REGENCY AND SUMMATION EFFECTS OF NONREWARD
EN CHILDREN
Thesis for the ‘Degree of M: A.
MECHiGAN STATE UNWERSHY
NANCY HENDERSfiO'E‘? DAVIDSON
196.8
Will/Ill!!!III/IllWilli/lIII/Ull/l/l/I/l/lIII/III/II/lifl
3 1293 10383 3707
ABSTRACT
RECENCY AND SUMMATION EFFECTS OF NONREWARD
IN CHILDREN
By
Nancy Hendershott Davidson
This study was designed to investigate the effects of
recency and summation of frustrative nonreward on children's
performance. Kindergarten children performed a lever pulling
task on a three-lever apparatus. Session I consisted of 100%
reinforcement; Sessions II and III consisted of three partial
reinforcement patterns presented in random orders.
Analyses of latency and movement times at the third lever
confirmed the recency hypothesis; i.e., when a single nonreward
was administered in two different patterns a greater FE occurred
after the more immediate nonreward than after a nonreward that
was separated from the time of measurement by a rewarded reSponse.
The data analyses also supported the notion of the summation
pr0perties of nonreward since two successive nonrewards yielded
a greater FE than a single nonreward. Theoretical discussion
focused on a new concept of reward expectancy.
Approved:
Date:
Thesis Committee
Dr. Hiram Fitzgerald, Chairman
Dr. Ellen Strommen
Dr. Mar? Rillinq
REGENCY AND SUMMATION EFFECTS OF NONREWARD
IN CHILDREN
By
Nancy Hendershott Davidson
A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
MASTER OF ARTS
College of Social Science
Department of Psychology
1969
ACKNOWLEDGMENTS
I am sincerely grateful to everyone who assisted in
this research. My Special thanks are extended to Dr. Hiram
Fitzgerald whose patience and active concern were so important
to the success of this study. The support and helpful criticism
of Dr. Ellen Strommen and Dr. Mark Rilling were also greatly
appreciated.
I am also indebted to Royal Olson for his adeptness and
persistence in constructing the apparatus, and to Suzanne Murphy,
Valerie Veres, Doug Williams, and my husband, Jim, who all served
as second experimenters and without whom the study would have
been a nonrewarding and highly frustrating eXperience.
ii
TABLE OF CONTENTS
Page
ACKNOWLEDGMENTS . . . . . . ii
LIST OF TABLES . . . . . . iv
LIST OF FIGURES . . . . . . v
INTRODUCTION . . . L . . 1
METHOD - . . . . . . 3
RESULTS . . . . . . . 8
DISCUSSION . . . . . . . 12
REFERENCES . . . . . . . 16
TABLES . . . . . . . . l7
FIGURES . . . . . . . 23
Table
LIST OF TABLES
Schedule of Reward and Nonreward for Two Days
of Testing
Summary of Data for All Statistical Analyses
Summary of the Analysis of Variance on Latency
to R
2
Summary of the Analysis of Variance on Movement
to R
2
Summary of the Analysis of Variance on Latency
tR
03
Summary of the Analysis of Variance on Movement
to R
3
iv
Page
l7
l8
19
2O
21
22
LIST OF FIGURES
Page
Schematic Representation of the Apparatus . 24
Latency to‘R as a Function of Reward
Conditiog at R1. . . . . 25
Movement to‘R as a Function of Reward
Condition at Rl . . . . . 26
REGENCY AND SUMMATION EFFECTS OF NONREWARD
IN CHILDREN
Considerable research has been generated by Amsel's (1958)
interpretation of the 'frustrative nonreward effect'. By
'frustration' Amsel refers to the ”active prOperties of non-
reward following reward [Amsel, 1962, p. 306]." Amsel (1958,
1962) used a criterion response (running in the second length
of a double runway) that was Spatially different from the non-
reinforced instrumental reSponse (running in the first length).
Amsel suggested that subsequent to the development of reward
expectancy nonreward leads to frustration, an aversive emotion
which increases the reSponse that immediately follows the
unrewarded reSponSe.
Study of the frustration effect (FE) in children was
begun by Penney, whose findings are in accord with those of
Amsel in that "nonreinforcement of a reSponse at one lever (R1)
increases the Speed of a subsequent reSponse at a second lever
(R2), where both R1 and R2 have been reinforced by the same
reward [Penney, 1960, p. 214]." Penney (1960) reports that
the increment in Speed of R2 is a function of the number of
continuous reinforcements the child received prior to the
introduction of nonreinforcement. Nevertheless, as Ryan and
Watson (1968) have suggested, Penney‘s results may Show only
an apparent frustration effect that is attributable to Slower
Speeds following nonrewarded trials.
In their recent review of the frustrative nonreward
literature, Ryan and Watson (1968) point out that the use of
massed trials (Short intertrial intervals) characterizes many
unsuccessful attempts to obtain the FE. Using Spaced trials
(long intertrial intervals) to counteract the carry-over of
frustration from one trial to another, Ryan (1965) found that
smaller R2 reSponse latencies in the 50% reinforcement group,
as compared with the 100% grOUp, did support an FE interpretation.
The 50% group also had smaller reSponse latencies than the 100%
group on rewarded trials--a phenomenon which Ryan interprets as
a "ceiling effect or perseveration of frustration from nonrewarded
to rewarded trials [Ryan and Watson, 1968, p. 114)."
A study of size of interreSponse intervals (IRI, i.e., time
between R1 and R2) reveals that transient nature of the FE
(Watson and Ryan, 1966). For an IRI greater than 5 Seconds, R
2
movement Speed was not at all determined byRl reward conditions.
Thus, this~serigs of studies strongly suggests a frustrationv
effect with children; yet the theory of nonreward may be further
Specified by looking at the possible effect of the patterning of
nonreward on the size of the FE. A look at the frequency and
recency aSpects of that patterning might determine whether these
are at all related to the FE. Will nonreward of two apatially
separate reSponseS have a greater frustration effect on a third
reSponse than nonreward at just one of them? What effect will
nonreward of a reSponse have on a later reSponse when the two
are separated in time and Space by a rewarded reSponse? The
first of these questions deals with the summation of nonreward.
There is to date no study reported in the child literature
directed to this question. Nevertheless, Bower (1962),
employing a three-alley apparatus--an extension of Amsel's
two-alley runway--found with rats as §s that experience with
two nonrewards led to faster running times than did experience
with one nonreward. There have been no attempts to replicate
Bower's demonstration of a summation effect nor to extrapolate
to the interpretation of frustration effects in children. The
first purpose of the present study is, therefore, to attempt to
test Bower's finding in a different Operant setting—-a three-
lever analogue of the three-alley runway-~with children as §S.
The second purpose of the study is to determine whether there is
an effect due to recency of nonreward.
METHOD
Subjects. Subjects were 50 children from the Elliott School
Kindergarten, Holt, Michigan. A total of 20 children were
discarded due to absence on one of the three consecutive days
of testing (5 children were lost in this manner), and due to
apparatus malfunction (failure of relays to stOp timing devices).
Consequently, the final sample in the study consisted of 30
kindergarten children (Mean age::5.8 yr, sd =.36 yr.). There
were 15 boys and 15 girls.
Apparatus. The apparatus was a child—Sized metallic gray
desk--a right-angled box with adjustable legs attached. As
indicated in Fig. 1, three levers (Rl,‘R2,‘R3) were ,gi,
mounted on the 22 x 30 in. sloping front. Also mounted on
the leping front surface of the desk were three correSponding
green stimulus arrows (81, S2, S3) and a metal goal cup. A
smaller separate console housed a power switch, a relay
switch for lever selection, and two clocks (one latency, one
movement), as well as two amplifiers and a power supply.
The three levers were moved in the order‘Rl-R2AR3. (R1
was pushed to the left along a 16 in. horizontal track;'R2
began Slightly below the end of the R1 track and was pulled
down the leping surface toward § along an 11 in. vertical
track; and R3 began at the endpoint of the R2 excursion and
pulled toward the right along an 8 in. horizontal track.
Procedure. The.§ manipulated the incidence of nonreward
at R1 and/or atR2 and measured the effect of these manipulations
on §'s reSponse at R3. The manipulations of nonreward yielded
three nonreward-reward conditions which were randomly
administered to each S. A trial-sequence was represented by
any one of the following conditions:
at R2 and
Condition'R—NRAR Reward at R1 followed by nonreward
at R2 and reward at R3.
Condition NR-R-R Nonreward at RlRfollowed by reward
30
Condition NR—NR-R Nonreward at R1 and R2 and reward
at R .
3
A lever pull was initiated by a start Switch, which in
turn illuminated a stimulus light and activated an electronic
timer. The stimulus lights served to direct the child's
attention to the appropriate lever and to Signal the beginning
of a lever pull. As soon as a lever was moved, a photoelectric
cell stOpped the first timer and started the second. The
first timer recorded latency (time between stimulus onset
and initial movement of the lever) while the second timer
recorded movement time (duration of the lever pull. After a
lever was moved through its entire excursion, a marble was
deliVered manually from behind the hand shield via a trans-
parent tube located at the left side of the desk. After
each marble was deposited in the goal cup, § retrieved the
marble and drOpped it into a vertical plastic tube mounted
on a plastic ledge. The IRI was maintained at 5 seconds
Since Watson and Ryan (1966) report that with longer intervals
there is no relation between events at two successive
levers. The ITI was two minutes So as to minimize possible
carryover effects from one trial-sequence to another, while
not prolonging the session unnecessarily. During the two-
minutes ITIS,‘§S were read short picture stories. The story
reading was to provide a relatively uniform intertrial
interval activity for all §S which could be easily interrUpted
when resuming the lever pull task.
The experimental design consisted of one day of training
under continuous reinforcement, followed by two days of the
test procedure during which the three reward-nonreward
conditions were introduced. The situation was introduced
to the child aS a game and story time. Each.§ was shown
an array of small toys and asked to select the one he would
like to win. The child placed the toy on the ledge at the
t0p of the marble rackup tube and was told that he might
keep the toy he had chosen if he could completely fill the
marble tube with marbles (capacity==20 marbles). Then the
child was told that he could "win" marbles by moving a lever
from one end of its track completely to the other end. The
Signal light was pointed out and the child was instructed to
move the lever "as soon as the light goes on". In addition,
he was instructed to move the levers in the proper sequence:
first the top lever, then the side lever, and finally the
bottom lever, and then begin anew with the top lever. §
placed his hand on the front edge of the desk after each
manipulation of a lever.
Simultaneous with the instructions, each child completed
two practice runs (i.e., performed all procedural steps for
two lever pulls-AR and R2). The remainder of the training
1
session consisted of 6 trial-sequences (Rl-R 4R3) for eighteen
2
marbles---which when combined with the two marbles gained in
practice enabled §Hto win the toy.
On the following two days, the trial-sequences randomly
comprised all three reward-nonreward conditions, yielding a
total of 12 trial—sequences and enabling S to acquire twenty
marbles. With the exception of the introduction of nonrewarded
trials all other procedural details in the training phase were
retained in the test phase.
Table 1 indicates the reinforcement schedule used for
all §s during both days of testing with nonreward. The
schedule was established by starting the day with the R-NR
pattern in order to reestablish the reward expectancy from
the previous day on the first lever pull. The introduction
of nonreward then followed according to the predetermined
random schedules.
The experiment was carried out in a small room near the
children's kindergarten room. One E gave directions to the
child, returned levers to their original positions, timed
intertrial and interreSponse intervals, and read the two-minute
stories. A second.§ manipulated the start switch, channel
selector and timing devices on the console (removed from the
child's line of vision), and recorded latency and movement
times.
Data Analysis. Four separate analyses of variance (Lindquist
Model AdaysX Breinforcement pattern XCsubjects) were performed--
one each for latency to R2, latency to R3, movement to R2, and
movement to R3. After the main effects were tested in the
main analysis of variance, the experimenter looked at the
nonreinforcement phase in terms of planned comparisons for the
R3 latency and R3 movement times. The comparisons were
made between the R—NR and the NR-R orders and between the
NR-NR order and the remaining two.
RESULTS
Table 2 summarizes the entire study in terms of total
reSponse times (or the time to reSpond, summing across all
trials on a given day and across all §S)- The reader should
bear in mind that the nonreward condition at R2 is actually
comprised of two of the reinforcement patterns. Therefore,
the NR-R and the NR-NR columns Should be collapsed into one
condition, nonreward at Rl, Since §s did not experience the
different consequences at R2 until after reSponse measures
for the second lever had been taken. For example, for
Latency to R2 on Day 1, the nonreward condition reSponse time
of 112.45 sec. (the average of 112.84 and 112.06) was
compared with the reward condition time of 119.69 sec.
Before analyzing for the summation and recency effects,
it was deemed necessary to demonstrate the fundamental FE
described in the nonreward literature. The results of the
study satisfy this requirement Since children's performance
at the second lever was characterized by shorter latency and
movement times following nonrewarded trials than following
rewarded trials. This conforms with other studies that have
demonstrated the FE in children, using latency measures
(Ryan, 1965; Watson & Ryan, 1966; Lobb, Moffitt, & Gamlin,
1966) and movement Speeds (Penney, 1960; Watson & Ryan,
1966).
Latency tp‘32.—— The analysis performed on latency to
R2, summarized in Table 3, yielded a significant order
effect. This indicates a shorter latency following trials
that are not rewarded at R than is the case when R1 is
l
rewarded. This is the frustration effect of nonreward.
It is apparent as Shown in Fig. 2 that, although not statis-
tically significant (F==1.22), the difference between the R1
reward condition (R-NR) and the nonreward conditions (NR-R/NR—NR)
is even greater on Day 2 than on Day 1.
Movement tp‘fi .-—— Measures ofR2 movement time reflect
2
very nearly the same pattern as those of latency. AS Shown in
Table 4, theorder effect was again significant, indicating
that movement reSponse time was faster after nonrewarded trials
than after those that were rewarded, and confirming the
‘10
fundamental frustration effect found in Latency to R2. It
is interesting to note that latency times for the reward
condition (R-NR) were considerably longer on Day 2 than on
Day 1 (see Fig. 2). 0n the other hand, movement times
were considerably shorter on Day 2 than on Day 1 (see Fig. 3)-
Latgncy.tg‘fl3.-— Table 5 summarizes the analysis of var-
iance on R3 latency. There was a Significant main effect due to
days reflecting a decrease in latency with the additional
day of testing. The main effect of order was also Significant
and, interestingly enough, latency to R3 decreased in the
order of NR-RJ>R-NRJ>NR-NR (see Table 2). This was the order
expected given that recency effects are Operating in the second
condition and both recency and summation effects are found in
the third condition.
A planned comparison of the NR-R and the R—NR conditions
indicated a Significant difference between them [£_(1,58)= 6.47,
p<.05]. This is evidence for the recency effect of nonreward,
since the R-NR condition had a greater effect on quickening
§fé reSponse at R3 than did the NR-R condition. The two
conditions employed differed procedurally only in the order
of the nonreward; they were equivalent as to the numbers of
ll
rewards, nonrewards, and transitions between the two--
important considerations in view of the work of Capaldi on
sequential effects (Capaldi, 1966; Capaldi, 1967). Any
Systematic differences between the conditions Should, therefore,
have been due to the Spacing of the nonreward. One
implication of the greater frustration effects under the
R-NR condition may be that with the more recent incidence of
nonreward, its frustrating effects on subsequent reSponseS
had less chance to dissipate with time than those of the
earlier nonreward-~or less chance to be cancelled out by the
effects of a subsequent reward.
A planned comparison between NR-NR and the remaining two
conditions also revealed a significant difference [£_(l,58)=l2.50,
up<..Ol]. The NR-NR order showed a significantly greater
frustration effect than either of the remaining orders. This
suggests that subsequent effects of successive nonreward may
be said to be cumulative, supporting the summation hypothesis
of the study.
Movemegt,tp‘R .-— AS shown in Table 6, the only Significant
u—e
effect was that due to order indicating, as with Movement to R2,
that reSponse was faster after nonreward trials than after
reward trials. A planned comparison between the R—NR and NR—R
conditions did not yield a Significant recency effect for
movement times E5 (1,58): 6.50, 941.025].
12
DISCUSSION
The present study was designed to investigate the effects
of recency and summation of nonreward with children. (The
results demonstrated that both recency and summation of
nonreward influenced children's performance. The study
suggests that the frustration effect depends not only on the
incidence of nonreward, the variable most previous investi-
gations of the FE have considered, but also on the patterning
of nonreward. It is the nonreward patterning that prompts
the recency and summation contentions.
The evidence gained in the present study supporting both
recency and summation effects in frustrative nonreward strengthens
the conclusion advanced by Ryan, Strawbridge, and Watters (1969);
for conditions receiving the same number of rewards but arranged
in different configurations, the reward expectancies may be
very dissimilar. )Ryan et. al. (1969) have suggested that a new
concept of reward expectancy may be necessary if Amsel's frus-
tration theory is to be useful for accounting for children's
behavior in partial reward Situations. In their experiment
different groups of children were given from one to four
N-lengths (N-length referring to the number of consecutive
nonrewards). The study was to determine what effect these
different N-lengths would have on the partial reinforcement
acquisition effect (PRAE). In addition to the four N-length
groups, a random 50% group and a 100% reward gIOUp were also
included in the study. Significantly faster Speeds were
13
obtained in the 3 N-length and 4 N—length groups as well
as in the random 50% group, when compared to the 100%
control group. Performance of the 1 N—length and 2
N—length grOUpS was not Significantly different from the
performance of the control grOUp.
On the basis of these results, Ryan et. al. suggest
that the expectancy for reward, and hence the FE, can be
modified by S's learning the nonreward pattern. Differences
among the reward grOUpS were Seen as a function of the
extent to which §S could learn the sequence of rewards and
nonrewards and thereby reduce their expectancy for reward
on nonrewarded trials. Since the patterns of Single and
double alternations are more likely to be learned by this
age group (kindergarten and first grade) than those of
triple and quadruple alternations, the FE of nonreward
would be expected to be less for the first two patterns.
Since the data on recency and summation effects in
the present study are not inconsistent with Ryan's notion
of expectancy, some methodological procedures can be con-
sidered for inSpecting that kind of reward expectancy.
One problem that has plagued investigators of children's
performance in frustrative nonreward Situations is that
they can at best only indirectly infer the child‘s expec-
tancy for reward. One way to assess expectancy for reward
would be to agk the child after reSponding at lever 1 or
lever 2, to predict the reward condition at the next lever.
Success at prediction would SUpport the Ryan notion above.
14
We are currently investigating this question.
A second way to deal experimentally with reward
expectancies would be to determine the effects of learning
on FE. In the present study, all reward-nonreward sequences
were randomly presented to each child and could not, therefore,
be reliably anticipated or learned. On the other hand, if
‘Ryan et. al.'s contention that learning can modify the FE is
correct, one would expect stddies designed to directly in-
vestigate learning influences to Show such a modification.
Another experiment currently underway is attempting to measure
the extent to which learning reward and nonreward patterning
affects the recency and summation aSpects of frustrative
nonreward.
Interestingly enough, the results of the research Show
considerable variation in‘R reSponse times across reinforcement
3 .
orders, even though the reward expectancy for R can be assumed
3
to be equal under all reward—nonreward conditions in the sense
that a reSponse at that lever was always rewarded--regardless
of the nonreward pattern. Some children verbalized an awareness
that they always received a marble at R3, yet this realization
did not seem to produce any systematic change in their R3
reSponses. The variability among the three conditions in
reSponse time at the third lever is perhaps supportative of a
"motivational" interpretation of the effects of nonreward in
this study rather than an "associative" one (Pederson, 1967).
A more cognitive interpfietation of reward expectancy would
predict an equal reSponse at the third lever for all orders.
-‘Ill’l'l.|ll|ln‘.[lvlr iltflilllll'ill
REFERENCES
REFERENCES
Amsel, A. The role of frustrative nonreward in noncontinuous
reward Situations. Psychological Bulletin, 1958,.§§,
102—119.
Amsel, A. Frustrative nonreward in partial reinforcement and
discrimination learning: some recent history and a
theoretical extension. Psychological Review, 1962, p2
306—328.
Bower, G. H. The influence of graded reductions in reward
and prior frustrating events upon the magnitude of the
frustrating effect. Journal of Com arative and
Physiological Psychology, I962:p§§, 582-587.
Capaldi, E. J. Partial reinforcement: a hypothesis of
sequential effects. Psychological Review, 1966, 13,
459-477.
Capaldi, E. J. A sequential hypothesis of instrumental
learning. In K. W. Spence & J. T. Spence (Eds.),
The psychology of learning and motivation. 221. 1.
New York: Academic Press, 1967, 67-156.
Lobb, H.,Moffitt, A. R., & Gamlin, P. Frustration and
adaptation in relation discrimination learning ability
of mentally defective children. American Journal pf
Mental Deficiengy, 1966, 11, 256-265.
Pederson, D. R. Associative versus motivational inter-
pretations of reward percentage effects on children's
performance. Psychonomic Science, 1967, 8, 139-140.
'Ryan, T. J. The effects of nonreinforcement and incentive
value on reSponse Speed. Child Development, 1965, g9,
1067-1081.
Ryan, T. J., Strawbridge, J. E., & Watters, R. G. Patterning
in partial reward schedules with children. Paper
presented at the Biennial Meeting of Society for Research
in Child DevelOpment, 1969.
Ryan, T. J., & Watson, P. Frustrative nonreward theory
applied to children's behavior. Psychological Bulletin,
1968, pg, 2, 111-125.
Watson, P., & Ryan, T. J. Duration of the frustration effect
in children. Journal 2f Experimental Child Psychology,
1966, 4, 242-247.
16
TRIAL
\OmflOUl-D-COMH
I—al—ar-o
nor-co
TABLE 1
SCHEDULE OF REWARD AND NONREWARD
FOR TWO DAYS OF TESTING
PATTERN OF REINFORCEMENT
Reward-No nreward-Reward
No nreward-Reward-R eward
Reward-No nreward-R eward
Nonr eward-No nr eward—Reward
No nreward-Reward-R eward
Nonreward-No nreward-Reward
Reward- No nreward-R eward
No nreward-No nreward-R eward
No nreward-Reward-Reward
Reward—No nreward—Reward
No nreward-Reward-R eward
No nr eward- No nr eward-R eward
l7
(R-NR-R)
(NR—R-R)
(R-NR-R)
(NR-NR-R)
(NR-R-R)
(NR-NR-R)
(R-NR—R)
(NR-NR-R)
(NR—R-R)
(R—NR—R)
(NR-R-R)
(NR-NR—R)
mo.moa om.maa 00.0HH
ma.oma om.vma om.HNH
Vm.moa o>.mba hm.vma
o©.awm Ho.mvw om.NNN
No.mm vo.mm ab.©m
ma.dm NN.O© mm.mm
mo.m© av.vo 00.0m
ma.ao mm.o> ¢D.H©
Ho.vm om.¢w om.om
mm.aw Ha.om no.00
mo.ooa NN.©NH 00.0HH
©0.NHH m©.mHH vm.NHH
mzlmz mzlm mlmz
m
m H2m2m>02
mzlmz mzlm. .mlmz
mm e2m2m>oz
cam.mw mmono< possum mama mpma
mzlmz mzlm mlmz
mm >ozmeozmemo cm>w0 m Hom mamfiah
.mpcoomm mm
commmHme mH< mmefih pcme>oz cam >ocmpmq HmPOH
umMW>Am<223m
N mqmdg.
mlmz
A