Group Title: effects of visual feedback from a chart upon the rate of academic performance of junior and senior high school students
Title: The Effects of visual feedback from a chart upon the rate of academic performance of junior and senior high school students
Full Citation
Permanent Link:
 Material Information
Title: The Effects of visual feedback from a chart upon the rate of academic performance of junior and senior high school students
Physical Description: x, 102 leaves. ; 28 cm.
Language: English
Creator: Quigley, Patrick Andrew, 1944-
Publication Date: 1972
Copyright Date: 1972
Subject: Academic achievement   ( lcsh )
Self-control   ( lcsh )
Foundations of Education thesis Ph. D   ( lcsh )
Dissertations, Academic -- Foundations of Education -- UF   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Thesis: Thesis--University of Florida.
Bibliography: Bibliography: leaves 98-101.
Additional Physical Form: Also available on World Wide Web
General Note: Typescript.
General Note: Vita.
Statement of Responsibility: by Patrick A. Quigley.
 Record Information
Bibliographic ID: UF00097635
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000585137
oclc - 14191653
notis - ADB3769


This item has the following downloads:

PDF ( 3 MBs ) ( PDF )

Full Text







To Ch airlotte


The task of sorting out the contribution of various

environmental changes with relationship to various be-

havioral changes is often a difficult task when precise

data are not kept. When the target behavior is professional

scholarship which has been shaped over a four-year period,

then accounting for the contribution of each environmental

cl:ancje becomes eve.n r.ore obfusciated. There are a few con-

tributors in the Ccse of this dissertation w\ho, however,

have had an e-ffect bey'ozd all others:

First, to Dr. John Newell, -who was the first to teach

r.e Icarning theory and then to reinforce my beliefs in this

aLCa over I t'i.cC-year per.od.

To Dr. 1Henrr. ennypacker who opened new vistas in my

traiixing as a .ciea .ingi theorist through precision teaching.

To Dr. Criile Jel:ter, who both as my advisor for the

master:'. dre'' and later a.s a committee member for the

do:!-orate IelLpcd me re alize the importance of psychology

a::- the !o.Lhr.r ncie for educational psychology.

To Dr. U Wolking, who hll'ped me realize the

.itpplica.ilit y of :Le L eC:perioiental. inalys.i.s of animal. behavior

in the area of sLirulus control t.o school learning. My

debt to him has not been paid in this dissertation and must

wait future payment.

To Dr. William Ware, who taught me a good deal of all

that I know in the areas of statistics and research design.

His presence on the committee helped me realize the positive

relationships between standard research methodology and the

experimental analysis of behavior.

To Joseph and Mary Otchin, I owe a wealth of thanks

for their support over the past few years.

Many thanks are due Pat Craig, Linda Howell, Pat

Cohles, and Margaret Tomlinson for initiating this re':;earch

and using Lheir classrooms to further our understanding of

junior and senior high school students.

Finally, and most important, to Charlotte, my wife,

and to Leigh, Catin, and Shannon, my children, who, as the

most significant features of my immediate environment,

provided most the incentive for finishing this work.




LIST OF TABLES. . . . . .



ABSTRACT . . . . . .




I [I




. . iii

. . vi

. . vii

. . viii

INTRODUCTION . . . . . . . .



CA ALYS.S . . . . . . . . .

CONCLUSIONS . . . . . . . .



IEFERENCES. . . . . . . . . . .

BIO;RP.APHICAL SKETCH . . . . . . . .





Table Page


2 DESIGN FOR ANALYSIS . . . . . 38

FOR ERROR . . . . . . . . 48




BRAC KET' ;. . . . . . ..... . . .55

LOGS C' L' kROR RATE r'11 TYPING . . . 57

MNE-.NS IN i:2.ACKNFTS . . . . ... . 69

BRACXKET'LS. . . . . . . .. . 70



TEMPORAL ORDER . . . . . .



FIRST SUBJECT . . . . . . . . 67

SECOND SUBJECT. . . . . . . . 68

STUDENT . . . . . . . . . . 72






RCS Rate Computation Sheet

SBC Standard Bchavior Chart

KOR Knowledge of Results


Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy



Patrick A. Quigley, Jr.

December, 1972

Chairman: John I1. .Howeell
Major Department: Foundations of Education

The purpose of the L-tudy w~:s to proc.dtic-e sel f--control

of academic behavior in ninth and tent: grade typing and

eleventh grade accounting students by providing continuous,

self-generated feedback through various: de-'.vices and to assess

the relative merits cf each devic.-. A e'.-ic'.; of prior self-

monitoring studies suggested apr- op:.-iat cornLrols used

in the research design. The de.vjces tl.'msel;.-es proved no

different from the controls. The .acceleraLiors in the pist

postLreatnment. pr-ove.d rsignl.i -: t .y different fLr'on

all other phases; fo t::o scts of c-t.e. A thi.d set of data

only yielded a trc.nd toward sign ific-.nt *..:':.: dilriing the

last posttrcatLn ent baseline. .', revie..; f some i:i-l i-'i.dual

data revealedl directional chancg-es in rote ' ',Jure 10ot

mes,:-ured, by the ana lysi; of variance t.: :hni'ne. I' ults

surggc Lcd th.. tL tudnpts 'lemonst'-at-l re l.f--. r'ctL=d c lHing

OC pcr-formance rates withL thL dcevjc.s; terminaingr i.le ist

phase coi junctively ilith the endl of tL..., sclhtul. year e::ilains

the fixed interval scallop; statistical significance may

not be sensitive enough to reflect meaningful changes in

individual behavior; and weighting daily work more may

reduce scalloping and assure the longevity of self-control

through the end of the school year.




The purpose of this study was to examine the effects

of self-charting on academic behavior in business edu-

cation classes. The charts served as a vehicle whereby

the student might generate his own feedback with respect

to his daily classroom performance. Efforts weire made

Lo leave the o'.going procedures of the classes intact so

that the effect: of the chart might-. be clearly

assessed. Third; research will have significance in two

respects: tirst, the- chart 'was studied as a device for

sliaping sclf-control of academic behaviors-; and second,

in terms of Lfu-ure wor with self-generated charts, the

experimenter will be informed as to the reacti'.v effects of

charting and thlIus the need to incorporate additional con-

Crol gC.oups or phcsen uwi.thin an experiment to asrcess the

effects of the j.r.(c.-:[)Ondcn variables.


Currently in t.he field of education there is a move

afoot to make ;oii;e of the goats of the educational process


behaviorally or concrtl measurable. The recent clamor

concerning such practices as performance contracting,

accountability, contingency contracting, merit pay, and

oven behavioral objtctives attests to this underlying move-

ment toward the concrete statement of aims in those areas

where such a thrust is technologically feasible. Generally

speaking, those subject areas where the acquisition of a

fundamental --kill is involved are receiving most of the

attention. These areas are encompassed in the traditional

domains of reFading, -iriting, and arithmetic. Other, more

difficult to dofino areas of educational goals are not

receiving as much pressure- for reform. Under this heading

may be lodged such goztls as moral education, cooperation,

and self-control, to name a fecw. Given the hurried atmosphe-_re

of most contemporary classrooms, Lhe task of refoirm Seemts

least likely where the teacher must try some types of inno-

vative practices, in addition to the handling of deimands For

skill acquisition.

Concerning the more ethereal goals. ol- the process,

the teacher at mosL may try to serve as a behavioral model,

to issue exlhortations, or to rcprimaind the studeJnt-s Tor:

undesirable conduct, T1husi, while L-hecse goals ma,y he of

equal importancr: to those of skill learnincj wit"h roferenice

to Coe 9tudenit as a fuburec citizen, they rece~ive7 little

attention and miust' take ,-econd place, until the klsa-cc

With the use of an experimental analysis of behavior,

there remains the possibility of combining both the skill

acquisition and the learning of self-control techniques.

This is only possible because an experimenLal analysis of

behavior requires the behavioral definition of any dependent

variable under study. While school systems do not often

state their goals in terms of experimental analogies, this

language easily lends itself to the statement of routconles

or goals. Thus the teachers, curriculum, and peer-:; of Lhe

student provide the major independent vari,:bles or environ--

ment in which the student's behavior is to be chnil-.-d over

a course of 12 or more years.

Self-control :within the context of an expe::rim.:ental

analysis of behavior entails two responses: tihe controlling

response and the controlled response (S!:innci., 1.953). The

controlling response is emitted by the behave: to mnin ru.clate

those variables which are functionally related to the

controlled response. Thus, the controlling response may

determine the probability of occu-urence .n-d Lo:og-jraphy of

the controlled response. In school learning situti -.onrs,

the controlled response may be identified :s the sitiuden:'s

acad...mic behavior. The controlling r::pon may be any

behavior which the student emits to r.; the occur-

renco and topography of his own academLc rer orrr:m,.nc o.

The problem, then, of -haping cel.--control in cla tsroom

learning situations becomes one of identifying con ILolling

responses which the student may emit to manipulate or monitor

his own academic behavior. Furthermore, in order to make

such self-control feasible, the form of the controlling

response should be one easily emitted by the student and

involve a minimum of teacher influence. The former cri-

terion should be tempered by the fact that at least initially

the controlling responses of the students may involve some

shaping to permit the students to emit the responses.

However, violation of the latter criterion in any extc-nded

sense would by definition not permit self-control to be


Within the context of this research, t.he controlling

response under study was the recording of academic behaviors

on a standard behavior chart used in slightly different

settings (Johnson, 1971; Starlin, 1971; Haughton, 1971;

Duncan, 1971). After a small amount of izistruction and

practice in the use of the chart, the students were left

alone to record their own daily academic behaviors. It

was hoped that, by generating their own feedback for a period

of days, the students would be able to manipula-.e the rate

of their responses in a dci.rable direction. It was the

thesis of this research that the visual diSiploy generated

by the student on the chart would enable him to control

his own academic behaviors.

In terms of research methodology, the study of self-

control explores the reactivity of any given setL'ting or

test. If the setting or testing process produces cues

as to the acquisition of behavior which in itself will

be detected or influence later testing, then the test or

setting is said to produce a reactive effect (Campbell

and Stanley, 1963). The same may be said of sampling

procedures which change the members in any fashion. While

such an artifact may be a probleri to experimenters, it may

be of benefit to educators searching for devices to shape

self-control in students.



This chapter surveys the research and methodological

literature pertinent to this study. For purposes of

organization and clarity, this survey of the literautre

i-s presented in a number of sections which include: self-

cowii:rol, self-monitoring methodology, self-monitoring in

clinical settings; :elf-monitoring in school settings,

and knowledge of results through behavior graphs. A final

section presents a summary of the literature discussed in

ecach of the sections previously mentioned.

Se lf-Control

Over tlhe past half--century, behaviorists have worked

per; isLently to detail a precise science of behavior.

.:t times, this search for law and order led them to confuse

objectivity with empirical verifiable f.icts. These

1'ela'I/io.:is ts argued that unless a phenomenon were visible

aci. I could be reliably measured, it had no place in a science.

'SuchI .imitation:; oin che bounds of psychology caused confusion

lbo:!h within tlh field of behaviorism (Watson, 191.3) and to

a, gr,-ater extent outside of behaviorism (Scriven, 1972)

concerning those phenomena subject to empirical test. By


operating within the limited framework of accepting only

visible and observable behavior, psychology would never

adequately account for such subjective phenomena as moods,

feelings, urges, insights, thoughts, and free will. The

rejection of this last state resulted in little research

on self-control. To grant the existence of self-control

as a legitimate phenomena for study by psychologists

would be tantamount to admitting the existence of free will

and thus backsliding to a rationalist position (EPlles,


One notable exception in the arena of behaviorists was.

B. F. Skinner. Rather than admit the reliance of self-

control upon any special, internal mechanism (such as free

will), he described self-control as a behavioral phenomenon

which influences an individual's subsequent boh.avior

(1953). Thus the beaver emitted a response, calle-d a

controlling r_spon:se, which alters the occurrence oLf another

response, the controlled response within the r:cpe toirc.

The controlling and controlled responses may be either

public (outer or ohb_.ervable) or private (inner or unobserv7bhle

to the public) events (!ommin, 1965; Premack, 1971). Since

;-he private event '.-'..s said to share the same properties as

the public event (Skinner, 1953), self-control may become

a legitimate fi.eld of study for the behaviorist.

For behaviorists working w-itthin the frame-

work, the definition of sel.f-control was operational. zed

in terms of the context of research or therapy under

consideration. Goldiamond (1965) defined self-control as

a functional relationship between behavior and the environment.

Ferster, Nurnberger, and Levitt (1962) referred to self-

control as self-selected change in the environment to

produce changes in the individual's repertoire. For

Duncan (1969), self-control involved the self-selection

and application of behavior modification techniques. Later

work in the area of self-control led Duncan to adopt a

more euphemistic term (personal management) with relation

to the control of inner and outer behaviors (.971). B3rgjer,

one of Duncan's students, has produced the most exLtnsive

examination of personal management techniques for inner

(private) and outer (public) behaviors (1972). According

to :3lackw.ood (1970), self-control consisted of the emission

and reinforcement of verbal behaviors to courltercondition

sonic temptation to misbehave.

All of the above definitions may be encompasscd by

Sk nner's definition which emphasized the effect: oa self-

control on subsequent behavior. Where research has been

done in conjunction with the development of .e-lf-control,

self-monitor ing has been used as a data cotLecti.on device

(GoldJ.amond, 1965; Fer:,ter eI- al.., 1962; Crnca.n. L'99, 1971;

Berqer, 1972). Most of those studies use self-!oni:oring

in conjunction wiLh other indepen'l:ent variabl.os to change

behavior. However, self-monitoring may he studied as an

independent variable by itself (McFall, 1970). Other

articles which elaborate the possible use of inner behaviors

in conjunction with outer behaviors suggest the use of

self-monitoring without demonstrating its uses (Homme, 1965;

Premack, 1971). Cautela (1969) and other articles to be

mentioned later in this chapter suggested possible uses

of self-control procedures in clinical settings.

Methodological Considerations for the
Use of Self-Mionitorinq Procedures

Before examining the self-monitoring research, a review

of methodological considerations is necessary to assess

the validity and generalizability of the data in question.

Since self-monitoring as a data collection technique was

at first used extensively only in clinical settings, the

articles written with respect to validity and generali-

zability of results concern clinical data. These same

methodological considerations seem applicable to educational

settings with few exceptions and thus were applied to these


As result of studies by McFall (1970) and Marston and

McPall (1971) where self-monitoring was used as both a data

collection device and one of the major independent variables

for research on smoking behavior, the following methodological

discussions were prepared (Orne, 1970; Kanfer, 1970;

Mausner, 1971; Lichtenstein, 1971).

Orne's questions concerning self-monitoring data

centered upon the three major points which characterized

the other critiques. First, are behavior changes in

self-monitoring settings due to unique treatments or may

they be interpreted as a result of some demand characteristic

of the experiment? If the noted behavior change may be

attributed to some subtle cues from the experilmenter or

from the larger setting of just being in an experiment, then

demand characteristics may account for the treatment effects.

For example, if the subject either by volunteering or

by learning the goal or nature of an experiment behaved

accordingly, the validity of experimental effects must be

questioned. Orne suggested the use of additional control

groups to assess the effect of the major independent: variable

(or variables) beyond the demand characteristics of an

experiment. Second, in those instances a behavior

is occurring during all wakeful hours, what is the generality

of data taken in limited samples? This question is ans:wred

by either taking day-long samples or by being relatively

certain that the behavior under st'idy only occurs during

the sampling time. Third, Orne questioned the reliability

of the data gathered under self-monitored conditions. He

raised the possibility of the subject reporting faked data.

For example, when an observer of the subject'or behavior is

a colleague or relative, there could be some ccr'imnnicctLion

bct'.;een the observer and experimental subject. Thu:; relia-

bility measures taken to validate the self-monitored data

should be independent of the subject's recording. Having

an observer both personally or socially unknown to the

subject would constitute an ultimate answer. As will be

noted later, this objective was often compromised in experi-

mental studies which attempted to naturalize the setting

and facilitate the collection of reliability data.

Kanfer (1970) raised many of the same questions and

went further to state that the concurrent reliability and

validity of self-reports were maintained when reliability of

reports between subject and observer .:erLe high. The nature

of the observer given a broader definition to include

behavioral products as evidence of a behavior occurring.

Thus showing butts for smoking, weight loss for reduced

consumption rates, longer fingernails :'or nail biting,

fewer bruises on the wife's body for wife-beating urges

may all serve as concurrent validation of Lelf-monitored

data. Kanfer also noted that the self-monitoring may

serve as a controlling response and thus facilitate self-

control. In addition, he '.iewed the results of self-

recorded data .s, knowicdge of results. Hence, the self-

recording techniques used in this study were related to the

general discussion of knowledge of results data later in

This ch..pter. I.'nally,, lie proposed thai: s-el.f -monitorcd

data was a poor subst.iLtute for control procedures prior

to behavior m'lification. Thu:; some independent measure of

r:-sponding or response rates should be gaLthcred prior

to any ;clf-monitoring to capture t:he results of the


Mausner (1971) added little to prior methodological

considerations. However, as an alternative to recording

public behaviors, he mentioned as a possibility the

recording of a private event which represented a competing

response for the public's behavior under study. For

example, clicks on a wrist counter may be used to count

the number of times resistance to smoking urges occurs.

If there was a concurrent reduction in actual smoking rate,

the data on the private event are valid (Homme, 1965;

Premack, 1971).

Lichtenstein raised a further concern which was more

relevant to meaningful therapy than to the research method-

ology of self-monitoring. In the process of doing research,

the experimenter often categorized the treatments into

various groups to obtain experimental control. WLth

regard to the best outcomes for each individual in the

experiment, any one subject may not receive the most

promising treatment due to random assignment. One reply to

this concern would be to use counterbalanced designs which

permit each subject to receive each treatment or to base

the construction of manipulation upon previously demons Lrated

treatments from small sample or single subject research.

Self-Monitoring in
CliJnical Settings

The clinical literature on self-monitoring dealt with

a variety of human behaviors in many different settings.

References to self-monitoring as a self-control device were

not always made. A few studies were reported in detail to

illustrate the use of the criteria discussed in the last

section. McFall (1970) used his own class to study the

smoking behavior of college students. In this study,

M.cFall asked two questions: first, does self-monitoring

alter behavior; and second, does the nature of the target

behavior yield differential effects in the self-monitoring

situation? The validity of the data were judged by the

reliability of observers covertly monitoring two treatment

groups. The first group was asked to count the number and

1,ime spent smoking each cigarette in class and the second

group ..'as asked to count both resistance to urges and the

total number of cigarettes smoked. Each of the sixteen

smokers in the class was placed into one of tihe two

treatment groups and sixteen non-smokers were assigned to

monitor the smoking behavior of the smokers. A baseline

period of nine days was followed by thirteen days of treat-

ment and then an additional nine days were added a'3 a return

to baseline conditions. The smokers recorded only during

the treatment phase while the observers recorded throughout

the experiment. In order to facilitate smoking during class,

the experimenter lit a cigaj:ette and remarked to the students

thit the no smoking sign in the room only applied to after-

noon classes. Furthermore, the experimenter expressed an

interest in collecting data on smoking behavior from the

students. Pol:h treatment groups demonstrated changes

during the self-mo-nitoring phase and these changes persisted

during the rebase phase when self-monitoring was discon-

tinued. Two of th.e eight pairs of subjects reported data

for a day in which. no class was held, and thus all of their

data was suspect. For the other six pairs during the

treatment phase the reliability between subjects and observer

was .61 and for each pair taken separately the reliabilities

ranged from -.05 to 1.00,with three of the six pairs above

.75. McFall concluded that the self-recording was a reactive

measure and thus should be assessed independent of any other


Orne (1970) and Kanfer (1970), commenting upon this

study, suggested te possibility of demand characteristics

explaining the treatment effects rather than the treatments

themselves. By lihting a cigarette, the experimenter may

have communicated his hypothesis that self-monitoring would

change the smoking rates. The reduced smoking rate in the

"resistance to urgs" group may be due to instructions

alone. Stated mor specifically, the reviewers suggested

that the same resuts might have been obtained by merely

telling one group of students to smoke more and the other

less because they were in an experiment. Since it way be

safely assumed tha the smokers in the study continued

smoking all day, th generality of the data based upon the

limited sample was questioned by the authors. in lieu of

evidence to assure honest reports of Chose students not

caught cheating, th authors raised the Possybility of a

pact between smokers and observers. Given the opportunities

for socializing in most college classrooms, this suspicion

does not seem unreasonable.

Powell and Azrin (1968) studied the functional

relationships between the contingent, self-application of

electric shock and cigarette smoking. Three subjects

volunteered for the study and two of them gave names of

co-workers who would observe the smoking five to twelve

times a day. The subjects gave a self-report of their

smoking history and kept records of the count and time of the

day smoking occurred. The major treatment was the use of a

cigarette box which delivered an electric shock upon opening.

The value of the shock was varied from 0.0 milliamperes

to 2.0 milliamperes. The "participant-observers" served to

insure the subject was using the device, collect smoking

data, and check the working order of the cigarette box.

The percentage of subject-observer agreement was considered

acceptable for both subjects (98% and 88%). The main

finding was that as the intensity of the shock increased,

the smoking and the use of the box decelerated. Thus

while the subjects found other means to obtain cigarettes,

there was also a concurrent decrease in the total number of

cigarettes smoked. Once the shock was removed, the smoking

returned to its pretreatment level. The salient feature

of this self-monitoring study is obvious; the participant-

observer system insured reliability. While some suggestions

of a pact between the subject and the observer might be

raised, this was thought unlikely since the subjects

volunteered reports of not using the shocking device as

according to instructions. It would have been interesting

to have a covert observer gather additional data both when

the participant-observer was present and at another time

when the participant-observer was absent. These data would

both settle the reliability question and perhaps reveal the

extent to which the participant-observer may become a

discriminative stimulus for non-smoking. If this third

observer were paid, the social reasons for colluding would

also be minimized. The same authors have also reported

the use of the participant-observer system with similar

reliability in a smoking study using discriminated extinc-

tion through a cigarette box which sounded an alarm, and

unlocked after an elapsed time interval (Azrin and Pcwell,

1968) and in a study using response priming to improve

prescribed self-medication (Azrin and Powell, 1969).

None of the studies reported hence have incorporated

follow-up data collection to any significant extent to

demonstrate potent treatment effects. With a behavior

like smoking, long-term abstinence would be the only

obvious demonstration of experimental control. In an

attempt to compare several types of behavior modification

techniques (anti-smoking pill, stimulus satiation, extinc-

tion, and hierarchical reduction), Marston and McFall

(1971) set up a smoking clinic to attract volunteers.

Self-recording in a diary form was required of all subjects.

The smoking behavior patterns were typical of other data

using similar therapies. The only form of reliability data

available was through the number of cigarettes bought at

the clinic. Follow-up data revealed no differences due

to treatment as all subjects had approximately recovered

pre-experimental smoking rates.

After a review of previous stop-smoking literature

when adequate follow-ups were done, McFall and Hanmmen

(1941) agreed with Bernstein's (1969) conclusion that non-

specific or secondary factors accounted for the reliable

behavior patterns of temporary reduction and then long-term

resumption of smoking. The non-specific factors isolated

were: motivated subjects, structured participation in a

program, and self-monitoring. In order to create a climate

conducive to these factors, the authors advertised a clinic

in the student newspaper, asked participants to postdate

a twenty-five dollar ($25) check to insure against subject

withdrawing during the study, and set up four treatment

conditions. In the first condition, the subjects only

handed in a daily report of smoking; in the second, the

subjects used a wrist counter to monitor smoking and they

said aloud, "I do not want to smoke" with each urge; in the

third, the subjects awarded themselves twenty non-negotiable

points to be counted on a wrist counter for each urge resisted;

and in the fourth, the subjects recorded resistances to

temptations to smoke and award themselves one non-negotiable

point for each occurrence on a wrist counter. Data collected

six weeks and six months after treatment revealed no

differences due to treatments. The authors concluded that

the non-specific factors yielded as great a result as

the more elaborate, theoretically derived, and presumably

more potent procedures. In an earlier study comparing

self-monitoring alone with other self-control treatments

for smoking behavior, Rutner (1967) obtained the same

results. As for the smoking therapies themselves, Bernstein

(1970) has published a more extensive review of the litera-

ture than a review cited earlier.

Self-monitoring has also been used in other clinical

settings. Leitenberg, Agras, Thompson, and Wright (1968)

gave phobic subjects a stopwatch to self-monitor their

own relaxation times in desensitization therapies. The

self-monitoring was validated against automatic timing

devices. Initially, the self-monitoring was confirmed and

praised by the experimenter in conference with the subjects.

Eventually, the experimenter feedback was faded and follow-

up data revealed positive results of the therapy. Weight

control studies have shown positive results with self-

monitoring either by itself (Stollak, 1967) or in connection

with other environmental changes (Ferster et al., 1962).

Self-monitoring in diary or chart form with other

environmental changes (Goldiamond, 1965; Duncan, 1969)

and chart form alone (Kolb, Winter, and Berlew, 1968)

have proved successful in changing various social behaviors.

In a study which revealed unreliable self-reporting and

peer-reporting, Fixsen, Phillips, and Wolf (1972) demonstrated

the use of various progressive ratio schedules of rein-

forcement to increase reliability among delinquents in an

innovative therapy center.

Finally, Duncan (1971) and Berger (1972) demonstrated

the use of self-monitoring for inner and outer behaviors.

Both studies used the Standard Behavior Charts (Koenig,

1972) to record self-monitoring. Duncan had a three-and-

a-half-year-old girl self-monitor selfish thoughts and acts

and a twelve-year-old girl self-monitor feelings of anger

and outbursts. Since the inner and outer behavior patterns

were similar, it assumed that the concurrent reports

confirmed one another. Berger enlarged Duncan's strategies

to include outer behaviors alone (twenty-three cases),

inner behaviors alone (thirty-seven cases), inner and outer

behaviors (three cases in which the urge or the act were

recorded as the same response), and inner-outer behaviors

(eight cases in which the movement cycle was in part outer

and in part inner). These last behaviors included such

pinpoints as eating when hungry and the occurrence of a

meaningful conversation. The author offered as criteria

for valid data such criteria as amount of "bounce" from

day to day on the chart, sequence of high and low rate days,

correlation between charts from the same person, and length

of charting period. These criteria were offered to make

self-charted data valid in themselves without continuous

observer checks. An example of the predictive validity of

these criteria was given for an inner behavior of one

subject. While the need for continual assessment of relia-

bility remains an open issue, Berger's work made a sub-

stantial contribution in exploring new movement cycles

and suggesting criteria for assessing the validity of

self-charted behavior itself. It may prove that the use

of movement cycles such as inner and outer behaviors and

inner-outer behaviors will never satisfy the traditional

methodologist and that criteria such as Berger listed will

provide the only means of judging validity.

Self-Monitoring in
School Settings

Self-monitoring has been used less frequently in

school settings than clinical settings. Two studies of

reliability under peer-monitoring conditions for peer-

tutoring revealed 90 percent reliability between a fifth

grader and his teacher for monitoring the academic behaviors

of fourth graders (Surratt, Ulrich, and Hawkins, 1969), and

90 percent accuracy of plotting rates on the Standard

Behavior Chart for a first grader tutoring classmates

(Starlin, 1971). Both studies also pointed to the successful

and timesaving features of peer-tutoring.

In an investigation of the relationship between non-

verbal and verbal behaviors of preschool children, Risley

and Hart (1968) found a low correspondence between the self-

reports of play equipment use and the actual use of the ma-

terials. When food snacks were made contingent upon corre-

spondence between the use of materials and the self-report,

the use of the materials accelerated to the level of the self-

reported use. This effect also generated to the use of other

materials in which the correspondence was not made contingent

upon food snacks.

Broden, Hall, and Mitts (1971) studied the effects

of self-recording on the classroom behavior of two junior

high school students. The first student went to the

school counselor and volunteered to record the frequency

with which she was on and off task during a history class.

Two observers entered the class to record the percentage

of ten-second intervals the girl was on or off task. Self-

monitoring brougb.t an increase in study time. The agreement

between the two observers varied from 87 to 96 percent.

Since the subject did not record on a fixed interval basis

and even failed to record at times, the direct reliability

between the observers' and the student's report was ques-

tionable even though the overall percentages were equivalent.

Once teacher attention was increased, self-recording was no

longer necessary to maintain study behavior. The second

subject did not volunteer. He was told by the teacher to

record on paper slips the number of talk outs for half of

the class session. Observers monitored the student as in

the previous case (agreement ranged from 84 to 100 percent).

There was little correspondence between the observer and

the subject report. After the slips were removed, the talk-

out rate accelerated. In this case, teacher praise was not

manipulated with the fading of self-recording. The authors

concluded that this fact accounted for accelerated talk-out

rates during the last phase of the study.

The criteria listed in the methodology section of this

chapter may be applied to the self-recording studies in

school settings. A critique of Broden et al. (1971) suf-

fices as a demonstration. Data based upon the self-

recording of the first subject do not lend clear support

to the positive effects of self-monitoring since the absence

or presence of the recording slips alone may have accounted

for the results. Broden et al. (1971) stated that physical

presence of the paper slips alone may account for the

results since the. self-recording was often omitted by the

subjects. To reiterate a previous question, could the

same results be obtained by merely telling a subject at

various times that an experiment is in progress? Since the

data collected by the observers and the subject were gathered

under different conditions (interval sampling as opposed

to unspecified recording), the reliability must be

questioned. As McFall and Hammcn (1971) concluded,

unspecific factors such as a formal setting or subject

motivation may underlie other treatment effects. In this

study, the formal setting and motivation (i.e., subject

initiated counseling) were confounded with the effects of

self-recording. For the second subject, motivation due to

volunteering was not a confounding feature. However, by

having the teacher in control of the intervention, differential

teacher attention may have accounted for some the effects.

While the authors reported taking observations of teacher

attention to both subjects, the data were missing for the

second subject and thus the omission raises additional

questions. Since the reliability for the second subject's

recording was poor, this study represented, perhaps, more

the effects of non-specific factors than of self-monitoring.

Knowledge of Results Through
Behavior Charts

As Kanfer (1970) noted, information provided through

self-monitored data qualified as knowledge of results (KOR).

Studies demonstrating KOR through self-monitoring and

self-charting have already been reviewed (eg., Duncan, 1971;

Berger, 1972). There were a few examples in the literature

of KOR through behavior charts which were not self-generated.

The behavior rates were plotted on a chart by a supervisor

or teacher and then shown to the behaver for evaluating his

performance. Johnson (1971) used the Standard Behavior Chart

to help one student in her first-grade class learn mathematics

and another decelerate his rate of ya.wning. Many more

examples of the effects of the Standard Behavior Chart as

a decelerator or accelerator may be found in the handbook

of Precise Behavior Facts (1971).

Jeus and Shores (1969) working with three trainable

mentally retarded adolescents found that showing a chart of

the previous day's work on a simple assembly task accelerated

performance on the same task. Equivalent procedures did

not yield similar results for ring assembly and packaging


With regards to the KOR literature itself, much

work needs to be done to demonstrate the reinforcing

or non-reinforcing value of KOR (Geis and Chapman, 1971).

While pointing to self-monitored data collection as an

example of KOR, the authors stated that the general reinforcing

value of KOR in relation to schedule control, immediate

versus delayed, and nature of the reinforcement remained

an open question.


This chapter has reviewed the historical context

of self-control %with respect to the realm of behavioristic

inquiry, the methodological considerations necessary for

valid data collection in regards to one form of self-control--

self-monitoring, and the behavioral literature of self-

monitoring in clinical and educational settings. Self-control

was defined as encompassing two responses--the controlling

response and the controlled response. Self-monitoring

of behavior as a controlling response the focal point

of the review. Criteria for valid data collection included

controlling for demand characteristics of experiment so

that instructions alone would not account for treatment

effects, reliability of self-monitored data with some

unbiased source, assessment of the response rates or

other measures of the behavior before self-monitoring

began, control for any non-specific factors underlying

the experimental setting (e.g., structured setting, and

motivation of the subjects), and adequate sampling of the

controlled response to insure generality of the findings.

All of the literature reviewed pointed to the positive

effects of self-monitoring and thus qualified the self-

monitoring response as a controlling response for the

acquisition of self-control. The main types of self-

monitoring responses were in diary form, note taking, and

on behavioral charts. Furthermore, the behavior chart

proved effective as a device for behavior change whether

it was self-generated or not.

With the exception of the McFall and Hammen (1971)

study, all of the studies failed to control for one or more

of variables demanded by the criteria. Since the study

by Berger (1972) both included some new behavioral defini-

tions and cited new sets of criteria for the assessment of

validity, the future course of scientific inquiry alone

must judge the validity of the new criteria. The contem-

porary technology cannot adequately assess the reliability

and validity of the movement cycles used.

As a result of this review, it is possible to restate

the purpose of the research at hand and justify the methods


and design using the criteria cited. As the title of the

present study states, this study was designed to assess the
feedback through a chart. Not all of the non-specific

features of the setting were controlled, rather a few
were purposely left unmanipulated. Thus the structured

setting for instruction in recording on unfamiliar forms
was tested against a familiar form of response recording

(i.e., posting responses and work times on manila
envelopes). Two new types of recording, one of which was

the Standard Behavior Chart, were studied in relationship
to another, more familiar type. The control for demand
characteristics through subjects guessing the hypotheses

of the experiment by instructions was accomplished by

exposing all subjects to the hypothesis or instruction that
all self-monitoring improves performance. Short baselines
were inserted between treatment pliases to assess the

persistence of experimental effects. In order to eliminate
a bias in the results due to teacher commitment or

differential teacher attention which might influence student

performance, the treatments wore counterbalanced in a Latin
squares fashion with the use of the old re-cording type
continuing while the new recording types weirc introduced.

As it was clear from the literatuire, the self-monitoring
is reactive for a number of reasons and thus serves as

a poor control technique prior to manipulation. Therefore,
ana~lt trtfcto rlvn opro lsro

performance was obtained to assess the approximate levels

or rates of the individual performance before any type

of self-monitoring was begun. The reliability of the.

self-recording was checked by each teacher during the term.

The teacher made comments concerning the high reliability.

One set of data was statistically examined to reinforce

these comments. The observers and subjects were unknown

to each other.

Finally, two points deserve further note. First,

if certain variables are known to be reactive, it is only

possible to examine the effects of one reactive variable

against other reactive variables with the treatment of

concern being hierarchically eliminated from other reactive

variables as controls. Thus the use of Standard Behavior

Chart was contrasted with the use of Rate Computation

Sheet (RCS) not because of any special properties ascribed

to the RCS,rather because it required recording responses.

Second, the generality of findings based upon academic

behavior sampled in classrooms is not as questionable as

taking a small sample of behavior in a clinical setting.

In a good number of cases, the academic behavior is under

the stimulus control of classroom. The behaviors observed

in the clinic can usually occur all day long. Thus the

methods and design of this study proved adequate to assess

the effects of visual feedback through a chart upon the

rate of academic performance of junior and senior high

school students.



Subjects and Settings

As a result of a workshop given to business education

teachers, three classrooms were secured for research pur-

poses. Two of the classes were introductory typing at the

ninth- and tenth-grade levels while the third class was an

- introductory accounting class at the eleventh-grade level.

The academic behaviors for charting the two typing classes

consisted of three-minute, timed typing of a sample from

the texts.

In the introductory accounting class, the curriculum

consisted of workbook exercises and bookkeeping activities.

The students in the two typing classes were accustomed

to identifying the errors made during the three-minute

typing sample and making note of these along with the

total number of words or words per minute on envelopes

which the students used to file each exercise. This

recording on envelopes was introduced well in advance of

the first baseline for all three classes. For the ninth-

grade class in which the students posted the words per

minute, the teacher had student-aides proofread the



exercises at a later time in order to identify any additional

errors overlooked by the typist. The other class, which just

posted the total number of words, exchanged their papers

with fellow students who then proofread the paper. In

both cases, number of additional errors identified by the

proofreader was posted on the typist's envelope. The

essential differences then, between the typing classes,

consisted of posting the typing rate for three minutes and

the delayed feedback of a day on the number of additional

errors found by a student-aide in the ninth-grade class as

opposed to posting the total number of words typed for

three minutes and the immediate feedback from a fellow

student concerning additional errors identified in the

tenth-grade class. Students in both typing classes used

the vertical scale in the texts to compute the total number

of words typed. The sample for typing was changed each

day during the experiment.

The ninth-grade class was composed of 29 students

of which 26 were females and 3 were males. Of this

number, 24 females and 3 males were included in the study

for analysis purposes. Of those not included for statistical

analysis of this class and the other two classes, some

were discarded for reasons of prominent absenteeism and the

rest were discarded by the randomization procedures employed

in the selection of subjects. These students did, however,

undergo the same treatments as the rest.


The tenth-grade class was composed of 25 students of

which 15 were females and 10 were males. Of this number

10 females and 8 males were included in the study for

analysis purposes. The remaining students underwent the
same treatments but their data were not included in the

statistical analysis.

The third class consisted of eleventh-grade, introductory

accounting students. These students worked at their own

pace through workbook exercises of fill-in-the-blank

questions and longer problems from the text of bookkeeping
activities. The sequence of classroom activity involved

reading the introductory material in a unit of the test;

answering short questions concerning the reading in the

workbook; and finally completing the bookkeeping problems

from the text in the workbook. The teacher made herself

available during class time on an individual basis t-o help

students and to provide answer sheets. The longer problems

were self-correcting because columns would not balance if

numbers were entered incorrectly or if there were mistakes

in addition or subtraction. By providing answ,.er sheets, the

teacher made attempts to have errors corrected as soon as

they occurred. These students were accustomed to posting

responses on envelopes before experimentation bcgan. The

response unit used for analysis consisted of any fill-ins,
column entries, and mathematical computations accomplished
duigtetta ls ie Det hIoulrshdln

at this school, the duration of class time varied from 50

minutes on three days of the week to 33 minutes for the
remaining two days.
In the 'Accounting class, a total of 21 students were

available for the study. Sixteen were female and 5 were

male. Of the students selected for the experiment, there

were 13 females and 5 males.


In addition to the daily classwork performance by

all students in each of the three groups, each group of stu-

dents was assigned three different methods for recording the

results of their classwork for the day. The first method

of recording was "no method." This recording procedure

consisted of employing the usual method of recording

classwork performance that had been used in the classes

before the treatment phase of this study began. The second

method was the use of Rate Computation Sheets (RCS) only.

A rate computation sheet is designed to facilitate the

computation of behavior rates. Spaces are provided for

entering the amount of time recording occurs and the number

of movements observed during that time. Each division

panel is keyed to the day of the week. Appendix 1 provides

a sample frame for a week's recording on the RCS. The RCS

forms were used by the students. The third method of

recording consisted of the Standard Behavior Chart (SBC)

which is designed to provide a visual display of rates in

graphic form. The ordinate axis allows for recording

behavioral rates occurring as slow as once in a thousand

minutes or as fast as a thousand per minute. This axis is

arranged in a six-cycle, semilogarithmic fashion. The

abscissa permits recording by the calendar day. Samples

of the SBC are provided in Figure 4, 5, 6, and 7 (pp. 67,

68, 72, and 73) without the grid usually given to facilitate


During a given time period of the treatment phase of

this study, each group of students employed only one of the

three methods described above but over the entire treatment

phase each group employed all three of the methods of record-

ing. The method of assigning each recording method to the

three groups is described more fully in the design section

of this chapter.

Instruction in the use of the rate computation sheets

was given during the first two days of each treatment phase.

During the first 20 minutes of each day, the group using

the rate computation sheet (RCS) alone and the group

using the RCS and the chart were instructed in the use of

the RCS together. At the end of the 20 minutes, the group

using the RCS and the chart together remained with the

experimenter to practice using the chart. Generally, this

further instruction took more than an additional 15 minutes

and thus another day of instruction was planned. A second

day was used for review of procedures for both groups. Each

student demonstrated his proficiency with both forms

by logging the previous week's data in addition to the

data from the first two treatment days. The group using the

chart was also given instructions concerning the means of

interpreting this visual display of academic behavior. This

instruction involved brief statements concerning distances

between the daily rates and the record floor (i.e., the

reciprocal of the amount of time behavior is observed for

charting) and the general trend suggested by any sequence

of rates.

At the end of the second day of instruction, the students

were asked to use the particular type of recording procedure

to which they were assigned for a duration of 10 school

days. The teachers were also asked to check the students'

envelopes occasionally during the treatment phases to

insure the daily recording. Without the recording, the

effect of unused charts would be difficult to interpret

and one which the present design was not constructed to

handle. Each treatment phase was terminated by the teacher

removing all forms from the students' envelopes.


In order to assess the effects of the standard behavior

charts as a device for affecting response levels, the

treatment phases incorporate the use of groups not keeping

any new forms, another group using the RCS alone, and a

group using the RCS and the standard behavior charts.

If no significant differences in increase in response

levels occurred for those groups not using the standard

behavior chart, when compared to the group using the chart,

the value of employing the standard behavior chart as a

device for effecting levels of classroom performance would

be questioned. Since all three classes were subjected to

special attention, novelty in the classroom as well as the

presence of a stranger in the classroom, it is assumed

that the effect of these variables on changes in classroom

levels of performance would be about equal for all three

classes during the treatment phase of this study. The

treatment phases are counterbalanced in a La'tin squares

fashion to permit all three groups to receive one of the

three experiences during the course of the three treatment

phases. There are also incorporated into the design four

baseline phases interspersed before, between, and after

each treatment phase. During these baselines, the students

merely did their classwork and posted the responses on

envelopes as mentioned in an earlier section of this

chapter. The actual assignment of the three treatment

methods as well as the overall experimental design of this

study are summarized in Table 1.

During the initial planning of this research, the

teachers expressed concern for the possible significance of

interactions of treatment with ability level of the students.

In order to study this possible interaction, it was agreed

to stratify the samples into three ability levels. C(hih,

average, and low). Since there were no standardized measures

available to permit an objective and replicable stratifica-

tion, the teachers merely used each student's previous

class performances and test scores to assign the students

to one of three ability levels.

Once three ability groups were identified, the students

in each ability group were randomly assigned to one of three

treatment sequences. As stated above, each student

received each treatment once during the course of the

experiment. Thus, there were equal numbers of students to

each strata in each treatment sequence. For the tenth-

grade typing class and the eleventh-grade accounting class

there were two students per each of the nine cells. The

ninth-grade class had three students per each of the nine


The three treatment sequences were randomly chosen

from a possible set of six treatment combinations with the

exception that any given treatment appeared only once in

a given column or'row. The first combination was assigned

to the first ability clustering, the second combination to

the second clustering, and so on for the last combination

to the last ability clustering.

The overall sequence of treatments and subject groupings

are presented in Table 1. This design held for all classes

with the exception that the ninth-grade typing class had

three subjects nested in each ability by treatment order

Table 1


Bl T1 B2 T2 B3 T3 B4

S4 Chart
C S5
i S7
S10 Chart

C2 S15



_____________________ A __________ L .1. A. a

B. .

1..5, 6..15, 16..20, 21..30, 31..35,

.Rale Computation Sheet
.High ability
.Average ability
.Low ability

36.. 45,


cluster. Each baseline lasted for five days and each

treatment phase for ten days. Whenever variations in the

number of days for each treatment phase occurred due to

difficulties in scheduling class time for instructing

students in the use of the Rate Computation chart or the

Standard Behavior Chart, the onset of the treatment phase

began at the earliest possible date. Hence, the baselines

prior to such treatment phases were extended. 1Where this

occurred, the phases were lengthened rather than protracted.

Research Questions

How data from the various phases and treatment

sequences were reorganized for purposes of statistical

analysis has been summarized in Table 2. According to

this design, an analysis of variance permitted the following

questions to be asked:

1. What are the contributions of the main effects

(Treatments, Order, and Ability) to the population variance?

The term treatments is used here to mean any one of the

seven phases, whether experimental manipulation phase or

baseline. These, of course, must show some differences

in order for additional and more pertinent questions both

to be asked and answered succinctly. If the treatments

are significant, there is justification for proceeding with

post hoc comparisons of means to identify the sources of

this variance. The treatment order was not a central

question to this study since the order variable was

Table 2


I -_ I *



3 C2 1
C3 S17

A....Order of treatments
C.... Ability levels
bl...baseline for classwork alone
b2...classwork alone
b3...baseline for classwork + RCS
b4...classwork + RCS
b5...baseline for classwork + RCS + Chart
b6...classwork + RCS + Chart baseline

.2 C2



i.. ..........

originally used as a device for counterbalancing three

treatment combinations for each treatment phase and for

assuring each subject's exposure to each treatment.

Traditionally, a design controlling for order effects is

used when there is some support for the belief that the

effect of the independent variable or variables on the

dependent variable is irreversible. Such is not the case

with the treatments used in this study. Ability in itself

is not a prime concern to this study. Rather, in relationship

to previous behavioral studies (Rosenfeld, 1972), there is

a concern for the Treatment X Ability interaction. Should

the main effect of ability prove statistically significant,

this would indicate that initial differences in ability

among students contributed to differences in levels of

performance but would not contribute to the major question

being examined in this study, the effect of recording

procedures on levels of performance.

2. Are there any differences among the interactions

of the main effects? Of concern here were the Treatment

X Order, Treatment X Ability, and Ability X Order interac-

tions. A significant Treatment X Order interaction would

yield valuable information with regard to the primacy or

recency of the effects or, by association, the effect of

other naturally occurring phenomena within the school

setting (i.e., grading periods or termination of the

school year). Interest in the Treatment X Ability has

already been suggested by past behavioral literature. A

significant Order X Ability interaction would be difficult

to explain in terms of past behavioral literature but would

nonetheless yield information with regards to the work

patterns of the various ability levels over the duration

of the school semester.

3. Are there differences between the seven treatments

themselves? If the treatment effects are significant,

then mean comparisons by the Scheffe's technique will be

done to identify the sources of variability. This question

will also permit a comparison of baselines (both pre- and

post-) and experimental manipulation phases. If the data

for the manipulation phases prove better than their

pretreatment baselines, there would be some evidence to

suggest the effectiveness of the manipulations and thus

self-control on the part of the students. If the mani-

pulations prove better than their posttreatment baselines,

then a reversal shall have occurred and self-control on the

part of the students thus diminished. If the opposite

occurs, self-control will have been retained, unless some

alternative explanation is possible because of the occurrence

of events beyond experimental control. Such events might

be an upcoming grading period in which a given amount of

work needs completion or possibly the completion of the

school year and the attendant rush to finish incomplete ....'

work. An inspection of the means for experimental mani-iii

pulation pha:sees:will suggest any differences among th i

.,aiiiiiiiiiiiiiiiiiiii ,,

three treatment combinations. If the recording in count

form on envelopes alone proves better than the Rate Compu-

tation Sheet (RCS) alone and the RCS with the chart, then

there would be reason to suspect the overall novelty of

a visitor, intraclass communication, and the fact that

the instruction in the type was given by teacher and not

by the experimenter account for the difference. If the

RCS recording alone proved superior to recording on

envelopes and charts, then just having a novel form or

practice using novel forms (the RCS alone is given after

the RCS and chart manipulations for the first two treatment

orders), or perhaps any difficulty the students might have

had using the chart with the RCS might explain the difference.

If the chart alone proved better than the other two manip-

ulations, then there is justification for believing

the usual display through the chart accounts for the

difference. Should there be no differences among all three

manipulations, then each must take credit as controlling

responses for varying reasons. The posting of a response

count would serve as a controlling response; however much

this recording can be explained on the basis of intraclass

communication or novelty of a visitor to a class. The

use of the RCS alone would also qualify as a controlling

response even though its use was originally intended by the

experimenter as a control for the novelty effects of intro-

ducing a new form to the students. While it has been argued

earlier in this study that the use of the chart, when under

tight experimental control, would alone represent a con-

trolling response less confounded with extraneous sources

the self-control acquired through the use of the chart may

be no different in terms of ends of performance than the

other two controlling responses in contrast.

If any one of the sets of data is so contaminated

with procedural difficulties that the data collected by the

students is rendered suspect, an accounting of these diffi-

culties will be made anecdotally and quantified to show

the extent of the damage done to the original experimental

strategies. Possible sources for such difficulties may

include: the experimenter failing to instruct the students

in the use of forms adequately enough to be used by the

students alone; the experimenter or teacher failing to

terminate experimental phases; teachers changing the class-

room procedures after the experiment has begun; teachers

not giving students the opportunity to use the forms;

students not being able to perform the simple, mathematical

computations necessary to use the forms correctly; or

students failing to use the forms for long periods of time.

While most or all of these difficulties could be remedied

as they occurred, this would entail a radical change in the

classroom environment and thus render clear demonstrations

of self-control impossible.

Finally, the last features of data to be analyzed will

be those pieces of individual data which either hold clini.bal

significance with regard to the control of academic behavior

or suggest meaningful significance beyond the significance

or non-significance indicated by the particular statistical

analysis used in this study. Since the import of this study

lies in the test of possible techniques to be used in class-

rooms, any data which may hold some application value is

considered worthy of mention. Furthermore, such individual

data may suggest tentative hypotheses for further classroom

research or intervention.

If there are missing data, there are two solutions which

may be used. Should there be missing data and there appears

to be no accounting for the loss due to the nature of the

treatments, an unweighted means solution will be employed.

If there are missing data and there is some reason to

believe that the nature of the treatments in some way

accounts for the missing data, then a least squares solution

must be employed (Kirk, 1968, p. 204). Thus, should a

student complain of his difficulty with the forms and

subsequently refuse to perform, there would be justification

for using the least squares solution. Alpha levels for all

statistical analyses were set at the (p < .05) level.

In summary then, three classes were secured for

research at the junior and senior high school levels. Two

were in typing and the remaining was in accounting. All

three classrooms were accustomed to posting the number of

responses on envelopes before experimentation began. Later

this response was to be contrasted with two other forms

of controlling responses: recording responses in rate form

on a sheet which lacked visual, graphic display of the

academic behavior and recording responses in rate form on

a chart which provided visual, graphic display of the

academic behaviors. The response units for the typing class

were words per minute and error rate for a three-minute,

once daily, timed typing. The response units for the

accounting class was any fill-in, row entry, column entry,

or mathematical computation occurring during the class

period. Counting responses and recording them in one of

three forms occurred once daily during the five-day school

week. Manipulations were counterbalanced in a Latin squares

fashion to insure equal exposure of all manipulations to

each student. Baselines were inserted before, between, and

after each manipulation phase to provide a contrast for

each manipulation phase. The students were stratified into

three ability levels and randomly assigned one of three

treatment sequences. The treatment sequences themselves

were randomly chosen from a possible set of six combinations.

The major research questions were listed and possible

answers were discussed with reference to the various

results which analyses may yield. Possible procedural

difficulties were suggested and the means of handling these

data were mentioned. The possibility of presenting individual

data was indicated and the reasons for this presentation


were given. Lastly, the techniques and rationale for

handling missing data were discussed.



This chapter is divided into four major sections:

analyses of group data, analyses of selected individual

data, reliability, and summary. There is a summary after the

first two subheadings to facilitate the reading. Before pro-

ceeding to review the analyses of group data, the reader

should refer back to Table 2 and review the treatment

designations since cross-reference will be made to these

designations in the body of the current chapter.

Group Data Analyses

Three analyses were run: these were on the accounting

work rates for the eleventh-grade class, words per minute

and error rate for the ninth-grade typing class. The

data from the tenth-grade class in typing were not treated

in a similar fashion for reasons to be given later. A

log (base 10) transformation was performed initially on

all rates because the effects of one of the major independent

variables, the Standard Behavior Chart, contained a lo910

transformation. Due to absenteeism and to one student

refusing to perform when asked to use the recording forms,

there were missing data and these were generated by us51n
..c...... '::

the least squares estimates provided by biomedical computer

program X-63 (Dixon, 1971). This program performed regression

analysis on the data. Using those parameters from the

regression analysis, an estimate of the missing values could

be obtained which would not radically effect the total

variance. A check on the validity of these estimates was

obtained by comparing the sums of squares for error using

the X-63 program and the sums of squares obtained from the

analysis of variance table.

If the estimates are satisfactory there should be very

little difference between the t'w.o error terms. The differ-

ences were felt to be negligible as indicated in Table 3.

The analysis on logs of accounting work rates is

presented in Table 4. The treatments were significant (p>.05),

yet there were no significant differences between the

manipulation phases (B2, B4, B6) and their pretreatment

baselines (Bl, B3, B ) using the Scheff4's multiple comparison

technique. There was no difference between the manipulation

phases and their posttreatment phases (B3, B5, B7). Ranking

the means of the treatments and making all pair-wise com-

parisons showed that B2 was significantly smaller and B7

was significantly larger than B1, B4, B6, B5, and B3 using

Scheff6's criterion. The only two treatments that differ

are classwork alone in a manipulation phase (B2) and the

last posttreatment baseline for all treatment orders (B7).

Table 3


Data Source Sum of Squares for Error

X-63 ANOVA Difference

Accounting .7010 .6998 .0012

Words per minute .0639 .0640 -.0001

Error rate 1.5048 1.5703 -.0655

o r- ,%
,-0 IN L0




0 0 0

or 00 0 "
N N c
0 0 0

0n (N -4i m

C0 C0 0 C C0

'.0 ( (N U -
* *





Mr U
H Ou


O4 0



4 Q)
C1- 4


0 E







r 0 O .-4 1-1 0 0
SO CC) m % CC
in LA *1 r'T %Z F r- A Ir m
r n n r-4 02 r) (N r- 0m
o 13 n r-4 m r- r"i V.0 '

C4 04



*-q > >1 S
-4 1-4 4.J 0
.- I --4 N

X0 E -4

*w ( 4 4- 0 0 \2
W '-I W *r 4 >
>o < l Q 0 *) a 5 x
1- O E 1 3
V *H -\ .0 O 4 O 4
0 pz 0 E- X X X X


The ability levels also differed significantly with

respect to means (p, < .05). Their means rank in the

expected order of high ability highest, average ability next,
and low ability last. The Treatment X Ability and Order X

Ability interactions were not significant.

The Treatment X Order interaction was also significant

(p < .05). An inspection of the cell means in Table 5 for
this interaction shows no significant patterns with the

exception being the last posttreatment baselines which for
the first and third orders are higher than all other cell

means. The cell means for classwork-alone group in a'

manipulation phase showed lower work rates than their

pretreatment baselines for the first two orders. Since
according to the original temporal sequence of the experiment

(see Table 1) the cell mean for the classwork alone in a

manipulation phase for the second order came in the next-
to-last phase of the experiment and since the rates were

higher for the last posttreatment baseline in com-rparison
to this treatment, even the second order tends to support the

conclusion that the high work rates during the last post-

treatment baseline contributed to the variance to this


With regard to self-control, it cannot be said that

the manipulations were effective as accelerators for per-

formance. Further, it appears that for the group continuing
to,,,rkasuua drngth ntodcio o ors oote





i- 22

H u
2 MZ

U U 0-
0 E- U

0 E- U)

2 <



< d



U C l









4 4

~ Ln

M -in
Q'> .-.LA


'.0 -
m ^o


m -.M

'ym *T
r -l

N -

coo -



0- .

-D ---.



- "--'
r- (N

rc oco

aM n
N -1



m rLn

rn r



Co -
m ,-1


u'l --4





groups, their exclusion served as a decelerator for per-

formance rates (except for the third treatment order and

here the cell means of .1718 for the first baseline and

of .1736 for the classwork alone during a treatment phase

were hardly different). With respect to the continuous

feedback through almost immediate knowledge of results

provided by the teacher and answer sheets, the self-

generated feedback did not implement accelerations signif-

icantly beyond those occurring through other available

sources. The superiority of the occurrence of final

examinatiDns and termination of the school year to serve

as accelerators was far more significant than any source

of feedback in this class environment.

The analyses of logs for the words per minute data

from the ninth-grade class showed treatments to be signifi-

cant (p <.05). Using Scheffe's multiple comparison

technique, it was found that the three manipulations

(B2, B4, B6) differed significantly from their pretreatment

baselines (Bl, B3, B5). There were no differences between

the manipulations and their posttreatment baselines (B3, B5,

B7). Ranking the means and making all possible pair-wise.

comparisons showed that the last baseline is significantly

higher than all of the other treatments while none of the

other means differ.

Ability levels differed significantly (p<.05) and

they were in the expected order: with high ability highiiti,

the average abi::ity next, and the i.,lrW ability last.

"' ,4 ii1111111 1

The Treatment X Order interaction differed significantly

(p <.05). Table 6 sheds some light on this interaction.

There is a positive acceleration of cell means for the

logs of the data with only two exceptions occurring as

noted. The rank order is almost precisely the same as

the temporal pacing of phases as in Table 1. This trend

was shown less pointedly in the accounting work rate data.

In the typing data, however, there are no reversals occurring

as a result of doing regular classroom work while some

students are receiving instruction in special forms. Using

the data from Table 6 and the mean comparisons of Treatment

effects, it can be shown that the data points are positively

accelerating with the only exceptions occurring where one

of the three manipulations is in effect. Thus the treat-

ments are functionally related to major upward movements

over baselines with the accelerations slowing during each

posttreatment baseline (with the exception of the seventh

phase for the second treatment). Without regard to treat-

ment order, however, the last posttreatment baseline

represents an acceleration over the highest manipulation

phase mean.

The words per minute data indicate the occurrence of

self-control during manipulation phases for all three devices

in comparison to pretreatment baselines. Further, there

are no data to suggest that any one of the manipulations was

any better than the others with regard to its properties for

CN in o 0o C iC en -
00 m r- o qzr r-4

0 0 0 0
4 4 4m r, *
l C -n *n *
0 E >x X .3 0 Q .0
S0 0 0 = 3
p a) N N n En In i

Wl E-4 Ei x

> C 0) I 0r < oo m 0 4 o
OH a0 C I 0 0 0 0 0 0 0
0 P Y: o 0n o o o o o o o


: D O
nO a
ko >4

a-) 20 W o0
(n c n U) o

E- I U () N rTC) rM ,r n
0 z a) (U) raN

H0 H
M 3 tnC 4

In Nn r -O 4 O %D

^ *-H

,-f .-
4-1 0) ko C,- m i-I m 4 O

00 0 0 C r r t H
U3 0o r- o -4 n r r-I (N .60
E 3 c-a C) r-' CD C) 0 o -,o

>1 ro -4f ri













O -

Lrl <
* r-

Ln -




r-l M

* o








Q; E-4

7 <

u Q

-m 4
n *T


* c

n *
m* N



in *

i- .

* n

N .

0 (

a) a)
c 0
E a

(1) a)

4-) 4-)
-i .-i

0 0


0 a

>4 4
G Cp

*H *.

0 0

o o

I .,4 4
a) (a

> >

o o

-4 P-4
'0 '
4-J )4

0 0
a4 a)

a) a)


0 0
14 4-i

- 4- a -4 a4

o0 o
a a

Qn aU

a) 4a)
(U rr l)
j=:0 :
E- 4i -i4-
*K *

helping students demonstrate self-control. Thus this self-

control must be explained as being confounded with novelty,

intraclass communication, and perhaps differential teacher

enthusiasm occurring at the beginning of manipulation phases.

The multiple comparisons data suggest that self-control was

lost after manipulation phases and the naturally occurring

events as final examination period and final exams possess

much more control over academic behaviors than the students

themselves possess.

The analyses of variance on the logs of error rate for

the ninth-grade typing class as indicated in Table 8 showed

only the Treatment X Order interaction to be significant

(p <.05). The data do not lend themselves to clear

interpretation since the treatment means do not progress

in any orderly fashion nor in the original temporal sequence

of phases as in the words per minute data. By inspecting

the magnitude of treatment means there appears to be an

acceleration of error rates for the last posttreatment

baseline. Thus it appears that as the words per minute

increase, there is an increase in error rate but not to

a commensurate extent. The use of the chart alone seems to

facilitate control of error rate however,its use is

not better than the precharting baseline. Even this

conclusion must be modified by adding that the precharting

baseline is only better than the charting phase in the fir:t

treatment order. Without even statistical support, many of



0 E

0 )


0 E






r'- Ln o
(N CN -4
Ln m '.

o o 0
\ S. \
r *r-I *r-)
0 0 0
Sr -
La a LA L



En w

F <

4 (0 0

0 0



00 n
mn e (N mO Ln D (Ni in
, ,N o ,00- 0 0-
Lin I0' Ln o -r -4 .-- r-4
-4 m 'I (N O 0 0 0 0

4N (N N 0 k. N CN r LA

'.o LO '. r-4 ON '. r- m
(N VA oo 0 L o 0
i-4 (N O (N Ln n r- 'J r-
n r- co co m LA H LAn

-4 > I XH
S0 .>1 ; .
H 4i X 0
--4 .I 3- -,. -

-4 -4 -i \ 4J O < O L
a) r- -r c X X x
0' -<- '0 .. E 4P 4. 4. 4.
O ,r O LA H E H

m N c
m (N m

in Uk OD 03
e a c a

N. 3 3
00 0 .0 0

U) L) LA 1n
4J 41 4 4-1








the above conclusions are at best tentative for this group

of students as a class. There is support in the typewriting

literature to anticipate increased error rates with increased

overall typing speed (West, 1969).

With regards to self-control, these group data offer

little support for the presence of controlling responses.

The effects of self-generated feedback through the chart

were only suggestive in two of the three treatment orders.

Furthermore, with the highest mean occurring in the last

posttreatment phase, the data share some of the same con-

clusion from the other data sets--that the occurrence of

events such as final grading periods and termination of

the school year possess more control over academic behavior

than the students themselves possess.

The data for the tenth-grade typing class was so contam-

inated with procedural errors that the analyses performed

for the other data sets was not done. A summary of the

difficulties collecting data representative of what the

original procedures demanded includes poor communication

between the experimenter and the teacher, teacher forgetting

to have the timed typing daily, rushing the students through

the recording of rates on the forms, starting the timed

typing before each student was ready with a piece of paper

in the carriage, and not checking to see that the forms were

being used daily. There was also an example of a few studeflts

calculating the rates one way during the manipulation ph.aaiis

and in another fashion during the baseline phases. Related

to all of these faults which must be shared by both experi-

menter and teacher, the students themselves failed to use

the forms as frequently as demanded by the procedures of this


Figure 1 presents the percentage of form usage by the

entire class and by temporal order of introduction to the

RCS and the chart. These figures give some justification

for believing the loss of data for the chart is not solely

due to construction of the chart itself. There was some

expectation on the part of the experimenter and all three

teachers prior to introducing it to the classes that the

students would experience more difficulty with the chart

than the RCS due to the semi-logarithmic scale on the

ordinate of the chart. This, however, was not the case.

The greatest losses temporally occurred in the second order.

This may be explained by a four-day holiday occurring a few

days after instruction in the use of the forms and still

another four days before timed-typing exercises resumed.

The phase was thus extended to compensate for the holiday

and lost typing practice.

It is clear from Figure 2 that the greatest loss of

data by ability occurred in the low-ability groups. Even

here it may be noted that the chart fared no worse than the

RCS. Rather, by reviewing the forms of all three groups,

the experimenter found a number of division mistakes for

r, a
m \I nmP



;;; -c-
m r--

4-4 1-1
0 0


,-4- J


>1 0












0 U

.- 0 *
co ,-- o-,

11 11 1 fip, I- I(I

00 0 o 0 to -p o In0
00 o o
0n 0 0

afiesn wuoa 30 %

(Y) MD




AqT[TqV /q a6esn w2oj go %

* 0
\0 kD



0 o
0 o

O o


- h

the error rate data. The inability to do division when

not supervised by the experimenter perhaps accounts for

the greatest loss by the low-ability group with both forms.

Figure 3 presents the percentage of data loss for the

Treatment X Ability X Order interaction. Overall, the lines

restate the representations and explanations given for

the Treatment X Order interaction (Figure 1) and Treatment X

Ability interaction (Figure 2). The notable exceptions

occurred for the high- and low-ability groups in the second

and third order, respectively, using the RCS. There does

not appear to be any plausible explanation for handling these

data which would also explain the more frequent use of the

forms by the average-ability RCS users.

In summary, then, an analysis of variance was performed

on work rates for the eleventh-grade typing class and on the

words per minute and error rate data for the ninth-grade

typing class. Data in the form of percentage of form usage

alone was presented for the tenth-grade typing class because

the student performance data did not come in adequate contact

with the procedures for the study. The analysis of accounting

data revealed significant Treatment, Ability, and Treatment

X Order interaction effects. A Scheffe's mean comparison

revealed the last posttreatment baseline to be significantly

different from all other treatments. Furthermore the com-

parison indicated that for individuals continuing to do

regular classwork while others received instruction in the


I- 1
i H

In o 0

r- r - ,.
t-. x 4- F
\* J 4J 4J ? 0
/ 4 C4 I-c
S\. / 0 o !:
I. m w m m V
1. / ,cu u ,c ,c
-* \ ./ lC < >

L / O ) u r -I- > 0 F-4
C* H / U C!
SLI\ I.4 C. .
* *

\ / X -I
u a
oooo o
,\r oo,. ooo i-

0/ O r L --
A/ I ^

ii a
I / i


,/W'/ ,/_ :

I. / 4
I. /

,',' / o

"- ... .- --

om' o

"-to m. > fl
I I-. 0

Suc u C4 1 C4

C) 0 0 U -- .
010 0 9
-i.nTTTTqv pIu p o Aq arPisfl -oa 1o %h
OO O o 0 In 1 in =
C) to 0 O

use of novel forms there was a decelerating effect. It was

concluded that no self-control wad demonstrated but rather

that the control of academic behavior was under the control'

of final examinations and the approaching end of the school


An analysis of variance on the words per minute for

the ninth-grade typing class showed significant Treatment,

Ability, and Treatment X Order interaction effects. A

Scheff6's mean comparison showed all three manipulations to

be different from their pretreatment baselines but not

different from their posttreatment baselines. It was'con-

cluded that all three manipulations, and not any one in

particular, served as accelerators. An overall mean com-

parison indicated the difference between the last post-

treatment baseline and all other phases. Thus the self-

control that was demonstrated during manipulation phases

was confounded with a long-term, fixed interval scallop

peaking with the occurrence of final examinations and the

end of the school year. These data did not reveal any

decelerating effect of doing classwork alone during manipu-

lation phases.

The analysis of variance for the error rate data of

the ninth-grade typing class was less revealing and only

suggested the usefulness of the chart in reducing error

rate and in reconfirming the fixed interval scallop with

the impending examinations and end of the school year.

Both sets of typing data tended to reaffirm the finding of

increased error rates with increased typing speeds but not

to any demonstrable extent.

Because of procedural difficulties which did not occur

in the eleventh- and ninth-grade classrooms, the tenth-

grade class was analyzed only in terms of the percentage

of form usage. The data for the population did not indicate

that the students had any more difficulty using the charts

than the RCS. Looking at the loss by temporal order con-

firms the procedural interruptions and other weaknesses as

cited. The extreme loss of data with low-ability students

may, at least, in part be explained by the inability of

these students to readily perform the required division.

Inspection of the loss by Treatment X Order X Ability inter-

action reaffirm the conclusions drawn for Treatment X Order

and Treatment X Ability interactions. Where exceptions were

noted, no explanation was thought plausible.

The main conclusion to be drawn from all of data

analyzed was that where a demonstration of self-control

existed it was confounded by fixed scallops (a positively

accelerating increase in rate). This finding was clearly

indicated in the accounting and words per minute data but

only suggested by the error rate data. Suggestions concern-

ing the future prevention of data loss and the flattening

of the fixed interval scallop will be discussed in the

fifth chapter.


Selected Individual Data
The individual data are presented to suggest, frt
that the group statistical analysis used in this studyma
not represent meaningful significance with regard toefct
of the chart upon the rates of academic performance ad
second, that for at least one set of comparisons ofin-
vidual data there is demonstrated clinical value forth
inspection of individual charts. The first set of dat
came from two individual subjects in the eleventh-grad
accounting class.
Both charts contain the baselines prior to the chatn
phase for each student. In both cases the obtaine rte
are contrasted with the record floor (i.e., recipoafr
the amount of time the behavior was occurring forcatn

purposes). Most noticeable in both sets of dataarth
downward trends during baseline periods and the rvra
of these trends during the charting phase.
As may be seen in Table 9 for the mean comparsn
(Glass and Stanley, 1970) there is no differencebewnth
means of the two phases for each subject. The significac
then which seems meaningful is the directional change o h
trends and further this change is not statisticalyvsbe
It seems appropriate then to recommend the relianeuo
visual inspection of celebrations for ascertainingth
significance of effects in those cases %,here the en r
siia ndtevrineIresal

-I0 i" :


O t

- c )


-1 . W.CMMT~

0 Lg









I o
-' 0

(-I--;- 0

31in1N;I U',Jd G.l.;C J i3AO'l"



O .


o a

J x
S0 o

F )

..o o


:o ,

o,, <"3 o
q Q iiiiiiiiiiiiiiii
3 i n,,,,,3,1,,,,,,,,,,,,



E-i 0)

ilE- i

E-4 LI

E En


0 Cn W


.- 0-00
- r-(N

* m

o -
- c

in -



Ln -

C% r-4 r-q

co -

"r '0G'.



U) N
(N -

Imn .

o -m

(N -

co o
N -


Ln ) (N


r 1-

rI- -


,-q a)

a) -
r- (N. 0c

L* -
* m

in11 m


LA -4


* r

o -

* N

* N
* (N

r~ "
N'-- s


c r


Table 10


Phase Mean Variance N Test of

Subject 1

Baseline .4048 .0098 3
t = .0802
Charting .4027 .0160 10
phase [2.53]

Subject 2

Baseline .1920 .0218 5
t = .0036
Charting .1924 .0411 6
phase [1.56]

Critical value for Subject 1:
Critical value for Subject 2:

t.05, 11 = 1.796
t.05, 9 = 1.833

Figure 6 is a chart of a student who represents an

absentee problem to the school. Besides the scattered

instances of missing classes, there was one gross example

occurring from the middle of the third week lasting until

the end of the eighth week. While it is clear that this

student does not attend school frequently, it is also

evident that there is little declaration between the last

few work rates before the break and the work rates occurring

after the break. When this student was in school, he worked

quite consistently. The clinician's task would then become

largely one of keeping this student coming to class and not

one of further devising techniques to motivate the student

to work while there.

Figure 7 represents an entirely different case. This

student, as the chart depicts, dropped out of school. While

there are instances of absenteeism as with the previous

student, this student's behavior demonstrates the work of

punitive contingencies even while in school. Not only did

the student dislike coming to school, he also found the work

aversive. This conclusion was also supported by the

presence of various scatological comments peppering his

work envelope. Thus the clinician's task here would be to

reinforce both attendance and higher work rates.

The differences between the two students whose data

are represented on Figure 6 and 7 are only noticeable because

of the calendar-day comparisons of rates provided by the



C) i0



0 II)



0 0

JJ 0

0 0
*- *rq
> >
r3 ra



r,-.7 ?YVY'q~..

111111 1
0 o
O o

iI 0


N ..

. I

T *rT'. i~i' rar'-riT!!!"""--irrnyi~7*--;r O

* L)

9 '




co -

o o
0 o
0 ,o

I--I s.r tl~:1~ nill
'J1 i~,rJ .1,

Standard Behavior Chart. Without this information, it might

have been quite easy to mistake the two,given the frequent

occurrences of absenteeism. As may be noted by the various

recommendations, these students are different and the types

of intervention required would differ in each case as sug-

gested by the charts.

In summary for this section, it has been noted that

the reliance upon the visual inspection of individual data

is necessary at times to assess the meaningful significance

of manipulations and to suggest appropriate interventions

where they can be attempted. For those cases where treat-

ment means are similar and the variances are small, a

statistic relying upon mean comparison is not as useful as

a directional or trend change. Further, it was suggested

that the calendar-day comparisons of rates on the Standard

Behavior Chart are not only suggestive of diagnosis but also

of prognosis for intervention with academic behavior prob-



Due to the need for reliability with self-monitored

data, the teachers were asked to check the students' en-

velopes on an intermittent basis. The teachers spent

additional class time checking the recording on the en-

velopes daily upon initial introduction and thereafter only

checked on a random basis after class time. All teachers

reported accurate recording.

As a demonstration of this reliability, the error rate

components (i.e., number of errors found by the typist and

the number of errors found by the proofreader) were analyzed

for the ninth-grade typing class and the aides who rotated5

the responsibility of checking the papers. A technique sug-

gested by Winer (1962, pp. 124-128) was used to compare

frequency of student error and the frequency of proofreader-

identified error. Five days were randomly sampled and the

analysis yielded the following coefficients for those days:

.94 for the first day, .90 for the tenth day, .96 for the

twenty-seventh day, .82 for the fifty-second day, and .96-

for the sixty-third day. Summary tables for the analysis of

variance used to compute the reliabilities are located in

Appendix 2. These coefficients were computed by subtracting

the quotient of the mean squares within people divided by_

the mean squares between people from 1. These reliabilities

were felt high enough to demonstrate accurate recording.


Both group analysis and individual data were reviewed.

The group analysis of the eleventh-grade accounting class

suggestsabove all else the significance of the occurrence of

final examinations and the termination of the school year as

reinforcers for academic behavior. The analysis of the

ninth-grade words per minute typing data confirms this

conclusion significantly but go on to lend support for the

experimental manipulations as accelerating the fixed interval

scallop at intermediate points along the curve. The analysis

of the error rate data for the ninth-grade typing class only

tends to show support for the fixed interval curve and to

suggest that the Standard Behavior Chart may have helped the

students gain control over their error rates. An inspection

of form usage by the tenth-grade typing class indicated that

the chart is no more difficult for adolescent students to

use than the RCS and that low-ability students may need

remedial instruction in simple division before these forms

may be used without the aid of some prosthetic device (i.e.,

calculator or frequency finder).

The individual data were presented to indicate a differ-

ence between meaningful direction change of rates as opposed

to statistical change for the chart users. Two data sets

were then added to indicate the usefulness of the chart in

depicting individual differences with regard to academic

performance over time.

Self-control is said to have been demonstrated in a

statistically significant manner during the manipulation

phases for all recording responses in the words per minute

data of the ninth-grade typing class. The presence of

self-control was only suggestive for chart users within

the error rate recording for the ninth-grade typing class.

Two individual cases were presented for the accounting classii

in which significant self-control was deiiiimonstrated.. meaniing-a....

fully, but not statistically.


Accurate self-reports of work rates were insured by

teacher checks throughout the semester. A statistical com-

parison was performed for reliabilities between the student-

and proofreader-identified errors for five days selected

at random. The coefficients were high enough to demonstrate

accurate recording.



This final chapter contains three major sections:

discussion, implications, and summary. The discussion

section provides a review of the results as related to

prior literature and how this study has added to that body
of knowledge. The implications section reviews the results

and presents recommendations for future research and
classroom practices. Finally, the summary provides a

precis of the entire text.


The literature pertinent to this discussion falls under

three headings: self-monitoring, self-control, and knowledge

of results. Previous research has indicated that self-

monitoring may serve as a device for behavior change

(Stollak, 1967; Rutner, 1967; Leitenberg et al., 1968;

Kolb, 1968; McFall, 1970; McFall and Hiammen, 1971), that
self-monitoring must be validated by an observer or ob-
serving device (McFall, 1970; Orne, 1970; Kanfer, 1970),

that self-monitoring may be included as one of many non-

specific factors such as structured situation and subject

motivation accounting for behavior changes in research

settings using elaborate treatments (Orne, 1970; Kanfer,

1970; McFall and Hammen, 1971), and that the effects of

self-monitoring are persistent for a period after self-

monitoring is terminated (Leitenberg et al., 1968;

McFall, 1970).

The analysis of the group data for the words per

minute data and for two of the pieces of individual data

indicated that self-monitoring during a treatment phase

served to change behavior in a desired direction. Besides

replicating previous research, this finding provided in-

formation with regard to the measure of the dependent

variable. Many of the prior studies use only gross number

or percentage data. The present study employed rate measures

and obtained many of the same findings reported by others.

Explanations for the non-replication in the accounting and

error rate data are presented under the discussion of

self-monitoring and non-specific factors.

Although steps were taken by the teachers (through spot

checks) in all three classes to insure reliable data

recording, there was only one set of data available for a

statistical description of the reliability. This set of

data was obtained from a comparison of self-recorded and

teacher-aide-recorded errors in the ninth-grade typing

class. Since the student and the aide were unknown to one

another, the reported high reliabilities were taken as

honest measures. It was unfortunate that the tenth-grade

typing class data could not be analyzed. The proofreading

technique employed in this class was a variant of the

"participant-observer" method used in clinical settings

(Powell and Azrin, 1968; Azrin and Powell, 1968; Azrin and

Powell, 1969). Thus the data would have served as a

systematic replication of this method for classroom settings.

As for the suggestion that self-recording serves as a poor

control device prior to behavior modification (Kanfer,

1970), the self-monitored data yielded the same ability

stratification as that given by teachers prior to the

study. Thus the self-recording served as a good control.

Just as McFall and Hammen (1971) used a sophisticated

design to isolate non-specific factors in smoking reduction,

the present study employed a design which would isolate non-

specific factors in self-recording through charts. This.

design permitted the most conservative test of self-monitoring

data up to the present since the recording on the chart was

contrasted to self-monitoring employed with relatively

non-reactive procedures (student recording on envelopes

after instruction by the teacher before the experiment

began) and to self-monitoring with reactive procedures

(students using the RCS after special instruction from

the experimenter). Thus the design recommends itself for

further use whenever behavioral devices are tested. A

simpler version of this design without the within-subjects

replication may be found in Azrin and Powell (1969).

::1:: .

The non-replication of self-monitoring effects for the

accounting work rate and error rate data was probably due

to the power of the design. Whereas prior research combined

self-monitoring with other reactive variables, this design

controlled for them. The results of mean comparisons for

the accounting work rates, i.e., the negative effects,

further demonstrate an instance of self-monitoring divorced

from the reactive variables. Therefore, it may be concluded

that the effects of self-monitoring are functionally related

to the immediate environmental context in which this

recording is accomplished.

As for the finding of persistent self-recording

effects lasting after this recording is terminated (McFall,

1970), both the data which did not reveal effects different

from baselines (the accounting work rate and error rate) and

the data which did reveal effects that were confounded

with other reactive variables (words per minute data)

indicated no lasting effects. In relationship to the

administratively scheduled events of exams and the termination

of the school year, the effects of self-monitoring were

shown to be evaporable.

The analysis of the group and individual data revealed

only a few instances of self-control. Where it was demon-

strated, the self-control was confounded with other

variables such as instructions and novelty. This finding

indicated that the controlling responses employed were

functionally related to the immediate environmental context.

Thus more powerful contingencies must be used and then later

faded for controlling responses if self-control is to

remain intact. The self-control acquired through the

recording responses used in this study was not sufficiently

buffered against the effects of examinations and the end of

the school year.

The use of the RCS and SBC as feedback devices for

knowledge of results showed only qualified results. Thus

the question of KOR as a reinforcer remains open. The error

rate data, which was gathered under delay conditions of a

day, pointed to the superiority of immediate feedback for

simple skill learning. A clear resolution to this question

may have to start in a more tightly controlled, laboratory

setting before clear generalizations to the school setting

may be drawn.


The implications of this research are twofold first,

with regard to future educational research and secondwith

regard to classroom practices. The impetus for this study

came from and was in part planned and conducted by four high

school teachers. Their active participation during work-

shops and in the initial planning insured a maximum of

benefits for each student during the course of the study

and at the same time permitted the conduct of this research

from escaping the ubiquitous demands of running a class:riom

on a daily basis. IThile communications problems still

existed, many difficulties listed in an earlier chapter

could have been remedied either by working with fewer

classrooms or by soliciting colleagues of the experimenter

to instruct the classes in the use of the recording forms.

Research conducted in this fashion has value for those most

involved in the educational process--teachers and students.

All of the teachers involved in this study noticed positive

changes in their students and thus planned to continue

self-monitoring in their future classes. Furthermore, the

teachers have expressed interest in doing further research

in their classes. Thus their participation will prove

mutually rewarding to both researcher and teacher.

Two additional remarks with regard to educational re-

search seem noteworthy because of events which failed to

occur. First, West (1969, P. 286) after a review of much

research on the point concluded that intensive practices of

timed typing incurs high error rates. This result was not

replicated under daily timed typings over a period of

twelve weeks. Perhaps the simultaneous self-monitoring of

both words per minute and error rates accounted for the


Prior research has limited the effectiveness of

behavior modification to low ability and special education

classrooms (Rosenfeld, 1972). The author reported a study

in which money was employed as a reinforcement for sixth-

grade students having a high I.Q. (above 110 on the Lorge

Thorndike) and money reinforcement plus stars on a publicly

located paper for average and high I.Q.'s (b06 or better)

showed significant achievement gains. Rosenfeld's findings,

which ran counter to prior research, are replicated here

because no significant Treatment X Ability interaction


The implications for classroom practices are directly

related to the analyses of the data. Even when the treat-

ments proved effective over pretreatment baselines, the

effects of these treatments did not effect the acceleration

of rates obtained at the end of the school year. While a

qualified demonstration of self-control was achieved, this

control passed to the occurrence of administratively

scheduled events. This finding suggests that, when possible,

testing or the deadline for large projects should be inter-

mittently scheduled to insure the longevity of student

self-control. Weighting daily work more and terminal work

rates less might have the same effects.

Another finding which holds implications was the fact

that rate measures may be used to achieve ability grouping.

While this measure has been used for such purposes in typing

class prior to this research, the replication of this finding

with a pinpoint as hetergeneously mixed as the accounting

work rate lends generality to the practice. Further study

with other types of curricula is necessary to insure this


Finally, as a result of inspecting some of individual

data, it was suggested that the satisfaction of criteria by

statistical techniques may not always be consistent with

significant behavior changes actually occurring. The sta-

tistical techniques may prove too time consuming, expensive,

and not immediate enough to satisfy the daily needs of

teachers and students. Data collection in the form of

rates does not suffer these drawbacks. Furthermore, when

rates are plotted on the SBC, long-range trends may be

assessed at a moment's glance and communication time is

commensurately reduced. These features alone urged a

number of students to remark that "really bad days" were

not taken as seriously as they were before the SEC was



The purpose of this study was to produce self-control

of academic behavior in high school business education

students by providing continuous, self-generated feedback

through various feedback devices and to assess the relative

merits of each device. This study was done to enlarge the

settings and vary the parameters under which the effects of

self-generated feedback have been found to be significant.

By using Skinner's (1953) definition of self-control, the

academic response rates were used as a controlled response

and the generation of feedback responses through the various

devices was used as the controlling response. For purposes

of this study, self-control is said to exist when celebrations

(acceleration or deceleration depending upon the nature of

the pinpointed responses) are better under treatment con-

ditions than under pretreatment baseline conditions.

The research was done in three classrooms. The first

was a ninth-grade, second semester typing class with

twenty-nine students. The second class was an eleventh-

grade, second semester accounting class with twenty-one

students. The last classroom was a tenth-grade, second

semester typing class with twenty-five students.

The response unit for the accounting class was any

fill-in-the-blank item, row or column entry, or any mathe-

matical computation. The dependent variable for this class

was work rate alone without respect to correct or incorrect

responding. Since the students had answer sheets and the

teacher available during the class for individual help, it

was felt that work rate alone was a sufficient pinpoint.

The response units for the typing classes were gross words

per minute and error rate for a three minute, once daily,

timed typing. The sample for typing was changed every day

during the experiment. Each student computed the total

number of words typed by using the vertical scale in his

text for each sample. The students then proofread their own

papersfor errors. Later, a student-aide or a neighbor

reread the papers to sift out any additional errors.,

during the treatment phases, error rates ":;re computel!!Bd a::i

day after the typing had been done for the class using the

aides and immediately for the class using the neighbor.

In all classes prior to the experiment, the students

were accustomed to counting the number of responses for

either the entire class period in the accounting class or

for the three minutes in the typing class and posting this

count on a manilla envelope used to hold all of their work

material. This type of self-recording was continued

throughout the experiment and contrasted with recording on a

rate computation form which lacked visual, graphic display

of the rates and Standard Behavior Chart which lent visual,

graphic display of the rates. During the first two days of

each treatment phase, the experimenter visited the classes

to teach the students the use of the new forms. At this

time, one-third of the students continued working and re-

cording as usual while the other two-thirds received

instruction on the rate computation sheet or the chart. This

was to control for any celebrations due to novelty due to the

presence of the experimenter, intraclass communication, or

teacher enthusiasm. After instruction in the forms, the

students were asked to use the forms daily for two weeks.

There was no differential reward given by the teachers for

using the forms. Since most of the students in these classes

were doing their daily work assignments, the reward was

thought superfluous and even contradictory to the goal of

self-control. Each treatment phase was interspersed with a

week's baseline condition before, between, and after the

three treatment-order combinations. Each treatment phase

was terminated by the teacher taking the forms from the

students' envelopes.

The three sets of data (accounting work rate for the

eleventh grade, typing rate, and error rate for the ninth

grade) were run independently through an analysis of

variance after appropriate transformations of the data

were performed. Since the tenth-grade typing class did not

experience the procedures as contended.. their data were not

analyzed in this fashion. Due to the data loss by subject

absenteeism, drop-out, and even one case of a student

complaining about her difficulty performing the operations

necessary to use the feedback devices, a least squares

solution was used to predict the missing scores. The

students were put into three ability levels according to

past class work and then randomly assigned to one of three

treatment orders. There were two subjects nested in each

Treatment X order X Ability crossing for the accounting

class and three students in the nesting for the typing class.

The main effects of treatments and ability were signifi-

cant for the accounting class ( p<.05 and p<.05, respec-

tively) and typing rate data (]p<.05 and p<.05 respectively).

These effects were not significant for the error rate data.

The Treatment X Order interaction was significant (p<.05)

for all sets of data. Using scheff6 comparisons, there was

a significant difference between the first six phases and

(p<.05) in the accounting and typing rate data. A comparison

of the first three pretreatment baselines and their treatments

reveals a difference in favor of the treatments (p<.05)

for the typing rate data. There were no significant

differences between the treatments themselves and their

posttreatment baselines (with the exceptions as noted above)

for the typing rate data. The reliability of the self-

monitored data was found to be acceptable for the error

rate data sampled.

The main conclusion to be drawn from the study is that

the students did evidence self-control for some of the

treatment conditions but not with regard to the last post-

treatment baseline. The Standard Behavior Chart while

serving as an accelerator of performance rates did no

better than the other devices. In the accounting class,

due to the presence of answer sheets and individual teacher

help, the devices alone were no better than these other

modes of feedback. The error rate data, which was gathered

under delay conditions of one day before feedback, confirms

the superiority of immediate feedback for simple skill

learning found by other researchers.

None of the treatment or baseline means were any better

than the last posttreatment baseline. In graphic form,

this behavior would when plotted cumulatively represent a

fixed interval scallop. Given the fact that the daily

classwork played little part in grade determination and that

this phase ended simultaneously with the end of the school

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs