Effects of contingent and non-contingent reinforcement on children's academic performance

MISSING IMAGE

Material Information

Title:
Effects of contingent and non-contingent reinforcement on children's academic performance
Physical Description:
iii, 150 leaves. : illus. ; 28 cm.
Language:
English
Creator:
Alvarez, Carlos M., 1944-
Publication Date:

Subjects

Subjects / Keywords:
Reinforcement (Psychology)   ( lcsh )
Psychology thesis Ph. D   ( lcsh )
Dissertations, Academic -- Psychology -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis -- University of Florida.
Bibliography:
Bibliography: leaves 145-149.
Statement of Responsibility:
by Carolos M. Alvarez.
General Note:
Typescript.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000580636
oclc - 14063448
notis - ADA8741
System ID:
AA00002223:00001

Full Text








EFFECTS OF CONTINGENT AND NON-CONTINGENT REINFORCEMENT ON
CHILDREN'S ACADEMIC PERFORMANCE
















By

CARLOS M. ALVAREZ


A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY













ACKNOWLEDGMENTS


I wish to express my most sincere gratitude to Dr.


James C.


Dixon,


Chairman, advisor and friend, whose patience,


support not only through the course of thi


guidance and continuous


investigation but also


throughout my graduate career have been invaluable and greatly


appreciated.


I also gratefully acknowledge Dr.


William D. Working,


Co-Chairman of my doctoral committee for hi


and critique of thi


help in the organization


manuscript


Special gratitude is extended to Dr.


Henry


Pennypacker for


the invaluable amount of teaching which served as an inspiration for


the present investigation.

are greatly appreciated.


interest, advice and creative suggestions


I am further indebted to the other members of


this


committee,


Dr. Ted Landsman and Dr.


Donald Avila


for their assist-


ance and criticism.


Special


thanks to my good friends Jose Valle and Alejo Vada for


long hours of hard work spent with me in the preparation and imple-


mentation of the pilot work that preceded this investigation.


Their


continuous encouragement and


incere testimony of friendship will


always


be remembered.


I would also


like to express my appreciation to Mr.


Phillip Alvers,


Principal of Moseley Elementary School, who facilitated the space and


- ~ .3-. .1 a .2 -2.. LL.. 1-1 -.-- 1 C --1.


1


I, r,,l,,









Finally, special


thanks to my wife Arminda, not only for her


direct participation in this research as co-experimenter


but also


for her encouragement and continuous support throughout my graduate

career.













TABLE


CONTENT


Page


ACKNOWLEDGMENTS.


ABSTRACT


CHAPTER


INTRODUCTION


a a a a aa a1


Review of
Statement


Literature


Problem.


.* a a a a a .19


METHOD


Expe


riment
riment


RESULTS.


. a .* . a . 3


Anal


Analysis


the Accuracy of
Ss' Productivity


Ss' Performanc


S S a a a a3


. . a a a 3


Sa a a a a a a a 44


UMMARY


APPENDIX


DESCRIPTION


NUMBER


SESSIONS


ATTENDED


S. a S 57


DIAGRAM
RELATIVE


EXPERIMENTAL


ITION


GETTING


E AND


SHOWING


S . a a S 5


BLUE


HEETS


CONTAINING


ARITHMETIC


PROBLEMS


YELLOW


SHEETS


CONTAINING ARITHMETIC


PROBLEMS


S . 64


DAILY


ACCURACY


RATIOS


EACH


S UNDER


EACH


EXPERIMENTAL


CONDITION


a a a a a . 67









ACCURACY RATIO AND PRODUCTIVITY CELEBRATIONS


FOR EACH


S UNDER EACH EXPERIMENTAL CONDITION


9 S 129


REFERENCES


BIOGRAPHICAL SKETCH.












stract


University


ssertation


Florida


Presented


Partial


Graduate


Fulfillment


Council


Requirements


Degree


Doctor of


Philosophy


EFFECTS


CONTINGENT


AND NON-CONTINGENT


REINFORCEMENT


CHILDREN


ACADEMIC


PERFORMANCE


Carlos


August,


Alvarez

1974


Chairman:


James


-Chairman:


Dixon


William


Working


Major


Department:


Psychology


present


investigation


children


s performance


in an academic


task


was s


tudied


under


types


environmental


arrangements


one in


which


consequences


were conditionally


related


their


performance


another


which


environmental


outcomes


were


admini


tered


independ-


ently


their


performance.


experiments


were


conducted.


Experiment


eight


first


-grade


children


were


individually


exposed


on a


daily


basi


to two different


experimental


conditions.


one o


them,


each


given


tokens


contingent


upon


correct


performance


on an arithmetic


task.


second


condition


each


S received


token


which


were


given


independent


of hi


performance


on the


same


task.


quality


performance was


measured


accuracy


responses


to the


arithmetic


problems


rate


improvement.


Productivity,


operationally


defined


as the


frequency


a ,. 4km a 4.4 a t.. n k 1 a .t44nmntnA r~ a In r a a- n n I- an ~ rI 1 a-~-~ a n a S S In -t S N


rr


r .A f


A U A LI 1


~LLnmnlnrl


I. A I.*U rl 1


~ A AIIIlIrInII


J- i


i


P^ n








session and rate of change in frequency over time.


during the performance could be


Tokens obtained


later exchanged for a variety of toys


and candies.


In Experiment II


seven first-grade children were individually


exposed on a daily basi


to two different experimental


conditions.


one of them,


tokens were delivered


contingent upon the


correct


responses on the arithmetic task, but withdrawal of tokens also occurred


everytime the


condition


basis,


that i


S made an incorrect response.


, delivery and withdrawal of token


In a second experimental

occurred on a non-contingent


, independently of the correctness of responses.


Experiment I,


tokens obtained during the performance could be


later


exchanged for a variety of toys and candi


Results


showed no


significant difference between


' average


evel


of accuracy under the contingent and non-contingent administration of


tokens

toward


Both experimental


conditions produced acceleration or a trend


improvement in accuracy in all


Ss participating in both experi-


ments with the exception


of one


S who showed deceleration in the


level


of accuracy over time under the non-contingent condition.


These findings then,

previously reported in the


failed to

literature,


show the dramatic difference

regarding the differential


effects produced by contingent and non-contingent reinforcement.


was argued, however,

probably accounts la


that generalization across experimental


rgely for the non-differential


conditions


effects found in this


study.








withdrawn


non-contingent upon


their


performance.


possibility


explaining


these


results


term


supers


titiou


responding was


explored.


It was


argued,


however,


that


future


replication


this


phenomenon


is necessary


before making


generalizations.













CHAPTER I

INTRODUCTION


One of the basic characteristic


of the work conducted with


instrumental


respon


ses,


beginning with


Thorndike and continuing with


Skinner and his


colleagues,


has been their emphasis


on the notion of


contingency.


For many years contingency has been considered necessary


for gaining control


over voluntary behavior.


Accordingly, contingency


became a fixture in experimental designs for operant conditioning

research, and has been a basic characteristic of the investigation of


schedule


of reinforcement conducted by


kinner and his colleagues


Contingency has been generally understood as a set of conditional


probabiliti


that is


, the probability that given one event another


event will occur


In instrumental


conditioning,


it is usually interpreted


as the probability that given a response,


reinforcement will


occur.


example, control


tingenci


over voluntary responses is obtained by arranging con-


between pre-selected responses and environmental consequences.


Implicit in the notion of contingency,


therefore,


is the prior emi


ssion


of a response in order for reinforcement to occur.


This


notion of contingency,


however, does not seem to explain all


the possible relations between response and outcomes to which organisms


are sensitive.


Environmental events can occur when no specific response


ha b noon orm iftforaF an^ +knh, ran aicn 2'Ffar+ nn..nnn hahaninrT


In Sh; r









relationships


between


response and


environmental


outcomes.


That


an inves


tigation


was conducted


tudy


some


possible


effect


produced


formance,


experimental


also


arrangement


compare


which


effects


independent


produced


per-


y arranging


contingency


between


performance


consequences.


this


investigation,


children


performance


an academic


task


was studied


under


two types


environmental


arrangements:


one in which


consequences


in which


were


conditionally


environmental


related


outcomes


were


their


admini


performance


stered


another


independently


their


performance.


Review


Literature


Until


common


followed


kinner


in most


reinforcement,


work


on schedule


reinforcement


that


intermittent


studied


every


response


was a


reinforcement,


response


necessary


efficient


condition


the delivery


reinforcement


Intermittent


reinforcement


dropped


sufficiency


condition


retained


necessity


rule


according


to which


the occurrence


respon


se was


till


required


procurement


reinforcement.


other word


whatever


intermittent


reinforcement


schedule was


effect,


based


on either


number


respon


or temporal


parameters,


reinforcement would


occur without


necessary


antecedent


response.


Intermittent


schedule


reinforcement


are usually


considered


the most


frequently


occurring


schedule


in the


natural


world


That


while continuous


reinforcement


can be


provided


quite


easily


laboratory,


unlikely


that


most


behaviors


could


reinforced








In non-laboratory situations behaviors are not only reinforced


on intermittent schedule


, but also,


there seem to be a number of out-


comes in the natural


environment that appear to occur independently


of the organism's responding. In

responding which usually may fail


Some instances,


for example, non-


to produce reinforcement


sometimes


produces it.


On such occasions there may be environmental arrangements


in operation that are quite different from the typical


schedules of


reinforcement discussed by Skinner,


in which responding is


a necessary


criterion for reinforcement to occur.


At thi


point,


it appears relevant to mention Seligman


Maier and


Solomon


s (1971) and Maier,


Seligman and Solomon's (1969) argument regard-


ing the independence between environmental


events and responses


and its


subsequent effects.


According to these authors,


traditional


learning


theorists have particularly emphasized the organism'


learning of two


relations between response and reinforcement, either that a certain


response produces reinforcement or that the response no


longer produces


the rewarding consequence.


However, a third relation can also be


learned


by the organism,


that there is no relationship between responding and out-


come


ligman, Maier and Solomon (1971)


learning that reinforcement


independent of the presence or absence of a response i


also an active


way of learning,


which in fact may interfere with


learning new relation-


ships


in subsequent


situations


This is the basi


for their theory of


"learned helplessness that attempt


passivity observed in some animal


to explain the phenomenon of


when facing trauma.









organism'


clock-sc


heduled


responding. Skinner

reinforcements were


(1948) co

delivered


inducted


a study


pigeon


in which


without


reference


on-going


behavior


at the


time


reinforcement


delivery.


Most


bird


in the


tudy


developed


a pronounced


stereotyped


pattern


of behavior,


kinner


labelled


particular


pattern


pattern


varying with


behavi or


individual


as superstitiousu


animal


behavior"


since,


according


to the experimenter,


this


kind


of behavior


occurred


as a


consequence


"accidental


conditioning.


Skinner


areas


oned


that


although


an effective


no experimental


"contingency"


"contingency"


could


been


inferred


establi


from


shed


observed


beforehand,


stereo-


typed


pattern


behavior.


kinner describes


process


"accidental


conditioning"


following manner:


conditioning


to be


sult
the


executing
it tends t


next


place,


process


some


usually


response


repeat


sentation


second


response


become


more


as the


response.


is not so great


"contingency"


still


further


probable.


obvious
hopper


If
that


is probable.
subsequent


is true


that


bird


ears;


interval


extinction


happens


as a re-


before


takes


This strengthen
reinforcement


some


responses


reinforced


not just


some


been


reinforcement


appear when


net result


is the


response
development


a cons


iderable


state


strength


-169)


However,


according


Lachter,


Co le


choenfeld


(1971)


ible


conditioning,


kinner,


to salvage


description


the concept


process


contingency


accidental


appealing


"(a)


fact


that


behavior


stream


continuous


to the


inference


that a


reinforced


even


when


applied


without


pre-selection


response


must


contingent


a a a a .1-


upon s


ome


response;


- .- -


..~J... a -


r, La


presumption
a r .-


that
.-L I- -


whatever
.LL








of its meaning, because every reinforcement

asserted as being contingent, or conversely,


said to be non-contingent [p


schedule must then be

that no schedule can be


. 233]."


To Schoenfeld,


Cole, Lang and Mankoff (1973)


"Regard 1


of how


'superstition


' demonstration is viewed it made two things clear:


first,


'contingency'


conventionally described as


'procurement'


and '


pro-


duction'


could be adequately


tated in temporal


terms alone; and second,


that operant conditioning was possible without it,


experimentally


pecified dependence of the temporal


that is, without an

distribution of Sr


(reinforcement) upon the temporal distribution of R (response)


The review of the


156.]"


literature that follows includes another aspect


of contingency


that may be relevant to thi


study


Contingency has


traditionally implied temporal


ing event,


continuity between response and reinforc-


temporal continguity meaning the relatively immediate occurrence


of reinforcement after the organism's response.


notion that immediate


reinforcement or temporal continuity i


a criti


al condition for


hanging


behavior has been set out explicit by Skinner (1953) and supported by


many other behaviorists who explain the way


subjects


learn when they


are classically or instrumentally conditioned by the principle of


"association by continguity."

lever-press and food pellet.


The rat

The dog


learns the connection between

learns the connection between the


bell and food.

According to the principle of contiguity we ought to find a


correspondence between each response and reinforcement;


that is,


each instrumental


resDonse


S-..u


irned by any oraanism must be re-








contiguity


appears


more


difficult


to apply


the case


schedules


intermittent


reinforcement.


However,


Mowrer


Jone


(1945)


offered


"response-unit


hypothesis


" that


seems


to reconcile


them.


Accord-


Mowrer


Jone


some


studied


past


have


hown


that


reinforcing


effects


reward


applied


not only


immediately


preceding


behavior


, but


also,


esser


degree,


behavior


that


are somewhat more


remote


from


reinforcement.


Therefore,


respon


se-unit


hypothesis


argued


that


sponse"


must


re-defined


include


the whole


sequence


behaviors


leading


reinforcement.


only


recently that


principle


temporal


contiguity


been


challenged


(Baum,


1973;


Bloomfield,


Rachlin,


1970)


as a


universal


cription


relations


between


environmental


stimuli


response


ability


reinforcements.


that


example,


instrumental


Rachlin


conditioning


(1970)


organism


suggest


learn


correlations


give


between


an illustration


rates


organism


response


learning


reinforcements.


correlations:


us cons


ider


that


a chamber with


sometimes


is pre


low rate,


food


comes


rate.


When


is pre


ssed


at a rapid


rate,


food


comes


rapid


rate.


notice


that


situation


kind,


rats


tend


ess


rapidly.


poss


ible


that


critical


feature


relation


ship


between


response


reinforcement


their


correlation


rate,


158].


Rachlin


lation


so makes


argues


tinction


that correlation


between


between


connection


response


corre-


reinforcement


"Let








reinforcement immediately.


reinforcements,


When responses are only correlated with


reinforcing events may be received between responses,


as well a


preceding or immediately after responses.


For example,


organism's


learning of a positive correlation between its responses


and reinforcement will


be reflected in an increased rate of responding


produced by increased reinforcement and vice versa when


learning a


negative correlation.

A study by Herrnstein and Hineline (1966) appears to show results


that favor the correlation hypothesis. This expe

by Rachlin as testing whether correlations can be


riment is also discussed


earned by organisms


when not accompanied by direct connections between responses and reinforce-


ments.


Herrnstein and Hineline found that rats


learned bar-pressing


responses


when not followed by any other consequence than a reduction


in the frequency of electric shock received.


In this experiment the


experimental design was arranged in such a way that there was not a

1 : 1 connection between responses and shocks, or between responses and


periods of no-shock.


In the experiment, rapid rates of bar pressing


responses were correlated with a


low rate of


hocks


and when rats


pressed the bar at a


low rate, shocks came at a rapid rate.


found that under these conditions,


Results


rats


obtained in other studies are al


The authors


learned to press the bar rapidly.


o interpreted by Rachlin


under the


light of the "correlation hypothesis


For example,


in a


series


of investigations conducted by Maier, Seligman and Solomon


and Seligman, Maier and Solomon


(1971) some dogs


(1969)


were exposed to electric





8


According to the "correlation hypothesis," responses emitted by the dogs


in the harnes


were completely uncorrelated with the rate of shock


received.


where


On the next day,


ignalled


the same dogs were put into a shuttle box


hocks were delivered and the dogs given the possibility


of escaping and avoiding them by jumping over a barrier.


Most of the


dogs previously exposed to independent


hock did not


learn to escape.


These dogs appeared to have given up and passively accepted the electric


hock


This group of dogs


differed significantly in their reaction to the


shuttle box from the dog


who had been previously exposed to escapable


hock.


Some of the passive dogs eventually learned to


escape shock, but


not until


after being dragged across the shutti


several


time


by the


experimenters.


Obviously,


prior exposure to inescapable


shock or shock


delivered independently of their behavior produced a strong interference


effect on the dog'


subsequent behavior in the shuttle box.


Previou


experience severely retarded and interfered with their


learning abilities


to escape and avoid shock


Rachl in


as mentioned before


, interprets


these results


in terms of the correlation hypothesis as follows:


dogs


learned in the initial condition a "zero


orrel ation" between


responding and environmental events.


Learning this zero correlation


however


interfered with their


learning of the new positive correlation


present in the shuttle box.


Rachlin explains further:


"The evidence provided by the Seligman,


Maier and Solomon experiment is that correlations, even when they are zero


correlations


learned (in the sense that they have potent effect on









experiment when shocks and responses


were uncorrelated was that there


were no connections


could have been


established.


learned?


If there were no connection


In terms of correlations,


what


however,


experiment is easy to explain, because zero correlations exist on the

same continuum as positive correlations. There is not reason why they


cannot be


learned just as easily as positive correlations.


Then,


zero correlation, already


learned,


interferes


with the effect of the


positive correlation imposed later [p.


165]."


The review of the


literature that follows


leaves the theoretical


aspects of contingencies and concentrates on reviewing empirical


evidence


collected in several


areas of research regarding the differential


effects


of administering positive contingent and non-contingent outcomes.


most of the following studies contingency vs.


non-contingency were


operationally defined in the way proposed by Schoenfeld and Farmer,


contingency mean


1970:


"that the distribution in time of R [response]


determines the distribution in time of reinforcements" while, on the


other hand "non-contingency means that the temporal


distribution of


reinforcements i


not determined by the temporal


distribution of re-


sponses [p.


First,


the empirical


evidence collected so far regarding organisms'


preference for either type of reinforcement appears fairly conclusive at


the present time.


Organisms in general


seem to prefer response-contingent


reward over response-independent reward.


For example, Jensen


(1963)


found that when 200 rats were given a choice between eating pellets


1]."








resumed


pressing


to obtain


food


during


the 40 minute


choice


period.


Neuringer


(1969)


Carder


Berkowitz


(1970)


also


found


that


animal


seem


to prefer


response


-contingent


food


over food


delivered


on a


non-contingent


ingh


(1970)


found


that


only


rats


also


children,


prefer


to work


to obtain


reward


even


when


having


poss


ibility


obtaining


free.


And,


more


recently,


Tarte


Collier


school

choice


(197


children

between


conducted


inve


obtaining


three


tigate

candie


experiments


behavioral


or pennie


with


pre-s


preference


taking


school


when

them


elementary


allowed


from a


or by pulling


lever


ults


tudy


confirmed


previous


findings.


They


showed


children'


high


preference


(77%)


earned


rewards


With


regard


to the


effect


non-contingent


itive


outcomes,


studies


such


from different


phenomena


areas


example,


research


in the


have


area


reported


inves


Internal


tigating


. External


Control


developed


Rotter


(1966)


we find


studies


investigating


effe


"random or non-contingent


reinforcement"


upon


expectancie


future


reinforcement.


According


Rotter'


(1954)


social


learning


theory


probability


behavior


occur


in a


particular


situation


a function


important


variables


first,


person


expectancy


that


given


behavior


will


instrumental


obtaining


available


reinforcement,


second,


value of


reinforcement


individual


function


behaving.


strengthen


Rotter


' expectancy


theory,


that a


reinforcement


particular


s main


behavior will


followed


that


reinforcement


in the


future.









the first condition


to be performed i


Ss are usually instructed that reward in the tasks

determined by their own skill, in other words, con-


tingent upon their performance.


In the second condition,


Ss are instructed


that obtaining reward is outside their control, that is dependent upon


chance, fate or powerful others.


The basi


hypothesis underlying these


experiments i


tingent upon hi


that if a person perceives a reinforcement as being con-


own behavior, the occurrence of positive or negative


reinforcement will strengthen or weaken the probability for that behavior


to occur again in the same or similar condition.


sees


reinforce-


ment


as being out of his control or non-contingent, then the behavior


preceding


reinforcement will have a lower probability to re-occur.


To test the above hypothesis


some studies were carried out comparing


verbal expectancies for future reinforcement under chance and skill con-


editions.


One of the first studies was that undertaken by Phares (1957) who


used an ambiguous color matching task and instructed half of the


Ss that


the task wa


so difficult that


success would be a matter of luck and


another half that success would be a matter of skill.


measured by the number of chips a


Expectancy was


S would be willing to bet on hi


probability of being correct on the succeeding trial.


Despite the fact


that both groups received the same number and sequence of reinforcements,

the skill conditions had greater effect on raising or lowering expectancies

for future reinforcements than chance conditions.

In a second study reported one year later, James and Rotter (1958)


investigated the effects of partial vs.


100% reinforcement on the extinction









it was reasoned that under chance conditions of 100% reinforcement,

S received no cues of non-reinforcement during the training series.

When extinction begins he utilizes the first non-reinforcements as


cues that the


situation has changed.


is perceived as external


there i


Since control of the


udden shift


situation


in expectancy and


extinction is rapid.


With partial reinforcement under chance conditions


S receives cues of non-reinforcement as well


as reinforcement.


When


extinction is begun non-reinforcement does not produce a new cue which


he can use to discriminate a change in


situations.


Consequently,


extinction is more gradual and more trial


are required before the


situation i


recategorized."


It 1i


interesting to note that


heffield


(1949) had already offered a


However,


similar interpretation of this phenomenon.


hers was primarily based upon stimulus-response concepts while


James and Rotter introduced the expectancy variable.


To investigate the validity of Rotter


social


learning theory


hypothesis


, James and Rotter (1958) used a simple but ambiguous card-


guessing task in which success


was totally controlled by the experimenter,


although it did not appear


o to the


Instructions were then used to


create


skill vs.


chance conditions


in the skill


condition were


instructed that success in the task depended primarily upon their own


kill


, while


in the chance condition were told that success in such


a task was entirely a matter of luck.


To summari


ze, results of the study


seemed to be in agreement with the hypothesis postulated by the authors


of the study.


Under the skill


condition,


the 50% reinforcement schedule









previous studies using partial


reinforcement:


100% reinforcement


schedule produced faster extinction than the 50% or partial


reinforce-


ment schedule.


on an


Expectancies of success were measured by


11-point scale.


Ss' own rating


Extinction was considered to occur when the


verbalized expectancy fell


to either 0 or


1 on three consecutive trial


Another experiment reported by Rotter


Liverant and Crowne (1961)


replicated James and Rotter's


findings without using different


instructions


to create chance vs.


kill


conditions.


In this


study,


Ss were exposed to


two different tasks which would be regarded as


skill


and chance-


controlled on the basi


of the


Ss' previous


cultural


experience.


skill-determined task involved a motor-skill apparatus


(Vertical


Aspiration Board) while the chance-determined task involved the card-


guessing procedure used by James and Rotter.


Results of thi


study


indicate that verbal


expectancies


showed 1


increments and decrements


under the situation where reinforcement was perceived by

non-contingent upon their behavior (chance condition).


greater fluctuations of verbal


Ss as being

ss showed much


expectancies under the skill condition.


This study also replicated James and Rotter


(1958)


findings regarding


the resistance to extinction offered by verbal expectancies from


exposed to


partial


reinforcement under chance vs.


kill


con-


editions


Other studies have also tested


at least


indirectly,


for any


differential effects produced by contingent vs.


non-contingent positive


outcomes.


For example,


Bandura and Perloff (1967) conducted a study with









group of children was exposed to an experimental


condition in which


had to set their own standards of behavior required to attain reinforce-


ment,


and also had to administer the rewards to themselves when reaching


their self-prescribed standard of performance (contingent reward).


second group of


children was exposed to another experimental


condition


that consisted of an externally controlled reward system.

condition reward was administered on a non-contingent basi


each child


In this


that is,


"inherited" a number of tokens before playing the game.


third group of children was treated as


a control


group


. Results of the


study


showed that both the self-contingently monitored (contingent re-


ward) and the externally controlled


non-contingent) reward conditions


were equally efficaciou


in inducing behavioral


productivity as compared


to the control


condition.


second study also using the


sel f-monitored


stem in one of the


conditions,


tested for the differential


effect upon behavioral


produc-


tivity of contingently administered vs


non-contingently imposed reward


In this study


, Liebert, Spiegler and Hall


(1970) required


Ss to set their


own standards and to turn a crank to achieve them.


in the contingent-


reward condition self-administered tokens whenever the


f-se


elected


standards were attained.


in the non-contingent reward condition re-


ceived the tokens


at equal


time intervals while playing,


being yoked to


in the


contingent condition with respect to density of reinforcement


and interval

of tokens (n


of time for delivery of reinforcement


lone,


low and high) was also an experimental


exchange value


variable manipu-








the control group, who recieved no tokens.


The number of responses


(cranks) emitted per session was directly related to the value of the


tokens,


in both experimental


groups.


However, an interesting result was


obtained despite the fact that


in the externally-controlled reward


system were explicitly told that they were being rewarded on a non-


contingent basis,


they,


too,


responded differentially to the different


value


assigned to the tokens.


Even more significant,


in the non-


contingent condition changed their standards


significantly more than


Ss in the contingent-reward condition.

interpreted these results as follows:


Liebert

"Thus, i


, Spiegler and Hall


t appear


(1970)


that non-


contingently rewarded


Ss may well


have changed their standard frequently


in an effort to "understand" or maximize the contingency, a phenomenon


that may have numerous counterparts in


for productivity are ambiguous


life situations where the criteria


246]."


One of the most typical methods of comparing the effects of con-


tingent vs


non-contingent positive events has been to shift, within the


same


schedule

by the


individual,


e


from a functioning contingent schedule to a non-contingent


and vice versa.


The effects on the rate of responding produced


shift are then carefully observed and measured


For example,


results most often reported in the animal and human


this method has been a decrease


literature when using


in the rate of responding when organisms


are exposed to non-contingent positive events as compared to the rate


maintained during the stage of contingent reinforcement.


A typical


study


in the human


literature is that conducted by Hart,


Reynolds


Baer, Brawley








upon cooperative play responses of a 5-year-old girl


in a pre-school


setting.


A baseline phase was recorded,


then non-contingent social


reinforcement from teachers wa


introduced.


After several


sessions of


receiving non-contingent social


attention and approval


teachers


began to administer the social


reinforcement only contingent upon


cooperative play or approximations to it.


Following evidence of


changes


in the rate of cooperative play behaviors


each condition was


discontinued and returned to a prior condition for the purpose of


demonstrating experimental


control


over the behavioral


changes.


Hart


et al.


(1968) reported data showing that non-contingent social attention


and approval did not appreciably develop cooperative play in the girl


However


, significant rate changes were observed everytime the social


reinforcement was made contingent upon cooperative play behavior.


Schoenfeld


, Cole,


Land and Mankoff


(1973) argue, however,


that


shifts from contingent to non-contingent schedule


usually produce an


of reinforcement


increased variability of responding since no constraint


is placed on the topography of the


behaviors when exposed to non-


contingent reinforcement as compared to


contingent reinforcement.


This


greater variability in behavior,


they believe, possibly accounts for the


greater reduction in rates of responding.


They also claim that a


lower rate of responding under non-contingent


reward i


not a universal


finding.


Other occasional


reports have been


made of indefinite maintenance


of ongoing rates of responding during


prolonged exposure to what the authors called non-contingent schedules









maintained by a non-contingent fixed interval 6-sec.


enforcement.


schedule of re-


Neuringer (1970) reinforced with grain the first three


pecks emitted by pigeons on a response-key.


pigeons


Neuringer found later that


continued to peck the response-key "superstitiously" throughout


50 experimental


subjects'


sessions in which grain was presented independent of the


behaviors.


The following


studies are


especially relevant to the present


investigation because they are concerned with children's academic

performance, although they deal only with the effects of contingent


reinforcement.


For example,


Kirby and Shields


(1972) reported increases


in the rate of


ratio


Ss' responses in arithmetic tasks when using a fixed


schedule of praise and immediate correctness feedback contingent


upon


performance.


They also reported a collateral


increase in


percentage of time


in attending behavior when the above procedure was


used.


Removal


of such treatment produced decreases


in both rate of


responses and attending behavior.


Hopkins, Schutte and Garton


(1971)


using an ABAB experimental design also found greater


effectiveness


using the procedure of allowing access to playground contingent upon


children


completion of their academic assignments as compared to a


baseline condition.


These authors reported the superiority of using


reinforcement contingent upon children's performance in producing increases


in work rates and trends towards fewer errors.


Lahey


McNees and Brown


(1973)


have al


so reported significant improvements


in children's reading


comprehension with the utilization of praise and pennie


made contingent








children


academic


performance.


Three


dependent


variables


were measured


by the


authors


this


investigation:


percentage


time


at work


work


output


per minute


accuracy


performance.


three


dependent


variable


showed


higher


increases


when


tangible


social


reinforce-


ments


were


admini


tered


contingent


upon


Ss' performances


as compared


a baseline


condition.


Other


studies


which


have


reported


similar


finding


when


using


contingent


reinforcement condition


include


Knapczyk


Living


ston


(197


Brigham,


Graubard


tans


(197


A few


studied


have


reported


diffential


effect


contingent


as well


as non-contingent


reward


upon


academic


performance.


Glynn


(1970),


example,


utili


four


different


groups


children


compare


effect


produced


elf-determined


experimenter-


determined


token


reinforcement


(contingent


conditions


chance-determined


token


enforcement


(non


-contingent


condition)


no token


treatment


(control


condition)


. Re


f thi


tudy


showed


that


non-contingent


con-


edition


produced


a level


accuracy


Ss' performance


that


was not


only


lower


than


those


produced


two contingent


conditions


also


lower


than


level


accuracy maintained


control


group


another


tudy,


Brigham


Finfrock,


Breunig


ushell


expose


school


kindergarten


children


three


different


reinforcement


con-


editions:


baseline


condition


which


no tokens


were


adminis


tered,


another


condition


which


token


were


admini


tered


on Ss


' correct


responses


finally


a condition


which


token


were


delivered


on a


non-c


ontingent


basi


. They


found


that


children


were more


accurate when


token


were









hand,


these authors reported that an analysis of the absolute rate of


responses--correct and incorrect responses--also showed that children

had nearly doubled their productivity during the non-contingent phase

as compared to the contingent phase.


%In sum,


then,


studies measuring the differential


effects produced


on academic and non-academic behaviors, by the contingent and non-con-


tingent administration of rewards have reported evidence of


significant


increases in rates of responding under contingent reinforcement con-


editions.


When both conditions of reinforcement were compared, contingent


reinforcement typically showed superiority over non-contingent reinforce-


ment in producing positive effects.


Furthermore, some studio


reported


decreases or deterioration of

contingent condition. Only o


ss' rates of responding under the non-


ne study reported that although contingent


reinforcement produced higher accuracy in


Ss' academic performance, an


analysis


of their absolute rate of


responding showed that


Ss nearly


doubled their work output under the non-contingent phase.


studies


reviewed utilized either reversal


or group-s


statistical design


in their


comparison of the effects of contingent and non-contingent reinforcement.

Statement of the Problem


It was the purpose of thi


effects on children'


investigation to test for any differential


academic performances produced by the exposure of


each child to two different experimental conditions:


external


one in which


consequences were arranged to be administered contingent upon


their responding on the academic task and another one in which outcomes








lines proposed by Schoenfeld and Farmer


1970).


That is,


in the


contingent condition,


the temporal


distribution of the experimentally


pre-specified responses determined the distribution in time of the


external consequences,


while in the non-contingent condition the temporal


distribution of outcomes was not determined by the temporal distribution

of responses.












CHAPTER II

METHOD


Two experiments were conducted.


grade children were


In Experiment


individually exposed on a daily basi


I a group of first-


to two different


experimental


conditions.


In one of them


each


was given tokens con-


tingent upon his correct performance on an arithmetic task.


condition each


In a second


S received tokens which were given independent of his


performance on the same task.


The quality of performance was


measured


by accuracy of responses to arithmetic problems and rate of improvement.

Productivity, operationally defined as the frequency of arithmetic


problems attempted per session regardless o

in terms of rates of problems completed per


f accuracy


sess


was also measured


ion and rate of change


in that frequency over time.


Tokens obtained during the performance


could be


later exchanged for a variety of toy


and candies.


In Experiment II another group of first-grade children was exposed


on a daily basis to two different experimental


conditions.


In one of them


tokens were delivered contingent upon the


arithmetic task


correct responses on the


, but withdrawal of tokens also occurred everytime the


S made an incorrect response.


In a second experimental


condition, delivery


and withdrawal of tokens occurred on a non-contingent basis,


that i


independently of the correctness of responses.


in Experiment I,


tokens


obtained during the performance could be


later exchanged for a variety of








investigation


a multiple


schedule


design


(Ferster


Skinner,


1957)


was adopted.


The multiple


schedule


sign


generally


cons


idered


more


sens


itive


individual


differences


than


ABAB


sing


a multiple


schedule,


finding


differential


respond-


are quite


powerful


experimental


result


particular


type


schedule,


being


differential


a product


rates


specific


responding


control


are more


produced


clearly


seen


different


timuli


schedule


reinforcement


used


each


component.


Two main


advantages


are generally


attributed


to the


multiple


schedule


esign,


however


First,


usually


quite


difficult


, especially when


using


human


same


, to employ


subject,


or more


as compared


experimental


to employing


conditions


ingle condition


on the


several


days.


Second,


there


always


poss


ibility


generalization


from


experimental


condition


to another.


During


pilot work


conducted


previous


to thi


inve


tigation,


interesting


phenomenon was


erved.


Children


generally


evidence of


differential


performance when


exposed


to both


contingent


non-contingent


reward


condition


They


seemed


not to d


criminate


between


experimental


conditions


furthermore,


they


appeared


to work


hard


under


both


types


arrangements.


Several


environmental


stimuli


were


then


systematically manipulated


throughout


pilot work


later


introduced


inve


stigation,


especially


those


that


appeared


to facilitate


process


crimi-


nation.


Environmental


timuli


were


made


quite


different


both








environmental


variabi


were, of course, counterbalanced among


to control


for their possible direct effects.


Finally, each


S had


a prolonged exposure to both experimental conditions that was also

expected to facilitate discrimination.


These procedures were used in Experiments I an

facilitate the process of discrimination even further


d II; however,


another variable


was introduced in Experiment II:


withdrawal of rewards.


Introduction


of the withdrawal of reinforcement procedure was


also made in response


to the findings of the pilot work that the administration of contingent

vs. non-contingent positive events alone did not seem to produce any

significant difference in the way children performed under both


experimental


conditions.


The withdrawal of reinforcement procedure commonly known as


"response


cost" ha


been extensively used in


laboratory research with


human


Ss (Weiner,


1962;


1963


1964;


1965a;


1965b;


1969)


This procedure


has also proved to be a powerful


contingent upon


one in controlling behavior when used


Ss' behaviors outside the laboratory and in therapeutic


situations


, Boren and Colman,


1970;


Kazdin


, 1971; Winkler,


1970)


Phillips,


Phillips,


Fixen and Wolf


(1971) for example, contrasted


the relative efficacy of token reinforcement,

cedures combined with predelinquent boys. Th


response cost and both pro-


ese experimenters found that


the administration of positive .reinforcement plus a response cost procedure


resulted in a greater percentage of correct response


on a quiz adminis-


tered after watching a


TV news program.


In another study,


Schmauk (1970)








Experiment I


Subjects


Eight


students


(4 boys and 4 girl


selected by the teacher from a


first grade


ass


at Moseley Elementary


school,


Palatka, Florida, served


as subjects.


Ss were between 6 and 7 years old


Appendix


A for


further description


Materials


Two distinctively different rooms at Moseley Elementary School were


used for the


experiment.


In each room there was an armchair where the


could


it on on


side of a wooden screen.


On the other


ide of the


screen


there was a table containing a carousel


projector, a whistle and a bell


that were used by the.experimenter (E)


The screen did not allow


s to


fully


see E on the other side of the


creen but


could observe the


written performance from the other side (see Appendi


The carousel


projector contained slide


with numbers 0 through 20


number of tokens to be collected by


slide


, equivalent to the


were projected on a


screen placed in front of


Next to the


s chair a box


was placed con-


training a


large number of tokens that could be picked up by


S at the end


of hi


performance


depending on the


last number proj


ected on the screen.


Also, on the floor next to the


chair was


a pile of


children


comics.


sets


of sheets of paper containing


simple counting tasks were


used.


heets of one set were blue and the other set wa


yellow.


On top


of the blue


circle


heet


s, a circle appeared at each corner.


were 5 different arithmetic problems.


Below each of the


Each problem


consisted of


,, _'IL-...~_L-_I


rr r 1


r .r II. I I


n.. II~ -1. L I~ c L




25



different blue sheets containing a total of 20 problems were presented

to every child during each session (see Appendix C).


The yellow


sheets were identical


in format except that a


square


appeared at each corner instead of a circle.


On the right hand


ide of


each problem four numbers appeared in column.


One of the numbers represented


the correct number of squares contained in that specific problem.


different yellow


sheets containing a total


0 problems were presented


to every child during each session (see Appendix D).


To control


and blue


for the


level of difficulty of problems on both yellow


sheets, problems appearing on the two different color


sheets


contained equivalent numbers of circles and


quares.


Also,


the alterna-


tive answers were the same for each type of


heet.


Procedure


Preparatory stage:


For three consecutive day


, each child was brought to the rooms by


each experimenter to get acquainted with the task,


tokens and prizes.


Each child was primarily familiarized with the way in which numbers were


projected on the screen in front of him.


The procedure to follow was


explained to him by


E in the following manner:


Everytime he came to the


room he would have to sit in the armchair


and at the sound of the whistle


he would begin to work on the problems placed in front of him.


procedure of counting the number of circles or squares in each problem,


depending on the sheet presented, was also explained to him.


explained that it would be his task to choose the number of circles or









problems


on the


right


hand


column.


case


completing


one sheet,


could


continue working


on the


second


heet


available on


desk.


During


three days


trial


period,


each


child wa


hown


procedure


obtaining


tokens


that


could


exchanged


later


toys


or candi


explained


to S that


during


performance,


number


on the


screen would


change-


-the


noise


carouse


projector


bell


ringing would


indicate


that


higher


number


token


was projected


screen.


Numbers


projected


on the


screen


would


represent


number


tokens


he was


getting.


E explained


S that


minutes


would


hear


the whi


again


at that


time


he would


have


to stop


working.


Then,


he would


pick


number


tokens


equivalent


to the


last


number


hown


on the


screen


This


procedure was


practiced


several


times


during


three


preparation


experiment.


During


trial


period,


E let


child


exchange


token


a variety


toys


candie


different


token


-prices,


attempting


establi


tokens


as conditioned


reinforcers


child.


explained


S the


purpose


comic


on the


floor


next


to him,


that


time


felt


tired,


could


pick


comics


read


them a


he wi


Experimental


procedure:


Every


each


conducted


one o


rooms


have


minute


sess


with


one of


first


sess


with


one of


after


collecting


token


S went


second


room


to work


with


second


E for


more minutes.


A period


time


ranging









In one of the rooms, one of the


Es administered the contingent


condition.


That i


after the sound of the whistle


E added one point on


the screen paired with the sound of the bell


ringing at the end of every


correct choice made to a problem.


At the end of


minutes


E sounded the


whistle again indicating that the


sess


ion was over.


In another room, a second


E administered the non-contingent con-


edition.


That is, after the sound of a whistle


on the screen paired with the bell


E began adding one point


ring, according to a time


schedule


prepared in advance of the


sess


This schedule was


independent of


performance on this


task.


The number of points projected on the


screen,


to be exchanged later for tokens, was presented according to the


number of tokens obtained by


S during the alternate session of contingent


delivery of tokens.


Both


experimental


conditions were presented daily and their order


of presentation alternated,


to control


for any effects produced by the


order in which they were presented to


stimuli


Also,


Es and environmental


such as rooms and their corresponding set of task sheets


or yellow) were counterbalanced among all


Ss to control


(blue


for their possible


effects.


At the end of the two daily sessions


S was allowed to exchange the


tokens for hi


choice of toys


and candies.


The number of days exposed to


the experimental


conditions varied with each child.


Although the study


was conducted for a total


of 21


school


days, children were absent from


school


some days during this period.







Experiment II


Subjects


Eight students (4 boys and 4 girls) were originally selected by


their teacher and assigned to this experiment.


However


, one of the girls


was ill


14 days


out of the 21


days


in which the experiment was conducted


, consequently,


the data on her performance were


later discarded from


the final


results.


for this experiment were students from a first


grade class at Moseley Elementary School,


Palatka, Florida.


ages


ranged between 6 and 7 years old.


in Experiment II were different from


those


Ss participating in Experiment


ee Appendix A for further


description of


ss).


Material


The same two rooms and equipment used in Experiment I were also used


for this experiment.


The same two


Es were involved in this experiment,


Only a new instrument was


introduced


"cricket" wa


utilized dur-


ing the


sess


ions to announce the


losing of points projected on the screen.


Two sets of sheets,


identical to those used in Experiment I, were


also used in this experiment.

Procedure


Preparatory


stage:


The first three day


of the preparatory stage were conducted in the


same way a


in Experiment I, except that in this experiment


acquainted with the procedure used by


Ss were also


E to communicate to them the


loss of


a point or token.


Loss of a token was


indicated by a change in the number


projected on the screen (to a


lower number) paired with the sound of the








a change in the number projected on the screen (to a higher number

the sound of the bell.


Experimental


procedure


Every day each


the experimental


S was conducted to one of the rooms


conditions was administered.


where one of


At the end of the first


sess


ion in that room S was conducted to a second room for the adminis-


traction of a second experimental condition.


A period of time ranging


between 5 and 10 minutes usually elapsed between one


sess


ion and the


other.


In one of the rooms, one of the


Es administered the contingent


condition


After


signaling the beginning of the session by using the


sound of the whistle the experimenter began adding one point to the


screen paired with th


In this experime


sound of the bell

nt, however, E al


at the end of every choice


so took away one point (pro-


jected on the screen and paired with the sound of a cricket) every time


an incorrect choice occurred.


At the end of the 2-minute session the


E sounded the whistle indicating that the


session was over.


The same


type of sheets were used for each


S throughout his exposure to this


experimental


condition.


At the end of the


session


S was allowed to


pick up the number of tokens equivalent to the points projected on the

screen.


When conducted to the other room, a second


E administered the non-


contingent condition.


That i


after the initial


signal of the whistle


E began adding one point paired with the sound of the bell


or taking away








initial


signal


E whistled


again


indicating .the


sess


ion.


same


type


heet was


used


on a


daily


basi


condition


sess


ion,


allowed


again


to collect


token


equiv-


alent


to the


last


number


projected


on the


screen


. At


the end


two daily


sess


ions


was allowed


exchange


tokens


obtained


both


conditions


toys


or candies.


Both


experimental


condition


were


ented


daily


as i


Experiment


their order of


presentation


alternated.


Es and


environmental


timuli


were


counterbalanced


among


. At












CHAPTER III

RESULTS


The differential effects produced by the experimental


were measured on two dependent variabi


daily performance on the arithmetic task.


conditions


accuracy and productivity of


The basic measure of


accuracy utilized in this investigation was the Accuracy Ratio.


Accuracy Ratio is defined as the ratio between frequency of correct


responses and frequency of incorrect responses


(Pennypacker


, Koenig and


Lindsley,


1972).


The productivity of S was measured by the daily


frequency of arithmetic problems attempted per session,


the accuracy of his performance.


Since information regarding proportional


changes of a behavior is


valuable,


the accuracy ratios and frequency


of problems completed per


sess


ion were plotted on Standard Behavior Charts


(Pennypacker


et al.,


1972).


Figures


through 30 (Appendix E) show graphically the daily


accuracy ratios for each


S under each experimental condition.


Figures


through 60 (Appendix F) present graphically the daily frequency of


problems completed by each


S during each experimental


plotting behavior frequency data on a


sess


logarithmic scale a


representation of proportional rather

frequencies can be obtained (Koenig,


!r than absolute changes in behavior

1972).


In nrder to facilitate an understanding of the results nresented


regard


linear graphic








Standard


Behavior


Chart


as well


as some


procedures


used


analy


Standard


Behavior


data.

r Chart


standard


Behavior


Chart


actually


logarithmic


chart,


whose


vertical


scale


include


logarithmic


cycles


vertical


dimension


frequencies


standard


or rates


Behavior


respon


Chart


ses.


represent


unit


scale


of measurement


movement


or responses


per minute.


hori


zontal


scale


represents


time


scale


that


runs


140 consecutive


or a


total


calendar weeks

Celeration


means


ure of behavior


change


over time.


Frequen


ratios


played


on the


Standard


Behavior


Chart


usually


either


accelerating


or decelerating


as time


passes.


term


cel eration


used


to de


cribe


acceleration


or deceleration


relations


hips.


Celeration


actually


measure


change


frequency


responding


that


occurs


within


a week


s period


time.


celebration


coefficient


ratio


value


that


cribes


slope


line of best


calculated


series


frequencies.


line of


best


obtained


using


squares


method


regrets


the most


important


character


celebration


that


provide


measure


of behavior


change which


independent


initial


rate of behavior,


so that


gives


a measure of


improvement without


reference


initial


accuracy


or s


peed.









across


Ss the celebrating effects on accuracy ratios produced by


different experimental


arrangements.


Measuring celebrating effects


rather than the absolute values of changes


in performance facilitates


comparison among


Ss allowing for individual differences


in levels of


performance.


Figures 61


through 75 (Appendi


G) graphically show on


left side of each figure the accuracy ratio celebrations


for each


S under both experimental


conditions in Experiments


I and II.


Productivity Celeration


is a measure of change in productivity or frequency of


problems attempted that occurs over time.


Figures 61


through 75


(Appendi


how graphically on the right side of each figure the


productivity celebrations for each


conditions in Experiments


S under the two experimental


I and II.


Analysis of the Accuracy of


Performance


Table


shows the geometric means* and standard deviations of the


accuracy ratio


per experimental


condition for each


S in Experiment I.


A matched t-test computed on the difference between the geometric means


in the two experimental conditions shows a non-significant differ-


ence (


.31)


Delivery of tokens contingent upon the


correct


responses


did not seem to affect significantly the average performance


Ss as compared to the condition in which tokens were delivered on a


non-contingent basis.


*The geometric mean i


fl't ?% C. S Iff C.


considered to be the proper average for any ratio


aL a- -C -C


; n n n I L~ n m n ~ ~ i I u Ar. 1 ir h rl







34











as r-. aD cM co
to (I I-






en qt CM








C I-. I-I


OC,

LU--S t
IC t 0N
I-US



zw w wD

CI trn S (C)



rx LUCQ





S S3



ZC,.



CI- %,

(n O CM C t
Co -S I-r C


OW
-0.


S

CI- tO C Mxco


C-Dact CO c


C') I-
1+Z ccl 4-'~
cC aE 9- a
Lt U)' U) U,
S: 0)) Li 0)
4-']c C3 C1 4r C
0 Cr c
I-. C) .4-. C P4
cx cx, o- aLU
I- C 0 C a 0 e
Lii 0r 0 0 r









Table


shows the geometric means and standard deviations of


the accuracy ratios per experimental


condition for each


S in Experiment


A matched t-test


shows a non


-significant difference between the


geometric means of the accuracy ratios produced by the condition in

which delivery and withdrawal of tokens were made contingent upon

ss' performance and those produced by the other experimental condition


in which delivery and withdrawal were made non-contingent upon


behavior (t =-


.42)


A matched


t-test computed for all


Ss from both experiments also


showed a non-significant difference


.05).


Results appear to


indicate that administration of tokens


contingent basi


accuracy perform


Finally


the over-all average 1

Experiment I and those


on a contingent and on a non-


did not seem to affect differently the average

nce of Ss.


a matched t-test was computed on the difference between


evel


of accuracy obtained by


obtained by


in Experiment II


participating in

The t-test


showed a statistically significant difference between both groups


75; p


< .02,


two tailed;


= 28)


That i


in Experiment I showed


on the average a significantly higher


level


of accuracy in their per-


formance than


in Experiment II


Table


shows the Accuracy Ratio Celeration


for each


S under each


experimental


condition


in Experiment I.


A matched


t-test computed on the


log of the celebrations produced by the two experimental conditions


showed


a non-significant difference (t


That is,


delivery of tokens



















































OLii
CO
n U
o w






37
















cr i






o N.



*) S

I- -





CL





o 10 I
U- 'CrY
o 0J
o c r)r



00 U)

o '

I-I
ccl
UzuJ
Co -U t
I- LUI'(V
w 0
LU -l



OC



o- c' c
ct
0:
'C '
00 C \

0<






I-


'C
r I VS
Q"c



rS f
Ir 0 )

C
C 0
0 C\J
-Li









performances.


also


important


to note


that


accelerating


effects


occurred


cases


under


contingent


non-contingent


conditions.


Table 4


hows


Accuracy


Rati o


Celeration


each


S under


two experimental


condition


Experiment


This


Table


shows


that


accelerating


effects


were


found


cases,


except


S #10 who


showed


deceleration


accuracy


under


non-contingent


condition.


seven


showed


greater


celebration


on accuracy


under


contingent


condition.


However


a matched


t-test


computed


on the


log of


the celera-


tions


under both


experimental


condition


showed


that


difference


cel erations


delivery


states


tically


withdrawal


non-significant


token


contingent


91).


upon


That


correct


correct


greater


responses


celebrating


to the


effect


arithmetic


on the


task


accuracy


produce


significantly


Ss' performance when


compared


to that


produced


contingent


delivery


Also


withdrawal


Experiment


most


token


Ss (with


on a non-


one exception)


showed


acceleration


in the


accuracy


their


performances


under


both


experimental


conditions.


However,


when


a matched


was conducted


on the


accuracy


ratio


celebration


produced


contingent


non-contingent


admini


traction


tokens,


from


Experiment


combined,


a stati


stically


significant


difference


was obtained


tailed


--= 14)


favor' of


contingent


conditions.


Analysi


Ss' Productivity


Table


geometric


means


standard


deviations






























































LU HJ

0 C=
-LIs ..
oc.t


I-LU
cC
r::


0 CO


04
do







04




















































.- I-


I-I-.









computed on the geometric means of problems attempted under the


experimental


upon


condition in which delivery of tokens was made contingent


Ss' performance and that obtained for the condition in which tokens


were delivered non-contingently, was non-significant


.39).


That


the delivery of tokens contingent upon


ss' accuracy of performance


the non-contingent delivery of tokens did not seem to differentially


affect the average


level


productivity


Table 6


shows the geometric means and standard deviations of the


productivity of


ss under each experimental


In one of the experimental


condition in Experiment II.


conditions in Experiment II,


tokens were


delivered and withdrawn from


Ss contingent upon the accuracy of their


performance,


in another condition the delivery and withdrawal


was made on a non-contingent basis.


A matched


of tokens


t-test on the average


productivity under both experimental conditions also showed a non-


significant difference (


.79).


Table 7 shows the productivity celebrations for each.


S under each


experimental condition in Experiment I


A matched t-test on the


log of


the celebrations of the two experimental conditions was non-significant


.25)


That is,


the delivery of tokens contingent upon


correct


responses did not appear to affect the


productivity celebration differ-


ently from the delivery of tokens independent of the


Table 8 shows the productivity celebrations for each


performance


S under the


two experimental


conditions


in Experiment II


Surprisingly,


out of


seven


Ss s


showed greater acceleration of their productivity under the





42














U) In ci. LI) CM ci-
C" c~)







LU~ C') C')





t-


0; 10L LA eN CMcM
o (Y) C")I-
0
0o



F- 4-' 4-
O r U dFt
C) CM 0) 0) CM CM CM
U) =P r

LU cC(V9



I- *

LU CM r r
00..


~I F
I- c



<0




su-E



en er, *



(DL CO'rC I-



O Cr C '
0) -0 0 ci )
Cr: En C Ct En C
F- 'C 0 cc a C
LU C.) 0 0
A-) -






43










Cr)









I- -







CCO





I-


o CM
Li a o


o 0u


0l =
W x A
o M


c) -- 0Y




-I-- cn
0: 0
cc CUH C



o a

>-- 0


orr

HG r .

XS C





a a
a a)

a






C-
42 Ir


i-)





























































-I -





45



celebrations produced under the contingent compared to the non-contingent


experimental


condition showed a


, two tailed; d.f.


= 6).


significant difference (


= 3.


That is, delivery and withdrawal


of tokens


non-contingently upon


acceleration in the produce


performance produced a significantly higher

tivity of Ss as compared to that produced by


the delivery and withdrawal


of tokens contingent upon


Ss' academic


performance.













CHAPTER


Results


obtained


this


investigation


regarding


differen-


tial


effects


produced


contingent and


non-contingent


admini


traction


tokens


station.


on children


academic


Beginning with


performance


effect


require


on accuracy


cautiou


performance


interpre-

e. Table


shows


no s


uperiority


the contingent


over


non-contingent


condition


Indeed


of 8


Experiment


achieved


higher


(absolute)


accuracy


under


non-contingent


condition.


Table


which


shows


same


kind


data


under


response


cost


procedure,


hows


no significant


difference


in the


average


level


accuracy


produced


y contingent


non-contingent


delivery


and withdrawal


tokens.


These


findings


failed


support most


findings


literature


which


have


generally


found


uperiority


contingent


reinforcement


[Chadwick


Day,


1971


Ki rby


shields


(197


Knapc


Livings


(197


Hopkins


Schutte


Garton


1971);


Brigham,


Graubard


Stans


(197


La hey


McNees


Brown


(1973)]


Turning


from mean


accuracy


ratios


accuracy


ratio


celebrations,


a means


ure of


change


over


time,


Tables


show


that


both


experimental


condition


produced


acceleration


or trend


towards


improvement


in accuracy


ss participating


both


experiments,


except


one case.


Only


.


..









withdrawal


of tokens.


Also, when comparing the rates of celebration pro-


duced by contingency and non-contingency conditions for each experiment,


the differences were


statistically non-significant except when comparing


the differential


effects produced in all


Ss participating in Experiments


I and II.


That i


when all


ss from both experiments were combined,


contingent administration of environmental consequences produced an


over all


higher rate of improvement in accuracy over time as compared


to that produced by non-contingent administration


Results of this investigation then,


failed to show the dramatic


difference in effects produced by contingent and non-contingent reinforce-


ment previously reported by Glynn


(1970)


Brigham


et al.


Hart


et at.


(1968).


Two particular characteristics of the method employed in this

estigation deserve careful evaluation especially when comparing present


findings to previous ones.


First of all,


the task involved the presentation


of the same


experiment.


set of problems over and over again throughout the whole


s likely that this procedure practically insured that


a certain degree of improvement would take place under any conditions


just as a consequence of the


Ss' familiarity with


the task and repeated


practice with the same problems.


Second, most studies comparing the


differential effects of contingent and non-contingent reinforcement have


either exposed different groups of

conditions, or have used the revers


Ss to the different experimental

al procedure of shifting within the


same


S from a functioning contingent


schedule to a non-contingent








both experimental conditions on a daily basis.


It is possible that,


although special care was taken to make stimuli under the two experi-


mental conditions quite different from each other,


the tasks remained


similar in nature, especially when contrasted with their other activities,


and were presented quite close in time.


very likely therefore,


that


some generalization across experimental


conditions occurred.


Finally, a comparison between the overall


average


levels of accuracy


achieved by


in Experiments I and II


showed that the group of


Ss who


participated in Experiment I performed with a


significantly higher


level


of accuracy than


in Experiment II.


Although the difference may be


attributed to the procedure of token administration,


it is also possible


that the difference in accuracy may reflect the initial differences in


performance


level of


in the two experiments.


from both groups


seemed to differ in their


levels of performance from the beginning of


the experiment.


In fact, 5 out of 8


ss participating in Experiment I


showed during their initial


performance accuracy ratios


equal


to or above


1, as compared to only


out of 7


Ss participating in Experiment II.


However, a t-test of the difference between the initial


accuracy of both groups, defined as


level of


performance during the first


three days,


failed to achieve statistical


significance (


--= 1.55)


While the accuracy ratio emphasized correct responses in relation


to incorrect ones


, the productivity measure is a more direct measure of


output.


An analysis


of the mean productivity


hown on


Tables 5 and 6


indicate that the average


level of children'


productivity under both








Ss participating in both experiments increased their output over time


under both experimental


conditions.


Previous studies had consistently


reported increases


in work output when children were reinforced con-


tingently upon different aspects of their academic performances


Hopkins,


Schutte


and Garton,


1971


Brigham, Graubard and Stan


, 1972).


These investigations, however


did not attempt to compare productivity


under contingent and non-contingent reinforcement.


findings then,


Part of the present


seem to be consistent with previously reported ones,


since an increase in productivity over time was a1


o found when children


were exposed to contingent reward


The most interesting results are found, however, when a comparison


is made between the rates


of increase in productivity shown by


Ss under


both


experimental


conditions.


Contrary to expectations, a test of


significance


showed that the non-contingent delivery and withdrawal


of tokens accelerated the rate of


Ss' productivity


significantly more


than their administration on a contingent bast


difference was


statistically significant only when the response cost procedure was


used in Experiment II


Although thi


last result appears difficult to interpret


should be noted that a previous study (Brigham, Finfrock, Breunig and


Bushell


, 1972), primarily interested in comparing the differential


effects of different reinforcement contingency


on the accuracy of


children's academic performances,


reports a


similar finding.


They


report that an analysis of the absolute rate of academic responses--


, it




50



which a non-contingent procedure of reinforcement was used, compared

to the phase in which a contingent procedure was used.

It is possible to interpret these results in terms of adventitious


reinforcement.


That is,


token


administered non-contingently on


accuracy of performance may have adventitiously reinforced their rate


of productivity


Liebert,


piegler and Hall


(1970) made a similar


interpretation of their findings in which


Ss who were non-contingently


rewarded changed their standards to obtain reinforcement significantly


more frequently than


Ss receiving reward contingent upon the appropriate


motor responses.


Theyargued that their findings


could be interpreted


in terms of superstitious responding, and furthermore


, using a cognitive


description of the results,


they interpreted


Ss' behaviors in the non-


contingent condition as possibly representing an effort on their part


to understand or maximi


contingency.


In regard to this


last interpretation,


it may be relevant to


the question of the possible influence played by past history of


reinforcement on the behaviors exhibited by


in thes


e investigations.


For example,


it is the impression of the experimenters participating


in this study that children generally came to the experimental


sess


ions


with the strong expectation that obtaining token


during their per-


formance would depend on how they did in the a


signed task.


Therefore,


based upon the assumption that many of the children


school


experiences in


frequently involve getting reward according to their academic


performances,


it i


possible to infer that such history of reinforcement





51



factor which interfered with the process of discrimination, especially


since the present experiment took place in a school


setting and also


involved an academic


task.


Future research should attempt to control


for the effects of past history of reinforcement of

ing to contingent and non-contingent environmental


Certain factors obviously


Ss on their respond-

arrangements.


limit the generalization of the present


findings.


First,


limited number of


ss participating in both


expert -


ments suggests the need for replication


Second,


length of time


that


Ss were exposed to the experimental


conditions may not have been


sufficient to insure appropriate discrimination between contingency and


non-contingency.


Further research


should investigate the extent to which


these findings are determined by the experimental designs utilized,


reversal


procedure vs.


multiple


schedule design.


choenfeld


(1973)


for example,


have claimed that shifts


from contingent


to non-contingent


chedul


of reinforcement (as


in reversal


design


usually produce an increased variability of response since no constraint


is placed on the topography of the


behavior when exposed to non-


contingent reinforcement.


They argued,


then,


that thi


greater


variability in behavior possibly accounts for the greater reduction

in rates of responding reported by other investigations.

Future research should also be addressed to the question of


generalization across conditions


when each


S is exposed to both con-


tingent and non-contingent environmental arrangements.


Procedures that


may facilitate the S's


discrimination of the experimental


conditions








under concurrent


schedules


involving


the procedures


proposed


Findley


(1958,


Findley


proposed


that


switching


behavior


conditions


switching


in a concurrent


schedule


need


to be more


explicit


more


subject


to the


experimenter'


control


pro-


cedures


not only


facilitate


the discriminative


process


could


also


provide


information


about


preference


a given


schedule


arrangement.













CHAPTER V

SUMMARY


In the present study,


it was argued that the traditional


notion


of contingency does not seem to explain all


the possible relation


between response and outcomes to which organisms are sensitive.


may also be outcomes when no


There


specific response has been emitted and


they can al


aimed at


so affect on-going behavior.


studying thi


The present investigation then,


latter dimension of the relationships between


response and outcome.


That is,


the effects produced by the administra-


tion of tokens independent of


Ss' behaviors were compared to the effects


produced by arranging a contingency between their performance and obtain-


ing tokens.


For this purpose,


children's performance in an academic task


was studied under two types of environmental arrangements:


the receipt of token


one in which


was conditionally related to their performance,


and another in which the tokens were received independent of their

performance.


A review of the


literature suggested the superiority of administer-


ing reward contingent upon responses over their non-contingent admini


tration,


to produce positive increases


in behavior, especially in higher


level


of academic performance.


Some


studies even reported decreases or


deterioration of


ss' behaviors when exposed to non-contingent reward.


c hn.,nrI cnmntahh+ m rtnnovn n rea+ i Ann nPC 1 -ham


nn iv +hrPO C+lrr(ioC


nna nC+ham








externally


imposed


reward


were


equally


efficacious


inducing


pro-


ductivity


among


A second


study


reported


that


Ss changed


their


standards


obtain


reinforcemen t


significantly more


frequently when


rewarded

to a self


on an externally


-monitored


based


schedule.


schedule


Finally,


than


third


when


rewarded


study


report


according

ed that


although

academic


contingent r

performance,


enforcementt


an analyst


produced


their


higher


solut


accuracy

e rates


children


responding


showed


that


Ss nearly


doubled


their


productivity


during


non-


contingent


phase of


experiment.


sent


inves


tigation


cons


two separate


experiments.


Experiment


eight


-grade


children


were


individually


expose


on a


daily


different


experimental


conditions


(multiple


schedule


ign)


one o


them,


each


was given


tokens


contingent


upon


correct


performance


on an arithmetic


task.


second


condition


eac h


S received


token


which


were


given


independent


performance


y accuracy


on the


same


responses


quality


to arithmetic


f performance


problems


rate


was measured


improvement.


Productivity,


attempted


rates


operationally


session


f problems


defined


regardless


completed


as the


accuracy,


session


frequency


was al


rate


arithmetic


measured


f change


problems


terms


in frequency


over


time.


Tokens


obtained


during


performance


could


ater


exchanged


a variety


toys


candies.


Experiment


seven


first


-grade


children


were


exposed


on a


daily


basi


different


experimental


conditions.


one o


them,









S made an incorrect response.

delivery and withdrawal of tc


In a second experimental


condition,


ikens occurred on a non-contingent basis


that


independently of the correctness of responses


As in Experiment I,


tokens


obtained during the performance could be


later


exchanged for


a variety of toys and candies.

Results of the experiments can be summarized as follows:


In terms of


accuracy of performance, no significant


difference was found between


' average


level of accuracy under the


contingent and non-contingent condition.


Also, both experimental


conditions


produced acceleration or a


trend toward


improvement in


accuracy in all


Ss participating in both experiments with the exception


of one


S who showed deceleration in the


level


of accuracy over time


under the non-contingent condition


These findings then,


failed to show the dramatic difference


previously reported


in the


literature regarding the differential


effects


produced by contingent and non-contingent conditions.


Neither did the


present results support most of the previous findings which have


hown


superiority of the contingent arrangements.


To help account for these di

istics of the method employed in th


fferences in results some character-


present investigation, such as the


kind of task involved and the multiple schedule


design utilized


were


discussed and contrasted with previous research.


generalization across experimental


It was argued that


conditions probably accounted largely


for the non-differential


effects found in thi


study








withdrawn


explaining


non-contingent


these


upon


results


their


terms


performance.


uperstititious


ability


responding wa


explored.


It was


argued


however,


that


future


replication


phenomenon


is necessary


before


making


any general


nations.


Finally,


some


suggestions


future


research were


offered.
































APPENDIX A


DESCRIPTION OF


AND NUMBER OF SESSIONS ATTENDED










Experiment I


Subject


Race


Number of
Contingent


sons


Attended


Non-Contingent


Male

Male


Male

Male


Female


Female

Female

Female


White

White


Black

White


Black


White

White

White


Experiment II


Subject


Race


Number of
Contingent


Sess


ions


Attended


Non-Contingent


Fema le

Female

Female

Male

Male


Male


Male


White

Black

Black

White

White


White


Black
































APPENDIX B

DIAGRAM OF EXPERIMENTAL SETTING SHOWING


RELATIVE POSITION OF


E AND













































/7j

/ '.4'
N I'
b I ,































APPENDIX


BLUE


SHEETS


CONTAINING


ARITHMETIC


PROBLEMS









SOOOOODOOODOO
000000000oOO00
OoOOO nOOr o
OcOOcOOco o


22
23
24
25


0 0 00 0 [000
0000 00000 0
OO00 0 0000


OOO OOOOO
00000 000 00
000O0000 o

OD


23
25
20


24


oo000o0000oo0 8
00 00000000 9


000


0 0o0 00 0 0
OO 0 o 0O 0
00 D3 D 0 D


0 0


0 0 0 0 000018
000000000 00


o0000 0
OO


00 00000
00 o 00
iCC0


000 00 0


000


0 O0 0 D OOOO 16


O 0 oOOO0 0


OOOO 0 0 0 15
000 0 o 0 ,


-










000 ] o 0


o000 oO


00000000
cOO000000
000oo000
0 [] 0o~o0


000000000
100000000
OCO c]OOQOr


0000


000000 219
000000 20

i0a 21


00000 0 000


0oOoO


23
22


0000 000
o0000 0
oooo o D O
D


000
S0 0
OOD
00


o0OO
000
300
Duo
El


000 0
o o 00oo 0
0 O 0 o 0
0 000


oc
0
0
0]
0


)0

0
0
D


000
0 0 0
00 0
0 00
Oo o


0 0000
000000































APPENDIX D

YELLOW SHEETS CONTAINING ARITHMETIC PROBLEMS









S0000OoOoo00
nOOOOuOoOu00
oO nO ] 0000
QOoQ 000000


25
23
24


22


00 0000000
00000 D 000
00000000
oo ooO~oooo


1100
OOl
00 c
D D


00[
00E
OJC

OO


0000
o000
)oOO
DO D


24
20


25


23


0000000000
00 0 o00OOD
000


)[0000
)oOOOOo
0000 D0


000 15
0
D


C 0OO 000
o0 00 O0
nOn


000 00000

00000ooo0
Doo


S19
n17


00 0o0000
0000 00 O
000f30000C


nOO


000 [0000


oOD


1110


000
00









ooo 00 o On
o 000000


000


000000


O


oo oo OO 00|o 9
S00000 0.0 o
ooo000000


O


00
00
00
O


0000ooOO
ODDOD
0oo0D


000000000 22
On00 OOOO O 21


SOo oo00000
0OnOo


20


oOoO000000 20
ODOQO. 21


00 0 O00 0 04
0000000


'0


D <
0
0


D cl
Us
LI


00 ] 000]


o































APPENDIX E


DAILY ACCURACY RATIOS FOR EACH


S UNDER


EACH EXPERIMENTAL CONDITION





















CII


6:I
:II-

4,-
a:




ri




'1I
C~ (X.?
2~
LI J..r------ --r --- --t---'"''~IT`^


-. I')


*0


-. FI Ii I S)I -- r% -*r A













-I -- ,--.4


O
0

jt.0
-- -

IVYS


O 4,
Ct
C)


n' n


0 --. 1-. -.--1---.-- .- ..-.. ..-----


1


c..
lii
~Ct


Ir I





,i ii









-I





I if








I T Ir '?If~












A- I~IIYf~~~I-IIC*)I U1LL .L -


mV




IIII







1*


i i



o r-----~--?--------1------------; ----------II
(3~~L I S oOr
ci Dt "myi


- 0
1Pmm:


O


0
0c
















___


U)



YII
2


C-)

I~I


,11










ILI K
--
* 9


(~) 0


0 to


IC?


0


a )
^.


I


(3 C)













I~A


0
C-
-S


0
-o3


.0 Co


0)
.1


LI
Lii

a:
c(


II


Q j


3LI. 1I
':C 3 *l----*-7 7~-I- r- - ~T1 ~- ---- 11.3


I)


C


L,-












1-n-.- -


n ,I


1It.
- i



d, i




CI

4-4*Y**)r IYUI~*lICrICI r l~C-I* r~ l--~


C)
-o


C.,













LL__ .I /C)


L (I,
~n0 >
c(3


zT I
g^"

LU
-J
0
4I


I





nl l


'J f(1


-- In


*0


~ (C)
Qc):


II


'0o


1Q
20




0


I


I1X 11.1\ 1 I













I._.. I i _i I L-- .


(I)






0ct)
-,I

iII











1


4 (


1'


O to


- u)


C*)


c3


c)>
SJ














L- -


r"S














\tI






rYI1-*,Wll~*l I TUY~ll~rIWLIrrC*r3ll*,l*Y f -iI


-D












1.~A 1.I____


ii- -


(I)

LI'
I i


O(1) c
C)'-
ci:


I





(3I -- -_- j-~---


O o
C) "


C.


0 IO


- It)


a3


Ci I I M xfnl.


o o












0


09


CI~ 4--"--"-


i

(0!













.,I
~YI,
I i






fl(J*
Iti I
OI


0 o


i
48Ut
I
C2


- IL')


rfl I5\ 1J \t4\ffC V


--


~ O






















CL)









I











-i



d~l-~----~-~-----.117~--,- --~~ CI


Qin


- 10


C)
0
* 1zi-


O
0f
tor


Q3


(I)






I-A


r"t? r ?J11 I Ib~\!l~rllH














0 .- --- I ---




II
cliJ
N*
*












CO-
*-






*w

N --


t~rF





0u
o;4


*1
















I I







. 'i7---I-r--l~I-.--..----t-- ., -_1- ,- - *-- -_i I


Ont


) C


tI'it U -ttt tJ r ,:L


(I')
4%-

''JI


C,
'2


lr)














CrJ --*--11-~--- I-L~-L
0




I. 2


II

tis


I I

-J
OI







IL 'I


Oi


0 u3
o CI-
0-) >-
0


0 0


O to


.- to


-. I. r













I I I _I


i t


.Oc
0C)


U)
Lii

I"ii

-

Cr


K I



r~- -"-11^^~.--.' .--lr- ....- *-T--- r


0


ni 2 i. *I~"Il 't-


00a
o C)


O


O













i I JI_ i __I O
i. tv I:


Cr)

CI! Ir
LU;


ICI
2:


0o
0o
0O


4









I
Ii





rcIt

L-~3I



~-----~- ~~'lr~-:`"Y-7-- -_-_I-~-.-r ----7---~ -----Iit


o In


L~---


S i)
0 8


Ut.


flu t r'..j I,%rrA-I..


i c)



-o



- -



-C)



a


















o~j






































.,I















r -r.
~~j
(1


- u)


tU I !.1I IrIt t rrt.


Z) U)











___ ___ 0L--
itrr~


(.0


-

.I





-r -i




'i -.


~~1'C


,-I


I


II


'(*) ~T rl*ri ~r~t* n~' *l/r~lUlt~l~ rr~r l2-~*~


O rC)


(.)C


Ott)


C)r


nl I













__ aL r I .- a- Li- amillumi m-mmu I- -r s----- ----^ -.-.^lA .I gi. -*

a *











I


II


II

I I



ii
I ~ I

--
I. C. I

#1 -l -. .,T j1 --- I --)--- (l--i
Ct 1 -7 -


*,- -


- -- .. aL a..l I












L____JII~ 0


O
to


(NJ


O






- -C
0)

0
0 (0


(7) -1t 0
tfl


0o
Oo
0 ^)'L


Q to


ow>r


OiLY A CrkLflnY:k/


CL i-
Cr
LI-
I)


*c


-i i


6-I-~-~--T- ~~ -- --------r-------r -r---.----7-.. -. I i. i


ct 8















































-
-I

c(03
C) o-
Pt--










11t








b3


- l _


I










*1




II


-~"---7~~r- ".-..r-F~."--:--~~T~ '---~r'7^'~'~-"' I


-it-)


i I X\-i i \Wt'


0
n-

I
?rmu*unyrJ~wL~


O rr,














I ___.~I_.--L



















(0










(I)
'A-y
Li
LU?



-J
o


--L----l_.---_-- C-----l--J-~T-^--7- r-III


a



tO


0
0



C, >-
o 0
0 ,.ct


0o
0 o
O Un


*1 e


0 ifl


t'd I -V wul1l-


C)


c 0
-*- 0


ll~~-----I ~r





























IJ


1



(rJ

1.
-J


C-
0)


O


FO
Ir
o





r)



cC)
0






* CO
C0 >-
C)


F ii i

_3 I3"O:


AO:)t' flQDV's






































(I')
'3

tti


IL.

* -w




*1

j


j.I I -~-L- -


Ic Is


*1

O j1~4


0 0


OW


K):


! 1 r'2J r'1


*C














-U-i-- -


(0$


















liii

rit


I"
t -.
U
-J
O.


I







--~i' r-~ -~1"~---- -~-~--- ---K-'--l' I" ^


0


Cj)



iiU;






1<)


C,, C)


ft I b' t'IIlffI


I--L-.L-.l


to)


0


(a