provide support for the pre-dicted effects of contingent reinforcement onthe academic performance of underachievingminority children. They worked longer, faster,and more accurately during the tangible andsocial reinforcement system designed for thisstudy.The final indicator of academic performance

utilized California Achievement Test scores.The pre-experimental average grade place-ment score was 3.60 years which increased dur-ing the program to 4.02. All but one childimproved their performance and the averageimprovement was .42 years in grade placementduring the 11 weeks of this summer program.

Teacher BehaviorThe behavior of the teacher and aides as

they attempted to control and direct the class-room were carefully recorded. During Baseline,the staff averaged 13 approving, supportiveresponses per hour; 27 disapproving responsesper hour; and 55 instructional responses perhour. The averages shifted as predicted duringTreatment I as approvals increased to 47 perhour, disapprovals dropped drastically toseven per hour, and instructions (requests,commands, etc.) dropped to 17 per hour.When the point contingencies were termi-

nated at the end of Treatment I, the averagenumber of approving responses dropped from47 per hour in Treatment I to 27 per hourduring Treatment II. Social control was main-tained during the Treatment II period withonly a slight increase in the rate of disapprov-ing responses (i.e., a shift from seven per hourto nine per hour) while instructional remarksremained essentially unchanged (16 per hour).

Several significant findings emerge from thisstudy. The first is that a limited teaching staffwas able to raise substantially the level of aca-demic performance of an entire class (25 stu-dents) of seriously underachieving students.One important consequence of this increase in

academic behavior was a significant improve-ment in academic skills and curriculum mate-rial learned as indicated by achievement testscores. The results suggest that if these under-achieving students had been able to partici-pate in the experimental program for a yearthey could have advanced academic skills andachievement up to a level comparable to theirpeers in the regular classroom.Another significant finding was the shift in

teacher behavior from social control responsesto supportive and instructional responses dur-ing the reinforcement program. In the experi-mental period, the teacher and aides were able,for the first time, to give instructions for newtasks, explain intellectual concepts and ideas,and present curriculum material without hav-ing to shout over general noise and withoutnumerous interruptions. The students' changein behavior allowed the staff to shift from con-trol over misbehavior to teaching functions.In fact, during the experimental period theteacher and two aides were hard pressed to re-spond to the flood of student requests for as-sistance with long division problems, wordmeanings, spelling, etc.A final finding to note is the "carry-over" ef-

fect or continuation of improved academicperformance after the tangible reinforcerswere terminated. Even though the total worktime dropped in Treatment II to a level com-parable to the Baseline, accuracy and efficiencywere maintained at a level equal to TreatmentI, after the tangible reinforcement was termi-nated. A number of explanations for these re-sults may be suggested, and further studiesutilizing more complete research designs willbe needed to test them. First, perhaps the sixweeks of reasonable pleasant interaction withthe teaching staff and the reinforcers associ-ated with their approval during Treatment Ielevated the reinforcing quality of teacherpraise, which in turn was able to maintain per-formance at a relatively high level. Also, theexperience of operating successfully at a highperformance level, may have resulted in thework becoming self-reinforcing and thus self-maintaining at a high level. Finally, perhapsimproved academic skills alone, developedduring the first nine weeks had a carry-over ef-fect during Treatment II. Whether one or allthree of these effects are operating, it is clearthat the efficiency and accuracy of work wasmaintained under the program of social rein-

318

SYSTEMATIC REINFORCEMENT 319

forcers in Treatment II. It is anticipated thatsuch improvement will assist the student toachieve at a higher level when returned to aregular classroom.A final comment is in order regarding the

practical implications of the findings describedabove for school districts with a significantnumber of underachieving students. It appearsto be possible, and perhaps more effective andless costly than present arrangements, to formseparate classrooms of problem children wherehigh-interest curricula and valued tangible re-inforcers are used to improve student aca-demic performance and skill levels so that theyeventually can succeed adequately in the regu-lar classroom with its more symbolic and socialreinforcement systems.

REFERENCESBruning, J. and Kintz, B. L. Computational handbook

of statistics, Glenview: Scott, Foresman and Co.,1968. Pp. 12-15.

Bushell, D., Jr., Wrobal, Patricia A., and Michaelis,Mary L. Applying "group" contingencies to theclassroom study behavior of preschool children.Journal of Applied Behavior Analysis, 1968, 1, 55-62.

Clark, Marilyn; Lachawicz, J., and Wolf, M. A pilotbasic education program for school dropouts incor-rating a token reinforcement system. Behavior Re-search and Therapy, 1968, 6, 183-188.

Clement, P. W. and Milne, D. Group play therapyand tangible reinforcers used to modify the behav-

ior of 8-yr.-old boys. Behavior Research and Ther-apy, 1967, 5, 301-312.

Hall, R. V., Panyan, Marion; Rabon, Deloris, and Bor-den, Marcia. Instructing beginning teachers in re-inforcement procedures which improve classroomcontrol. Journal of Applied Behavior Analysis, 1968,1, 315-322.

Hays, W. L. Statistics, New York: Holt, Rinehart andWinston, 1963.

Lewis, Don. Quantitative methods in psychology, NewYork: McGraw-Hill Company, Inc., 1960.

Lovitt, T. C. and Curtiss, Karen A. Academic responserate as a function of teacher- and self-imposed con-tingencies. Journal of Applied Behavior Analysis,1969, 2, 49-53.

Martin, Marian; Burksholder, Rachel; Rosenthal, T. L.,Tharp, G., and Thorne, L. Programming behaviorchange and reintegration into school milieux of ex-treme adolescent deviates. Behavior Research andTherapy, 1968, 6, 371-383.

Thomas, D. R., Becker, W. C., and Armstrong, M.Production and elimination of disruptive classroombehavior by systematically varying teacher's behav-ior. Journal of Applied Behavior Analysis, 1968, 1,35-45.

Wasik, B. H., Senn, K., Welch, R. H., and Cooper, B. R.Behavior modification with culturally deprivedschool children: two case studies. Journal of Ap-plied Behavioral Analysis, 1969, 2, 181-194.

Wolf, M. M., Giles, D. K., and Hall, R. V. Experi-ments with token reinforcement in a remedial class-room. Behavior Research and Therapy, 1968, 6, 51-64.

Received 20 August 1970.(Revised 18 January 1971.)