Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Reward Expectancy Strength As Related To The Magnitude Of Frustration In Children
(USC Thesis Other)
Reward Expectancy Strength As Related To The Magnitude Of Frustration In Children
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
REWARD EXPECTANCY STRENGTH AS RELATED TO THE
MAGNITUDE OF FRUSTRATION IN CHILDREN
by
Dunham Harding Gilbert
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Psychology)
January 1969
This dissertation has been
microfilmed exactly as received 69-6493
GILBERT, Dunham Harding, 1934-
REWARD EXPECTANCY STRENGTH AS RELATED
TO THE MAGNITUDE OF FRUSTRATION IN
CHILDREN.
U niversity of Southern California, Ph.D ., 1969
Psychology, experim ental
University Microfilms, Inc., Ann Arbor, Michigan
UNIVERSITY O F SO U T H E R N CALIFORNIA
T H E GRADUATE SC H O O L
U N IV ERSITY PA R K
LOS A N G ELE S. C A L IF O R N IA 9 0 0 0 7
This dissertation, written by
Dunham Harding Gilbert
under the direction of h.h$... Dissertation C om
mittee, and approved by all its members, has
been presented to and accepted by the Graduate
School, in partial fulfillment of requirements
for the degree of
D O C T O R OF P H I L O S O P H Y
Dean
S e p te m b e r 1 8 , 1968
DISSERTATION COMMITTEE
Chairman
ACKNOWLEDGMENT
To my dissertation committee: Dr. Langdon E.
Longstreth, Dr. William W. Grings, and Dr. Edward C.
McDonagh, thanks are due for guidance and cooperation.
The primary direction was provided by Dr. Longstreth who
served as the committee chairman; his critical evaluations
and suggestions were invaluable.
Mrs. Elizabeth Purcell, principal of the Thirty-
second Street School, Los Angeles, was generously
cooperative. I am particularly indebted to the teachers
involved and to the children who participated.
The major portion of the data analysis was
performed at the Western Data Processing Center,
University of California at Los Angeles.
I sincerely appreciate the patience and many
sacrifices of my wife, Doris. My parents, Mr. and Mrs.
John D. Gilbert, have provided a source of constant
encouragement through the years.
ii
TABLE OF CONTENTS
Page
ACKNOWLEDGMENT .................................. ii
LIST OF TABLES.................................. vi
LIST OF FIGURES................................ viii
Chapter
I. INTRODUCTION ............................ 1
II. LITERATURE REVIEW ...................... 8
Energizing Effects of Frustration
Literature on Expectations
(Ant ic ipa t ions)
The Relationship between Frustration and
Expectancies (Anticipations)
Reward Expectancies (Anticipations) and
Extinction
III. THEORETICAL ANALYSIS OF THE PROBLEM AND
THE EXPERIMENTAL P L A N ................ 25
Introduction
General Description of the Experimental
Plan
Analysis of Expectancies as Related to
the FE
First B Test Trial
Second B Test Trial
Third and Fourth B Test Trials
First and Second D Test Trials
Analysis of Expectancies as Related to
Extinction
Experimental Hypotheses Stated in
Operational Form
iii
Chapter
Hypothesis 1
Hypothesis 2
Hypothesis 3
IV. METHOD ..................................
Subjects
Apparatus
Procedure
Stimulus-naming Phase
Pairing Phase
Instrumental Training Phase
Test Phase
Extinction
Design
V. RESULTS ................................
Preliminary Analyses
R Amplitude
R Speed
Verbal Responses
Tests of Hypothesis 1
Second B Trial in Test
Third B Trial in Test
Fourth B Trial in Test
An Alternate Approach: Dunnett's Test
Tests of Hypothesis 2
Analysis of T4 Data
Analysis of T2 Data
The Dunnett Technique as Related to
Hypothesis 2
Tests of Hypothesis 3
Number of Subjects Reaching the
Extinction Criterion
Response Characteristics during
Extinction
Supplementary Analysis of Test Trials
VI. CONCLUSIONS AND DISCUSSION ..............
VII. SUMMARY
Chapter
REFERENCES
APPENDIX
LIST OF TABLES
Table Page
1. Outline of Experimental Plan and Procedure 28
2. Theoretical Values at the Time of Response
on the First and Second B Test Trials for
Nonrewarded and Rewarded Subjects .... 32
3. Theoretical Values at the Time of Response
on the First D in Test for Previously
Nonrewarded and Rewarded Subjects .... 39
4. Mean Response Amplitude in Mm. during Late
Instrumental Training and on the Second
Test T r i a l ........................ 79
5. Mean Response Amplitude in Mm. for T4
Subjects during Late Instrumental
Training and on the Third Test Trial . . 81
6. Mean Response Amplitude in Mm. for T4
Subjects during Late Instrumental
Training and on the Fourth Test Trial . . 82
7. Mean R-Amplitude Difference Scores in Mm.
and £ Test Outcomes for Subjects
Receiving No Marbles in Test: Second
through Fourth Test Trials .......... 85
8. Mean Response Amplitude in Mm. for T4
Subjects during Late Training and on the
First D in Test.................... 88
9. Mean Response Amplitude in Mm. for T4
Subjects during Late Training and on the
Second D in T e s t .................. 89
10. Mean Response Amplitude in Mm. for T2
Subjects during Late Training and on the
First D in Test.................... 90
vi
Table Page
11. Mean R-Amplitude Difference Scores in
Mm. for D Test Trials: Based upon
Only Subjects Receiving No Marbles
during Test............................ 92
12. Mean Response Speed in 1/Sec. for T4
Subjects during Late Instrumental
Training and on the First D Test Trial . 100
vii
LIST OF FIGURES
Figure Page
1. Mean Response Amplitude in Mm. as a
Function of Training Trial Blocks
for the T4 Condition.............. 61
2. Mean Response Amplitude in Mm. as a
Function of Training Trial Blocks
for the T2 Condition.............. 62
3. Mean Response Amplitude in Mm. as a
Function of Training Trial Blocks for
T4 and T2 Conditions Combined.......... 63
4. Mean Response Speed in 1/Sec. as a
Function of Training Trial Blocks for
the T4 Condition...................... 71
5. Mean Response Speed in 1/Sec. as a
Function of Training Trial Blocks for
the T2 Condition...................... 72
6. Mean Response Speed in 1/Sec. as a
Function of Training Trial Blocks for
T4 and T2 Conditions Combined.......... 73
7. Mean Response Amplitude in Mm. as a
Function of Late Training and Extinction
Fifths................................ 95
8. Mean Response Speed in 1/Sec. as a
Function of Late Training and Extinction
Fifths................................ 96
viii
CHAPTER I
INTRODUCTION
The present study was concerned primarily with the
concepts of frustration and reward expectancy, as well as
the possible relationship between the two theoretical
terms, as they pertain to certain aspects of children's
behavior. The prefatory remarks are directed to an
abridged statement of current ways of treating frustra
tion as a scientific concept, followed by a concise
consideration of the theoretical system from which the
experimental hypotheses were derived. Moreover, a pre
liminary explication of said research questions is
offered.
The historical antecedents of modern frustration
theory are an important and interesting study; however,
since several excellent expositions already exist (Lawson
and Marx, 1958; Lawson, 1965), no attempt is made to
present a detailed overview of early accounts of frustra
tion. Generally, the major trend has been away from the
1
more grandiose efforts in which a theoretical structure
was established to encompass frustration, to more recent
positions in which the term frustration has been assigned
conceptual status in an existing general behavior theory.
Two examples of the latter are the approaches of Brown and
Farber (1951), and Amsel (1958), which represent highly
similar integrations of frustration into Hull-Spence
behavior theory. Particular attention is devoted to
Amsel's notions of frustration since the experimental
hypotheses are closely related to such a position.
In early Hullian theory (Hull, 1943) no special
emphasis was placed upon the energizing properties of non
reward, but later (Hull, 1952) at least nominal recognition
was forthcoming since frustration was attributed drive-like
properties. Similarly, Amsel (1958) has defined frustra
tion very restrictively as ”... a hypothetical, implicit
reaction elicited by nonreward after a number of prior
rewards.” The result of frustration is to add an incre
ment to generalized drive, D, due to the aversive nature
of frustration, thus leading to higher reaction potential,
E, and finally to more vigorous responding. It is impor
tant to note that this energizing effect is predicted for
behavior following frustrative nonreward; the enhanced
behavior, as compared to appropriate control conditions,
defines the frustration effect (FE).
An additional assumption of Amsel involves the
importance of fractional anticipatory (antedating) goal
reactions, rg's> in instrumental behavior. Through the
processes of classical conditioning and stimulus generali
zation, pregoal stimuli come to elicit a fractional,
noncompetitive portion of the goal response, r^} and
this r. is also assumed to produce characteristic internal
O
stimulation, s . Spence (1956) has posited that the
S
r_-s mechanism has motivational properties which deter-
© S
mine the magnitude of incentive motivation, K, another
theoretical term in the Hull-Spence system. When rg“sg
is not confirmed, for example, when reward is not present
at the goal after a number of previous rewards, frustra
tion results. Theoretically it is crucial that prior
rewards should have permitted rg to develop in the situa
tion; otherwise very little, if any, frustration should
occur following nonreward.
Amsel further proposes a secondary, antedating form
of frustration which may be denoted as r£-Sf. This frac
tional anticipatory frustration develops in the same
manner as other forms of rg, and occurs after an
instrumental response has been rewarded as well as non
rewarded a number of times. The rf-s^ mechanism is
conceptualized as having inhibitory properties due to the
fact that frustration is aversive; hence it is suggested
that Sf becomes associated with avoidance responses which
compete with established approach responses. The result
is at least a temporary decrement in behavior leading up
to the goal. Fractional anticipatory frustration
responses, along with the associated stimulus consequence,
have been given an important role in accounts of phenomena
such as the partial reinforcement effect and extinction
(Amsel, 1958), and discrimination learning (Amsel, 1962;
Amsel and Ward, 1965).
The present experiment was concerned with the
relationship between anticipation of reward and frustra
tion. According to Amsel, the greater the strength of
anticipatory goal responses, the greater is the magnitude
of frustration, and hence the FE. This implication has
been substantiated at the infrahuman level (Amsel and
Hancock, 1957; Amsel, Ernhart, and Galbrecht, 1961), but
has received limited attention using human subjects (Ss).
Longstreth (1960) provided one of the few experiments
(children served as Ss) designed to evaluate the following
hypothesis: "... the stronger the symbolic expectative
responses of reward, the greater the amount of frustra
tion when that reward is removed." Basically, the major
purpose of the present study was to provide further evi
dence with respect to the hypothesis just stated, but
using a distinctly different design and directing atten
tion to various response measures.
An alternate way of testing the basic question
involves measurement of response characteristics to a
stimulus which has never been followed by reward. For
example, assume that a stimulus (SI) has been consistently
associated with an instrumental response (R), which has
continuously been rewarded for a number of trials.
Concurrently, the same response, when paired with a
second stimulus (S2), has never been rewarded. If Sl-R
were suddenly nonrewarded, presumably frustration should
occur; if S2-R occurred a short time later, and assuming
that frustration perseverates, then the magnitude of R
following S2 should be larger than a representative
response to S2 prior to nonreward of Sl-R. Further, the
greater the expectation of reward, the larger the amount
of frustration, and hence the larger the response incre
ment to S2 in the situation described above.
Another assumption of the Amsel position is that
resistance to extinction is inversely related to the
amount of frustration. As stated earlier, the mechanism
which is mainly responsible for the decline in instru
mental approach responses to a goal, as the result of
repeated nonreward, is r^-s^. Since may well have
properties which are similar to fear, the prediction of
more rapid extinction with greater quantities of frustra
tion appears to be reasonable. If the strength of reward
expectancy is positively related to the magnitude of
frustration, it logically follows that the strength of
reward expectation is inversely related to resistance to
extinction. The latter deduction was tested, but was of
secondary importance with respect to the principal goals
of the experiment.
In summary, the fundamental purpose of the
reported experiment was to test selected implications of
Amsel's frustration theory at the human level. The
experimental hypotheses, stated in general terms, were:
1. If frustration is introduced into an instru
mental reward situation, then the stronger
the expectations of reward, the larger is the
frustration effect (FE).
2. If Sl-R, a frustrated stimulus-response event,
is followed in close temporal succession by
S2-R, a second stimulus paired with the same
instrumental response but never rewarded, then
the magnitude of the facilitative effect
attributable to perseverative frustration on
S2-R is a positive function of reward expec
tancy strength associated with Sl-R.
3. The strength of reward expectancy is inversely
related to resistance to extinction in the
instrumental reward setting.
Operational statements of the hypotheses are
presented in Chapter III.
CHAPTER II
LITERATURE REVIEW
Energizing Effects of Frustration
Although many investigators have noted more vigor
ous behavior following nonreward (for example, Marzocco,
1950; Skinner, 1950), the first experiment specifically
designed to test implications of the Amsel theory was
that of Amsel and Roussel (1952). The apparatus consisted
of a start box followed by an enclosed runway (Al) leading
to a first goal box (GB1), then to a second alley (A2),
and finally to a second goal box (GB2). Under conditions
of hunger motivation, rats were rewarded with food at GB1
and GB2 for a number of trials, the primary response
measure being running speed in A2. During the second
phase, reward was present for half the trials in both
GB1 and GB2, while for the other half, reward was omitted
from GB1 but present in GB2. The latter nonrewarded
trials presumably should have resulted in frustration and
8
hence produced more vigorous responding in A2 (post-
frustration behavior). The data showed that nonreward
in GB1 was followed by greater running speed in A2 than
was the case where 6B1 was rewarded. A number of subse
quent studies have verified the phenomenon (for instance,
Roussel, 1952; Amsel et al., 1961).
Seward, Pereboom, Butler, and Jones (1957) criti
cized the Amsel and Roussel study by noting that the
observed performance difference could have been attribut
able to response depression after rewarded trials rather
than to response enhancement due to frustration. That is,
rewarded trials presumably reduced hunger drive, thus
leading to lowered motivation and response depression.
Wagner (1959), using a modified Amsel apparatus and an
appropriate control group (one never rewarded in GB1),
found no support for the contention of Seward et al.
(1957).
Graham (1963) demonstrated the FE in an instru
mental escape setting, thus extending the generality of
the effect, and showed that the noted intensification was
not attributable to a more basic operation of failure to
confirm any expectancy. Davenport, Flaherty, and Dyrud
(1966) reported an FE for monkeys using a bar-pressing
10
analogue to the tandem runway, thus Indicating that the
phenomenon is not restricted to rats.
A very recent study by Tacker and Way (1968)
investigated the effect of nonreward upon general activity,
rather than on a learned response, as has been the case in
the majority of FE experiments. Rats were trained to
traverse a straight alley for food reward, finally arriv
ing at a goal box which was essentially an open field.
Response measures included extent of ambulation and
rearing behaviors in the open field following both
rewarded and nonrewarded trials. The results indicated
an increase in both measures following nonreward.
The recent trend in the animal literature has
apparently been to determine the relationship between the
FE and various independent variables. For instance, the
magnitude of the FE has been found to be: (1) positively
related to the magnitude of previous reward (Peckham and
Amsel, 1964, 1967); (2) greater for rats subjected to more
severe drive conditions (McHose and Ludvigson, 1964);
(3) independent of the effortfulness of an instrumental
response (Grusec and Bower, 1965); and (4) essentially
nil as a result of incomplete reduction in reward in the
double-alley setting (Barrett, Peyser, and McHose, 1965).
The data from human studies also confirm Amsel's
assumption regarding the energizing effects of frustra
tion. In two relatively early studies (Screven, 1954;
Screven and Cummings, 1955), more vigorous crank-turning
responses were noted after interference (children served
as Ss). Additional data have been provided by Haner and
Brown (1955) and by Holton (1961), both experiments noting
instrumental response intensification presumably due to
frustration. Particularly impressive were Holton's data,
since 44 of 45 Ss showed an FE. Similarly, Ryan (1965)
found some evidence for an FE in a task which required
the child sequentially to depress two response levers.
Further experiments (Longstreth, 1965, 1966) have demon
strated an intensification of a joystick response attrib
utable to frustration.
Quite recently, an extensive review has appeared
which presented further confirmatory findings (Ryan and
Watson, 1968), as well as several which apparently con
tradict Amsel's position. The authors cite unpublished
studies performed by themselves which showed no FE in
situations where one was predicted. However, the authors
do point out that a ceiling effect may well have been
responsible for the lack of positive findings. While one
12
cannot completely discount these data, a thorough evalua
tion is not possible.
A related type of investigation was reported
(Kobasigawa, 1965; Whiteley and Ryan, 1967) in which one
S monitored the instrumental behavior of another subject.
When the performing subject was frustrated, the observer
was required to display an instrumental response shortly
thereafter. The first study found typically elevated
response speeds and amplitudes in the observer who had
not been directly nonrewarded, but the second study noted
changes in the opposite direction with regard to a speed
measure. Even though the experimental arrangement is
intriguing, the contradictory results suggest that further
experimentation is needed to evaluate the discrepancy.
In conclusion, it would seem reasonable to state
that the FE is a well-established phenomenon with human
Ss, too, and that attention should now be directed
toward the conditions which affect its magnitude.
Literature on Expectations (Anticipations)
Since a great deal of importance is attached to
the ?g~Sg mechanism, and similarly to an assumed symbolic
(verbal) human counterpart in the present experiment, the
13
nature of such anticipatory responses requires examination.
In the past, investigators using r as a theoretical term
o
have typically employed the concept as a hypothetical
construct which had only indirect reference to observable
events. Recently, data have been presented which would
seem to provide an empirical referent for the r concept.
O
Shapiro (1961, 1962) used dogs as Ss and measured lever-
pressing behavior for a food reward. The typical dog had
a tube implanted in the right salivary duct, providing
a measure of saliva flow which, in turn, could be related
to rate of lever pressing. In the first study, saliva
flow regularly followed the lever response, but the 1962
study showed that maximal salivation regularly preceded
the lever response after a number of training trials.
Thus, the latter study suggested that the anticipatory
salivation may have had the status of an r . The dis-
J g
crepancy between the two studies is perhaps related to
the different reinforcement schedules used. More defi
nitive evidence has been provided by Deaux and Patten
(1964). A harness arrangement was devised which supported
a drinking tube located immediately in front of S's
(a rat) mouth, and the tube could be licked at any time.
The animals were required to traverse a straight runway
14
to a goal box; arrival at the GB resulted in water being
delivered through the tube. Licking responses could be
continually measured by means of a drinkometer arrange
ment. In this way a measure of anticipatory responding
(licking prior to reaching the goal) could be obtained
and related to running speed in various runway segments.
The data indicated that licking rate increased in a
positively accelerated manner as J5 approached the goal
area, and in a negatively accelerated fashion with train
ing trials from an initially flat gradient. The results
are certainly consistent with relevant predictions from
the Hull-Spence position regarding r„.
©
A second important issue involves the manipulation
of the magnitude of rg, and the resulting effect on
various behavioral indices. One class of experiments
has been termed latent extinction (see Moltz, 1957 for a
review). In the typical latent extinction procedure,
S is given several nonrewarded placements in a goal area
following training in which an instrumental response has
led to reward in the goal area; comparison is made to
controls which had no goal placements. The usual effect
is a decrement in the strength of the instrumental
response leading to the goal region. A reference study
was performed by Deese (1951) in which rats were trained
in a position habit on a U-maze with food as the reward.
Following training, experimental animals were given four
one-minute placements in the now-empty goal area, while
controls received no placements. In extinction the
former group yielded a significantly smaller proportion
of previously-correct responses in the maze. According
to a Hull-Spence interpretation, the goal stimuli would
elicit r ’s which would be nonrewarded for the four
©
placements. If pregoal cues were similar to goal stimul
extinction effects of r would generalize to antedating
©
behavior segments. At the same time, the magnitude of
s would be reduced, and thus the strength of incentive
g
motivation would decline. Both factors should lead to
a diminished level of instrumental responding to non
rewarded cues.
A second type of study involves an opposite pro
cedure; that is, intentional pairing of goal stimuli
with reward rather than with nonreward. For example,
Marx and Murphy (1961) individually exposed rats in an
experimental group to a total of 90 buzzer-food pairings
in the goal area of a straight alley. A control group
received the same number of reinforcements and buzzer
presentations but in such a way that no pairings occurred.
All Ss were subsequently trained to traverse the entire
straight-alley apparatus for food. On selected extinc
tion trials the buzzer was sounded while S S was in the
starting box. The results showed that on the buzzer
occasions, starting time was increased for control rats,
but that a decrease in starting time was evident for
experimental Ss. One interpretation (Spence, 1956) would
assume that the initial rewarded pairings in the goal
situation allowed anticipations of reward to develop
independently of the instrumental response (running)
which would be measured at a later time. When the buzzer
was activated in the start box, presumably an rg~sg
sequence was evoked only for the Ss who had received
pairings. On the further assumption that incentive
motivation is proportional to the strength of r , one
O
would expect intensified performance for the paired
group as compared to the group which had received no
pairings. Stein (1957) reported similar experiments at
the animal level with equivocal results.
Less ambiguous findings exist for experiments that
have used human Ss. Longstreth (1962) gave preschool
children six trials in a runway situation, the locomotor
response being followed by either a red or a.blue light
(three times each), and not accompanied by obvious reward.
Candy was then presented to the children in conjunction
with the red light but not the blue light, and independent
of the instrumental approach response. The Ss were then
placed in the runway once again with the major result
that the approach speeds for the stimulus paired with
reward were faster than those associated with the non
paired condition. Additional data (Longstreth, 1960,
1966) also support the contention that the r -s_ formula-
© O
tion may be extended to humans where symbolic expectations
of reward are involved, as compared to more peripheral
anticipations in the case of animals. Particularly
impressive was the response-defined evidence of r in
O
the third experiment of the latter monograph. No less
important was the observation of increased amplitude of
a joystick response for most S>8 who displayed rg's, thus
suggesting that incentive motivation is dependent upon
the occurrence of expectative responses, as Spence (1956)
has theorized.
The data from human research show very clearly
that reward expectancies may be manipulated independently
of an instrumental response, and can affect subsequent
18
performance of an instrumental response. It is now
appropriate to consider evidence which corroborates the
positive nature of the expectancy strength-frustration
magnitude relationship.
The Relationship between Frustration
and Expectancies (Anticipations)
A search of the literature revealed only three
examples of experiments using infrahuman Ss, which have
examined the relationship between reward anticipation
and frustration. In the first such study, Amsel and
Hancock (1957) ran 42 rats in the double runway situation
to water reward in 6B1 and GB2. One group had similar
cues in the first alley and GB1, while for a second group
the stimulus cues were dissimilar. Then a series of
frustration trials (no GB1 reward) was administered on
a 50 per cent basis. The FE was greater for the group
which was exposed to identical stimuli in the first alley
and GBl. The authors concluded: "Nonreward of an instru
mental response is frustrating to the extent that stimuli
accompanying the instrumental sequence leading to a goal
elicit fractional antedating goal reactions appropriate
to that goal.”
19
Two experiments of a similar design constitute the
second and third instances of animal studies relevant to
the anticipation-frustration relation. Amsel et al_. (1961) ,
once again using the double-alley apparatus, varied the
length of the alley leading to GB1. It was reasoned that a
longer alley would serve to increase the probability that
r had been evoked and was near maximal strength when J S
s
entered GB1. If frustration were then introduced after a
number of training trials, the prediction would be that of
a relatively greater FE for animals exposed to the long-
alley condition. The results indicated limited data con
sistent with the hypothesis of a positive relationship be
tween the strength of reward anticipation and frustration.
On the other hand, Clifford and Schindelheim (1968) noted
no changes in running speed as a result of manipulating
the length of the first alley (15 vs. Ill in.).
Studies which have used children as £s have been
more numerous, but perhaps have often involved more indi
rect tests of the relationship. Longstreth (1960) had
children depress a lever continuously in order to receive
subgoal rewards (marbles). For group P a light was paired
with marble ejection, but for group U the light was not
paired with marble delivery. In extinction, only the
20
light was presented: group P extinguished faster, and
there was some anecdotal evidence of intensified respond
ing for group P (vibration and movement of the response
box). Moreover, recordings of verbalizations indicated
that during extinction, 7 of 16 Ss in group P demonstrated
an "awareness*1 of the relationship between the light and
marble ejection. No £ in group U manifested such verbal
responses. Since actual recordings of response amplitude
were not made, one must conclude that for human Ss the re
lation between expectation strength and the FE (as measured
by response amplitude change) is extremely tentative.
Several experiments, which involve indirect evi
dence, have argued as follows: assuming that magnitude
of reward expectancy increases as approaches the goal,
one would predict a greater post-frustration response
enhancement when frustration occurs near to the goal, as
compared to the case in which frustration is introduced
at a farther point from the goal. Haner and Brown (1955)
instructed children to fill a marble board which contained
36 holes in order to win a prize and candy. Ss were
"frustrated" (the accumulated marbles dropped out of
sight) when 9, 18, 27, or 32 marbles had been delivered.
21
As the marbles disappeared, a buzzer was activated which
continued until S pushed a plunger to begin a new attempt.
Plunger displacement (force) increased in a monotonic
fashion as a function of nearness to the goal.
A portion of the study by Holton (1961) yielded
comparable data when preschool children were trained to
push one of two stimulus windows for a marble reward.
A group frustrated nearer the goal had a greater mean
force of window pushing than did a group for which non
reward occurred at a more distant point.
Two additional studies involved "failure" in a task
as opposed to nonreward, a distinction stated by Ryan
and Watson (1968). Ford (1963) had children work on a
10-piece Seguin Formboard; when a buzzer sounded, £ had
to depress a plunger to terminate the buzzer. During the
course of the experiment several success and failure
trials were given. Expectancy of reward was response-
defined, and was inferred from the number of candy coins
£ was willing to bet on the outcome of an early trial.
The relationship between response speed and expectancy
level was nonmonotonic, and no relationship was discovered
when the dependent variable was response amplitude. The
adequacy of the reward-expectancy variable is questionable,
22
however, since Ss had only minimal knowledge of the
experiment at the time of betting. Endsley (1966) found
no consistent changes in response speed or amplitude
associated with disc pressing on a supplementary piece
of apparatus, measured immediately after Ss failed in
a task which involved raising a metal ball on a platform
to a goal. Distance to the goal was varied by arranging
to have the ball drop off the platform at different
vertical heights. Ford speculates that tasks which
increase the probability that £ perceives failure as
being the result of self inadequacy are less likely to
reveal "aggressive" responses than those for which S
sees failure as being due to the manipulations of some
external agent, for example, the experimenter. The latter
possibility may have been operating in the 1955 study of
Haner and Brown which also involved failure rather than
nonreward. Of course, it is plausible that the two
operations (failure vs. nonreward) are quite different
in terms of behavioral consequences.
As a whole, the data are supportive of Amsel's
assumption concerning the relationship between reward
expectancy strength and frustration magnitude. It should
be recalled that the single experiment which directly
23
tested the proposition (Longstreth, 1960) was primarily
concerned with a resistance-to-extinction measure.
Certainly the extent to which the size of the FE (as
defined by response speed or amplitude) is a function of
reward expectancy strength is far from clear at the human
level.
Reward Expectancies (Anticipations)
and Extinction
The Longstreth studies (I960, 1966) have provided
data confirming the proposition that greater reward expec
tancies in children lead to less resistance to extinction.
In Experiment 2 of the latter monograph, for example,
Ss were instructed to perform joystick responses whenever
a stimulus light was presented. A light which was paired
with subgoal attainment (marble delivery) resulted in more
rapid extinction (median number of responses • 38) than
did exposure to a second light intensity which had never
been rewarded (median number of responses ■ 117) .
In a conceptually similar experiment, Elmes (1964)
partially reinforced rats in one 6B on one half of the
trials, and never reinforced the Ss on the remaining
trials in a different 6B. During extinction trials one
24
group ran to both goal boxes; another group ran to the
previously reinforced GB; and a third group traversed to
the nonreinforced GB. As predicted by frustration theory,
the third group was most resistant to extinction, since
they presumably encountered no frustration (nonreward of
a previously rewarded response) during extinction.
The inverse relation between reward expectancy
strength and extinction duration has been verified a suf
ficient number of times so that there is little doubt as
to the reliability of the findings. In any event, the
test of the relationship in the present experiment was
of minor importance only.
CHAPTER III
THEORETICAL ANALYSIS OF THE PROBLEM
AND THE EXPERIMENTAL PLAN
Introduction
The basic logic underlying the problem has already
been presented in Chapter I, so that only an abbreviated
presentation of Amsel's position is required at this
point. Relatively more attention is devoted in this
chapter to the behavior theory of Spence (1956).
Spence's theoretical structure assumes that
observable behavior is due to two basic classes of
hypothetical factors, namely, excitatory and inhibitory
factors. Among the relevant excitatory elements are:
habit strength (H), generalized drive strength (D), and
the incentive motivational construct (K) . A higher-order
construct, reaction potential (E), is assumed to be
determined by D, H and K in the following way:
E - (D + K) x H
The magnitude of E is presumed to be at least monotonically
25
26
related to observable behavior. Other contributors to E
are present in Spence's system, but for the present
development are not particularly applicable.
Amsel (1958) postulated that the result of primary
frustration is an increment to generalized drive strength
which may be designated as Df. Thus for instrumental
behavior which takes place immediately following frustra-
tive nonreward, the associated reaction potential value
would be:
Ef » (D + Df + K) x H
which obviously indicates that E^ > E, the latter being
a representative prefrustration reaction potential (Df
not present). Hence behavior subsequent to frustration
should be more intense than that just prior to nonreward.
Another basic assumption of Spence and Amsel con
cerns r -s as the mechanism which determines the values
8 g
of K and Df. It is supposed that both K and Df magnitudes
are monotonic functions of the strength of rg”8g* That is
to say, if *g”8g * - 8 evoked, and has been established at
different strengths for various Ss, then the greater the
value of ?g~8g the larger will be K and also Df. Conse
quently, larger anticipations (or expectations) of reward
will result in both increased prefrustration instrumental
27
responding (due to K) as well as greater post-frustration
behavior (due to Df). As stated earlier, if repeated non
rewards are continued for a sufficient number of trials,
then the secondary form of frustration develops (r^-Sf),
and, having inhibitory properties, produces interference
with the goal approach response. The final result is
experimental extinction. The preceding statements repre
sent only a fragmentary account of Spence's theoretical
efforts, but additional details will be added as necessary
during the course of the theoretical analysis.
General Description of the Experimental Plan
Assume that children are individually placed in a
situation where they are instructed that in order to win
a prize it is necessary to name two different light inten
sities (bright, B, or dim, D) as they appear in a stimulus
window, which is located in direct view of S on a stimulus
response unit. Table 1 shows that the phase labeled
"stimulus naming" corresponds to this procedure. The
purpose of the series is to ensure that each S can reli
ably distinguish between the two stimulus intensities.
After completion of stimulus naming, the children are
TABLE 1. OUTLINE OF EXPERIMENTAL PLAN AND PROCEDURE
Pairing
Subgroup
Stimulus
Naming
Phase3
Pairing
Phase3
Instrumental
Tra ining
Phase3
Test
Phase
Extinction
Phase
PI (10 B, 10 D)
No Marbles
(60 B, 30 D)
All 60 B
Followed
by Marble
(55 D, 5 B)
All 5 B
Followed
by Marble
1/4: BBBB DD*
1/4: BBBB DD
1/4: BB DD*
1/4: BB DD
*B Trials Fol
lowed by
Marble
Repeated B
Events; No
Marbles
P2 Same as PI Same Sequence,
but Only Last
20 B Followed
by Marble
Same as PI Same as PI Same as PI
P3 Same as Pi Same Sequence,
but Only Last
2 B Followed
by Marble
Same as Pi Same as PI Same as PI
P4 Same as Pi Same Sequence,
No Marbles
Same as PI Same as PI Same as Pi
P5 Same as Pi Same Sequence,
No Marbles
Same, but
No Marbles
1/2: BBBB DD
1/2: BB DD
No Marbles
Same as Pi
aSee text for actual sequences.
N5
00
instructed that the game now consists of accumulating
a sufficient number of marbles to fill a transparent tube
positioned on the front of the stimulus-response unit.
Then "pairing** is introduced which involves presentation
of 60 B and 30 D events, arranged in the same random
sequence for each S,. The second column in Table 1 shows
that the various subgroups of Sa receive different numbers
of B-marble pairings. The children are told simply to
indicate verbally whenever a stimulus light appears, and
also if a marble is ejected into the tube: on rewarded
trials the onset of B occurs about one second prior to
marble delivery and B is terminated when the marble drops
into the tube. On nonrewarded trials either B or D
continues for an approximately equal length of time, but
is not associated with the fact of marble presentation.
At the end of the sequence the experimenter then manually
places a certain number of marbles in the tube so that all
Ss are at about equal distances from the goal; that is,
all have the same number of marbles remaining to fill the
tube.
At the conclusion of pairing, the 2 3 is informed
that he must now turn a response handle in order to get
further marbles. A total of 55 D and 5 B events then
30
take place. Only Instrumental handle-turning responses
(R8> to B are rewarded, while the D lights are never
associated with reward. The notable exception to this
statement is subgroup P5 for which no marbles are ever
forthcoming. Typically, either B or D is turned on;
£ then performs R, at which time a marble is either
delivered or not as required by the design; and B or D
is terminated.
The phase labeled "test" then begins unannounced
upon completion of the instrumental training sequence.
For one quarter of the children within each pairing sub
group (PI through P4), four consecutive B events occur,
each R being nonrewarded (frustrative nonreward), followed
by two nonreinforced D presentations. A second quarter
of the Ss receive the same treatment, except that B events
are accompanied by marble ejection (control). The third
and fourth quarters are subjected to analogous reward
conditions, but only two B presentations take place.
The P5 subgroup is divided into two halves, one receiving
the sequence BBBBDD, and the other having BBDD— no marbles
are delivered. The primary reason for inclusion of differ
ing numbers of nonrewarded B events (four vs. two) is to
31
Increase the probability of detecting frustration effects.
Since little information is available regarding the number
of nonrewarded events necessary to produce adaptation,
the tactic of using selected numbers of frustrating events
should increase the chances of noting facilitation due to
frustration (particularly on D trials). Finally,
"extinction" is initiated and consists of a maximum of
approximately 185 consecutive B events, none of which is
associated with marble delivery. Of course, all Ss are
several marbles short of filling the tube during test and
throughout extinction. Further, assume that the amplitude
of R may be reliably measured and recorded.
Analysis of Expectancies as Related to the FE
As will be apparent as the analysis proceeds, it is
convenient to consider the first and second B test trials
separately. The first presentation of D during test is
also treated in a different section.
First B Test Trial
Table 2 presents a summary of the theoretical
analysis derived from a Hull-Spence standpoint for the
first and second B test trials. Preliminary evidence
TABLE 2
THEORETICAL VALUES AT THE TIME OF RESPONSE ON THE FIRST AND SECOND
B TEST TRIALS FOR NONREWARDED AND REWARDED SUBJECTS
Subgroup
bhr dhr
rH-
B rg
Dw Df
K ber
NONREWARDED Ss (FIRST B TEST TRIAL), AND
REWARDED Ss (FIRST AND SECOND B TEST TRIALS) :
PI 100 100 65 5 0 65 70 7000
P2 100 100 25 5 0 25 30 3000
P3 100 100 7 5 0 7 12 1200
P4 100 100 5 5 0 5 10 1000
P5 100 100 1 5 0 1 6 600
NONREWARDED Ss ONLY (SECOND B TEST TRIAL):
Pi 100 100 65 5 65 65 135 13500
P2 100 100 25 5 25 25 55 5500
P3 100 100 7 5 7 7 19 1900
P4 100 100 5 5 5 5 15 1500
P5 100 100 1 5 1 1 7 700
u>
K>
33
from a pilot study indicated that stable R amplitude is
achieved in less than 60 trials; this being the case, it
was reasonable to assume that the D-R habit strength
(d* * r) f°r all subgroups would be at maximum, say 100 units,
by the end of instrumental training. Even though not
established by the administration of subgoal rewards
(marbles), the D-R habit presumably developed as a result
of instructional factors alone (and perhaps through gen
eralization from bHr). In addition, the B-R sequence was
accompanied by marble ejection five times during instru
mental training for all but P5 Ss. The RHR habit loadings,
by assumption, were assigned values of 100 units for all
subgroups, since presumably asymptotic values had been
reached. It should be emphasized that the preceding
assumptions regarding both the B-R and D-R habits may be
gross oversimplifications; on the other hand, even if the
two habit strengths had other than asymptotic values for
the various subgroups, it is easy to show that the theo
retical implications would remain unchanged. This is
primarily due to the fact that difference scores were
routinely used to evaluate the hypotheses.
During the pairing phase, Ss in subgroup PI received
60 marbles in the presence of B, while for P2, P3, P4 and
34
PS the respective numbers of pairings were 20, 2, 0 and 0.
Also, during the course of instrumental training, B had
been paired with marble delivery a total of five times
(even though Ss were required to perform R simultaneously).
Hence the total frequencies of pairing for subgroups Pi,
P2, P3, P4 and P5 were 65, 25, 7, 5 and 0, respectively.
It was assumed that expectative verbal "marble" responses
(rg) became classically conditioned to B resulting in
habits, , with strengths in proportion to the number
of pairings administered. Thus values of 65, 25, 7, 5
and 1 were given to the subgroups Pi through P5; the
latter quantity for P5 was assumed to be other than zero
due to possible minimal reward expectancies established
on the basis of instructions. A value of zero would not
alter the analysis, however. Incentive motivation, K,
assumed to be at least monotonically related to the
strength of r , was typically assigned a value correspond-
ing to that appearing in the column in Table 2.
The drive component, Dw, represents the drive to
win the prize, and due to instructional factors should
have been associated with attainment of marbles. Arbi
trarily the value of Dw was assumed to be five units for
all subgroups. Since there was no frustration on the
35
occasion of response to the first B in test (nonreward
of an R to B had not yet occurred), a value of zero was
assigned to all subgroups for D^. As defined previously,
Df represents the increment in generalized drive attribut
able to frustration. The column in Table 2 designated
refers to the total motivational level present, and was
presupposed to be equal to the sum of Dw, Df and K.
Finally, excitatory potential, of R to B, on the
first trial, was equal to the product of gHg and Dt, for
example, for Pi: gEg ■ 100 x 70 ■ 7000.
Second B Test Trial
For nonrewarded Ss the situation at the moment of
responding on the second B test trial was different in
that perseverative frustration effects from nonreward of
the R to the previous B presentation must be included.
Assuming that frustrative nonreward produces an increment
in generalized drive, the values assigned to Df would no
longer have values equal to zero as was the case for the
first test trial. According to Amsel's proposition, the
magnitude of Df is directly dependent upon the strength
of rg at the time of nonreward. In an identical fashion
as that used for assignment of numbers to K, the subgroups
PI, P2, P3, P4 and P5 were given Df values of 65, 25, 7,
5 and 1, respectively. Then associated Dt quantities
were 135, 55, 19, 15 and 7. The reaction potentials
(gEg) on the second test trial for nonrewarded children
had values of: PI « 135 x 100 - 13500; P2 - 55 x 100 -
5500; P3 - 19 x 100 - 1900; P4 - 15 x 100 - 1500; and
P5 - 7 x 100 ■ 700. For Ss who continued to receive
mm
marbles on the first and second B test trials, the state
of affairs on the second B test trial was assumed to be
unaltered from that which existed at the time of the
first test B.
The numerical values assigned to the various
theoretical terms have only relative meaning; that is,
predictive implications were based upon comparisons
between subgroups rather than absolute comparisons within
subgroups. Accordingly, the differences in gEg (nonreward
condition minus reward condition) for the subgroups were:
PI - 13500 - 7000 - 6500; P2 - 5500 - 3000 - 2500; P3 -
1900 - 1200 - 700; P4 - 1500 - 1000 - 500; and P5 -
700 - 600 « 100. Since all factors other than Df were
the same for nonrewarded and rewarded Ss within each sub
group, the difference between the reaction potential
difference scores were presumably attributable to
differential frustration only. Assuming that response
magnitude is at least monotonically related to reaction
potential strength, it followed that the relative diffe
rences between R-amplitude scores for nonrewarded (frus
trated) and rewarded (control) Ss should have had the
following order: PI > P2 > P3 > P4 (> P5). When the
results of the present study are presented in Chapter V,
it will be clear that a further variable of importance
concerned an estimate of R amplitude during late training
vs. test R intensity. While the issue is primarily one
of sensitivity, the procedure of computing test-minus-
training scores is another way to eliminate differential
K values, and hence extract frustration effects.
The P5 condition was included primarily for two
reasons: (1) to provide a control condition which could
be used in the event that the five rewarded B presenta
tions in instrumental training resulted in maximal
11
strength of B r , and (2) as an alternate control-
o
condition subgroup to evaluate frustration effects. With
respect to the second comment, assume that difference
scores were formed for nonrewarded Ss which involved
subtracting the value of gE^ for the first test B from
that ]}E r associated with the second test B. The obtained
difference scores for nonrewarded children in Pi through
P4 would presumably reflect differential frustration only
while P5 difference scores would provide an estimate of
reaction potential differential for Ss which had never
been rewarded. If each of the subgroups was then con
trasted with P5, these second-order gEg difference scores
should have the following values: Pi - 6500 - 100 « 6400
P2 - 2500 - 100 - 2400; P3 - 700 - 100 - 600; and P4 -
500 - 100 * 400. Thus the prediction follows that the
associated R-amplitude differentials should be ordered:
Pi > P2 > P3>P4, when compared to a single control, P5.
Third and Fourth B Test Trials
Identical predictions obtain for Ss which were
subjected to two additional B events in test. That is,
the rank ordering of R amplitude should once more be:
Pl>P2 > P3 > P4 (> P5) , on each further B presentation.
First and Second D Test Trials
Table 3 presents a summary of the theoretical
analysis for the first D test trial. Since D was never
paired with marble delivery, the value of DHrg would be
low for all Ss. Hence a value of one was assigned for
each subgroup (a value of zero would also have been
TABLE 3
THEORETICAL VALUES AT THE TIME OF RESPONSE ON THE
FIRST D IN TEST FOR PREVIOUSLY NONREWARDED
AND REWARDED SUBJECTS
Subgroup b“r D«R
_H
D r g
Dw Df
K
Dt
der
Ss REWARDED ON PREVIOUS B TEST TRIALS:
PI 100 100 1 5 0 1 6 600
P2 100 100 1 5 0 1 6 600
P3 100 100 1 5 0 1 6 600
P4 100 100 1 5 0 1 6 600
P5 100 100 1 5 0 1 6 600
Ss NONREWARDED ON PREVIOUS B TEST TRIALS:
PI 100 100 1 5 32.5 1 38.5 3850
P2 100 100 1 5 12.5 1 18.5 1850
P3 100 100 1 5 3.5 1 9.5 950
P4 100 100 1 5 2.5 1 8.5 850
P5 100 100 1 5 0.5 1 6.5 650
I
u>
vo
40
reasonable). For children exposed to previous nonrewarded
B events, it was assumed that perseverative frustration
effects would be noted on D-trial performance. It is
conceivable that the onset of D after a number of prior
B nonrewards would produce a degree of Df dissipation,
but, assuming that Df does not decline in such a way as
to reverse the rank ordering of previous Df strength, one
would expect a continuation of the relative Df values as
stated in Table 2. However, due to the possibility of a
Df reduction for all subgroups, each Df quantity in
Table 2 was arbitrarily halved and inserted in Table 3.
For Ss which had only rewarded B test trials, no persev
erative frustration would be expected; thus a quantity
of zero was assigned to Df for such children.
If a similar line of argument is followed as was
developed for the B test trials, the differences
(previous B nonreward minus previous B reward) for the
various pairing subgroups would be: Pi « 3850 - 600 «
3250; P2 - 1850 - 600 - 1250; P3 - 950 - 600 - 350;
P4 - 850 - 600 - 250; and P5 - 650 - 600 - 50. The predic
tion would be that the relative enhancement in R strength
attributable to perseverative frustration for the pairing
subgroups was: PI > P2 > P3 > P4 (> P5). Also the
41
alternate method of contrasting PI through P4 differences
with P5 once again would result in the prediction:
PI > P2 > P3 > P4, when compared to the single control,
P5.
A parallel line of reasoning for the second D test
trial results in identical predictions of subgroup order
ings. Due to the highly similar logic involved, no
description of the analysis is considered to be necessary.
Analysis of Expectancies as Related to Extinction
Amsel has contended that resistance to extinction
is inversely related to the magnitude of frustration, and
thus is inversely related to reward expectancy strength.
Reference to Table 1 shows that all Ss were exposed to
repeated, nonrewarded B events (maximum of about 185
trials) during the extinction phase. Since Ss in the
various pairing subgroups presumably had different expec
tancies of reward, and hence various amounts of frustra
tion, the prediction of subgroup ordering on a trials-to-
extinction criterion was: Pi < P2 < P3 < P4 (< P5).
42
Experimental Hypotheses Stated
in Operational Form
The essentials of the experimental predictions have
been stated at different places in the preceding sections,
but are restated more concisely at this time (Table 1
should be consulted again in order to clarify the hypothe
ses) . It should be noted that whenever a group designation
is mentioned, such as PI, a more accurate phrase would be
"a child in subgroup PI,” or a modification thereof con
sistent with intended meaning.
Hypothesis 1
If children have been trained according to the plan
given in Table 1, then the resulting increments in R
intensity on B test trials (after the first test B) will
have the ordering: PI > P2 > P3 > P4 ( > P5).
Hypothesis 2
If children have been trained according to the plan
given in Table 1, and if frustrative events (nonrewarded
B events) are followed closely in time by a stimulus-
response event which has never been rewarded (the D event
following the final test B), then the relative enhancement
43
of R intensity to D should have the following order:
PI > P2 > P3 > P4 ( > P5) .
Hypothesis 3
If children have been trained according to the plan
given in Table 1, then resistance to extinction for the
pairing subgroups will have the order: Pi < P2 < P3 < P4
( < P5).
It will be recalled that an alternate means of
testing the basic hypotheses was developed; namely, the
technique of contrasting PI through P4 changes to P5
alteration. Since this method of evaluation only repre
sents a second way of examining the data, and does not
rest upon a radically divergent theoretical analysis,
related operational forms of the hypotheses are not
formally presented.
CHAPTER IV
METHOD
Subjects
The sample consisted of 54 boys and 36 girls
(N « 90) obtained from a local elementary school. The
ages were distributed as follows: 10 were nine years of
age, 49 were eight years old, 30 were seven years of age,
and 1 was six years old. Data on class level (grade)
showed that the sample was comprised of 42 children in
the second grade and 48 children in the third grade.
As will be discussed in a later section, the same propor
tional representation of boys and girls (3:2 in favor of
boys) occurred for each treatment subgroup, the ratio
having been preselected on the basis of subject availa
bility.
Apparatus
The apparatus was a modified version of that
described by Longstreth (1966). Basically, the equipment
44
45
consisted of three major components: a stimulus-response
unit, a programming unit, and a recording unit. The
stimulus-response unit was a flat-black box, the front
measuring 44 in. in height and 34 in. in width. The
center of it housed an 8-in. square milk-glass stimulus
window, which was approximately at S/s eye level.
Either of two light intensities could be presented:
bright, B, having a value of 55.0 ftc., or dim, D, having
an associated intensity of 1.0 ftc. The stimulus data
were obtained by positioning a Weston Master IV Light
Meter one ft. from the center of the stimulus window and
reading reflected light intensity. Background lighting
was 0.6 ftc.
A response handle (2.25 in. long, 0.75 in. in
diameter) was situated directly below the stimulus window
and was capable of being turned in a counterclockwise
direction only. From an initially vertical orientation
the response handle could be turned through a maximum arc
of 105 degrees, at which point a stop prevented further
movement. When released, springs returned the response
handle to the original position. The force required to
hold the lever at full displacement was 20.0 lb. (as
measured with a Viking Jr. Model 890 Spring Scale).
To the right of the stimulus window a 29.5 in.
long, 1 in. diameter, transparent plastic tube was
mounted in a vertical position. The tube was clamped to,
and actually was supported by, a metal funnel arrangement.
The tube served as a receptacle for ejected marbles; in
addition, the experimenter (E) could manually insert a
desired quantity of marbles in the tube via the funnel
access. A piece of 0.35 in. wide red tape was wrapped
around the upper end of the plastic tubing, and a rubber
stopper was placed in the bottom of the tube to retain
the marbles. Located inside the stimulus-response unit
was a large receptacle where up to 69 marbles were held.
As required by the design, single marbles could be ejected
by a solenoid, either independently of handle turning (R)
or as a result of S performing R.
The programming unit consisted of complex relay
and timing circuitry, one function of which was to auto
matically activate the stimulus light behind the window
at the required intensity (either B or D) according to
predetermined time relationships. Secondly, E could pre
set certain switches so that a marble was, or was not,
delivered on a given trial.
The recording unit was a two-channel Offner Dyno-
47
graph Amplifier-Recorder, Type 542. A single channel was
connected to a linear-taper potentiometer, which was
physically linked to the shaft of the response handle.
The resistance value of the potentiometer varied directly
with handle-turning displacement, and thus the displace
ment of the first-channel recording pen was directly
proportional to R amplitude. Two marker pens indicated
the beginning and end of a light presentation and the
occurrence of marble delivery. Paper speed throughout
the major portion of the experiment was set at 10 mm./sec.,
so that response latency could be read to the nearest
tenth of a second. The primary dependent variable was
handle-turning amplitude (measured in millimeters of pen
deflection from an arbitrary baseline). Response latency,
as described above, served as a subsidiary measure. Full
scale pen deflection was 30 mm. on the recording paper.
Several additional items were used during the ex
periment. A large “prize," having dimensions of 13 in. x
8 in. x 5 in., and weighing 3 lb., 13 oz., was handled and
inspected by I S at the start of the session. The package
was wrapped in shiny gold paper, had three red bows, and
was tied with a single strip of 1.5 in. red ribbon. The
actual prize given to each £ (presented after all children
48
had finished the experiment) was one of a number of gifts,
such as marbles, kites, and necklaces. Other materials
included a chair for S , a movable screen which provided a
visual barrier between E and £, and 69 steel marbles,
each having a 0.625 in. diameter.
Procedure
The E proceeded to a classroom, selected S_ accord
ing to a prepared list, and returned with the child to
the experimental room (20 ft. long, 8 ft. wide). At this
time £ was seated on a low chair in front of the stimulus-
response unit.
Stimulus-naming Phase
The E then recited the following instructions:
"Today we are going to play a game. If you do well, you
will win this prize (S^ then lifted and inspected the
prize; subsequently E placed the package on a stand just
to the right of the stimulus-response unit, and in clear
view of f3) . Now, do you see this window here; sometimes
the light will be dim like it is now, and other times it
will be bright like this. Watch for the light to come on
in the window, and when it does, say out loud if it is
49
bright or dim. I am going out now. Remember, watch for
the light, and say out loud if it is bright or dim.1'
The E moved behind the screen and initiated the
sequence: BDDBBDBDBDDDBBDBDDBB. The interstimulus interval
(time between the cessation of one stimulus and the onset
of the next stimulus) was five seconds. The same quasi
random sequence was used for each Sf the restrictions
having been that 10 B and 10 D events were represented,
and that no more than three consecutive stimuli of the
same kind appeared in the series. The reason for including
the stimulus naming procedure was to permit S to acquire
the simple verbal discrimination, "bright” vs. "dim."
It was assumed that by having £[ indicate the presence of
B and D, attention would be maintained to the relevant
stimuli. After the last stimulus had been presented, E
told S to rest for a minute, and proceeded to prepare for
the next phase.
Pairing Phase
The E once again confronted the child and said:
"You are playing the game very well, so let's change it
so it's more fun. Do you see this tube here; sometimes
marbles will drop in the tube, and each time a marble
50
drops I want you to say out loud marble. If the tube gets
filled all the way up to this red mark, one marble on top
of another, then you win the prize. Also, the light will
come on sometimes. As soon as it does, say out loud if it
is bright or dim. Remember, watch for marbles, and for
the lights.”
The £ then returned to the programming unit and
activated the sequence: DDBBDBBBBDDBDBBDBBBDBDDBBBBDB
DDBBBBDBBBDBBBBDBDBDDBBBBDBBDBBDBDBBBBDDBBDBBBBDBBBD
BBBBDDBBB. As was the case for stimulus naming, the
interstimulus interval was five seconds. The sequence
was randomly chosen with the stipulations that 60 B and
30 D events occurred, and that no more than either four
consecutive B presentations or two adjacent D events were
generated. All Ss were given the same sequence of stimulus
intensities.
Reference to Table 1 indicates that different num
bers of B-marble pairings occurred for the subgroups PI
through P3, whereas marbles were never associated with B
for P4 and P5. A pairing consisted of the onset of B,
a verbal response of "bright” by S, and marble ejection
accompanied by B termination. The time between initiation
of B and marble delivery was approximately one second.
51
By having the children say “marble," it was assumed that
not only would marble delivery be noticed, but attention
would be continuously directed to the task, and the fact
that D was never associated with marbles would be evident.
For Pi children, each of the 60 B events was paired with
a marble. The Ss in P2 received marbles only on the final
20 B events, and for P3, B-marble pairings were limited to
the last two B occasions. The assumption was made that
differential r strengths had been established (due to the
8
varying frequency of pairing), and at the same time all
Ss had been exposed to equivalent stimulus experiences.
Upon completion of the pairing sequence, E entered
S's portion of the experimental room and indicated: “I am
going to put some marbles in the tube now." Then marbles
were manually introduced into the tube by the experimenter
so that each £ would be about 10 marbles short of the red
mark (the goal) by the time the test phase was begun.
The Ss in subgroups P2, P3, P4 and P5 had 40, 58, 60 and
65 marbles, respectively, placed in the tube by E. For
children in PI the statement was omitted, and no marbles
were given. This procedure theoretically resulted in
slightly discrepant distances from the goal, since some
Ss in certain subgroups received further marbles during
52
the test, while others did not. However, the variability
in the way that marbles became stacked upon one another
in the tube was sufficient to offset any systematic
differences.
Instrumental Training Phase
The response handle, which had been concealed up to
this time by a small box, was uncovered. The S, was then
informed: "Now we are going to play the game differently.
Do you see this handle here; as soon as a light comes on,
you turn the handle to the left, this way. Then wait for
the light to come on again, and then turn the handle
again. Also, when the light comes on, say out loud if it
is bright or dim, and if a marble drops in the tube say
marble. Remember, if you can fill the tube with marbles
up to this red mark, then you win the prize. Not every
body wins; only those who play very, very well can win.
You tell me if you get enough marbles to win, okay? What
is the light now? Turn the handle once to see how it works
(S performed R once to a D). Good. 1 am going out again,
and when I say start you can begin turning the handle.
Start." The stimulus light remained on until S turned the
handle, at which time a marble was (or was not) delivered
53
immediately, and at the same instant the light was
terminated.
A total of 55 D and 5 B events were then presented,
with B occurrences taking place on trials 5, 29, 33, 56
and 59. The B trials were intentionally distributed in
the series so as: (1) to minimize the chances of reward
expectancy strengths reaching maximum levels; that is, if
the five B events had taken place within a brief period,
then any effects of the previous pairing phase might have
been effectively masked, and (2) to accustom Sa (except
P5 children) to receiving B-R-marble sequences, so that
R amplitudes were not highly variable (or at asymptotes)
when the test phase began. The interstimulus interval was
reduced to two seconds, a value which then continued
throughout the experiment. The two-second interval has
been found to be adequate for detecting perseverative
frustration effects (Longstreth, 1966); the previous five-
second duration was obviously longer, and represented an
attempt to allow Ss an adequate time to perform possible
covert verbal response sequences, such as, "bright-marble-
win prize."
Early in instrumental training E said only once:
‘•You don't have to turn the handle that hard." The
54
procedure was necessary since initial R amplitudes were
near maximum. Had this trend been allowed to continue,
the probability of noting any frustration effects would
have been slight due to a ceiling effect. The statement
by E was not repeated, however, since any potential enhance-
»
ment attributable to frustration might have been removed
on the basis of instructions. The technique was indeed
successful in reducing the high-level responding observed
on the first few trials.
Test Phase
The test phase began unannounced. As shown in
Table 1, one quarter of the Ss in each of the PI, P2, P3
and P4 subgroups received the sequence BBBBDD, B events
not being associated with marbles. A second quarter had
an identical series, but marbles were delivered following
B and a response. A third and fourth quarter each had the
sequence BBDD, with analogous reward schedules to those
stated for the first and second quarters. Subgroup P5
was divided into two equal sections, one having BBBBDD and
the other BBDD; no marbles were presented. As before, D
was not paired with marble ejection for any subgroup.
55
Extinction
There was no interruption between test and extinc
tion, and each S was exposed to repeated B presentations,
but R never led to marble delivery. The extinction series
was continued until ceased responding voluntarily, or
until 184 responses (186 for T2 Ss) had occurred. After
64 trials had elapsed, E stopped S and stated: "Remember,
everyone can't play the game well enough to win. If you
want to stop, you can. You tell me if you want to stop
playing."
At the conclusion of extinction, S > was told: "You
have done very well, but I see that you have not filled
the tube with marbles. You may have won the prize anyway
since you played so well, but we'll have to wait and see
how the other boys and girls do. Remember, you played
very well and you have a good chance to win the prize."
The S, then returned to the classroom and escorted the next
child back to the experimental room. Normally the time to
process one S was about 45 minutes.
Design
Many aspects of control and methodology have been
mentioned in the preceding sections. It is desirable now
56
to state design characteristics in a more formal way.
From inspection of Table 1 it should be apparent
that the treatment subgroups PI, P2, P3 and P4 define a
dimension which may be designated pairing condition. The
P5 subgroup was basically a secondary control group which
served to evaluate selected questions, and hence may be
considered separately. A further variable was that of
test reward condition, reflecting either reward or non
reward on B test trials. A final dimension referred to
the number of B events during test, either four (T4
condition), or two (T2 condition). Thus the fundamental
design was a 4 x 2 x 2 complete factorial arrangement with
a supplementary control condition, P5.
The assignment of Ss to the treatment conditions
was essentially random with several restrictions. The
requirements were that representation within each cell of
the basic matrix be: (1) equal in terms of sample size,
and (2) proportional with respect to sex. The ratio of
3:2, with boys favored, was selected mainly due to the
fact that there were more boys available. A total of
three boys and two girls were subjected to each treatment
combination, the final over-all sample size for the basic
4x2x2 design being 80. The P5 subgroup consisted of
57
10 Ss, assigned in an identical manner. For P5, only
nonrewarded B presentations occurred during test.
No attempt was made to balance the treatment sub
groups with respect to age or class level. Examination
of relevant data after assignment of Ss revealed that
mean age was approximately equal for FI through P5.
Additionally, proportional representation of class level
was highly similar for each of the pairing subgroups.
After one entire replicate (one S per treatment
combination including P5) of the experiment has been
processed, it was decided to alter the procedure in
several ways. The five B events in instrumental training
were redistributed so as not to take place only during
the terminal ten trials. The interstimulus interval was
increased from two seconds to five seconds throughout
stimulus naming and pairing. And finally, the pairing-
phase stimulus ordering was changed so as to randomize
B and D occurrences, rather than merely exposing j3 to
60 consecutive B trials. The latter method possibly
involved a greater degree of differential boredom for
the pairing subgroups. The stated procedural modifications
held for the final four replicates of the experiment,
58
and are the same as described in previous parts of this
chapter.
CHAPTER V
RESULTS
Three response measures were used to evaluate the
experimental predictions. The first was amplitude of
handle turning (R) , and was measured in terms of milli
meters of pen deflection from an arbitrary baseline. The
second response variable was that of response speed (the
reciprocal of response latency as read to the nearest
tenth of a second from the recording paper). Finally,
a simple count was determined of the number of Ss who
attained the extinction criterion. Verbal responses made
by Ss during the course of the experiment are discussed.
The preselected significance level was a ■ 0.05 for all
statistical analyses.
Preliminary Analyses
R Amplitude
In order to investigate several secondary questions
of interest, an estimate was determined of the relation
ship between mean handle-turning intensity and the course
of instrumental training. A mean amplitude score was
calculated for each of the following blocks of two trials
(the letter designates the type of stimulus event, either
59
60
B or D, and numbers indicate the trials which were
averaged): 0(14,15), D(24,25), B(29,33), 0(34,35),
0(44,45), D(54,55), B(59,61), and D(58,60). The D-trial
blocks were chosen so that approximate fifths of instru
mental training were represented, hence a general impres
sion was obtained regarding mean R-intensity fluctuation
as a function of trial block. Since E was still in verbal
communication with Ss during the first 13 trials, such
trials were not considered. The two B-trial blocks were
included to render information which was relevant to
implications derived from incentive-motivation theory.
Also, B-trial data were necessary in order that estimates
of prefrustration performance be established.
Only the final four replicates of the experiment
were considered due to the procedural differences already
mentioned; that is, only the last four Ss assigned to each
of the treatment conditions received exactly the same
instrumental training sequence. Hence it was decided to
omit (at this point) Ss appearing in the first replicate.
Figures 1, 2 and 3 present mean R amplitude as a function
of instrumental training trial block for Ss receiving
four B events in test (T4 condition), two B events in
test (T2 condition), and for the T4 and T2 conditions
combined, respectively. The legends on all figures
convey, in an abbreviated form, the number of B-marble
pairings administered as well as the number of times B
MEAN RESPONSE AMPLITUDE I N MM.
14.0 -
13.0
12.0
11.0
10.0
Treatment Conditions
Pi (60 B+ - 5 B+)
P2 (20 B+ - 5 B+)
D(14,15) D(24,25) B(29,33) D(34,35) D(44,45) D(54,55) B(59,61) D(58,60)
INSTRUMENTAL TRAINING TRIAL BLOCKS
Figure 1. Mean Response Amplitude in Mm. as a Function
of Training Trial Blocks for the T4 Condition
On
I - *
MEAN RESPONSE AMPLITUDE IN MM
14.0
13.0
12.0 -
11.0 -
10.0 -
9.0 ■
8.0 -
Treatment Conditions
PI (60 B+ - ■ 5 B+)
o--o
P2 (20 B+ - • 5 B+) A —— A
P3 ( 2 B+ - ■ 5 B+) • — - •
P4
( o
B+ -■ 5 B+) A— —A
P5 ( o
B+ - • 0 B+)
□ — — □
7.0
D(14,15) D(24,25) B(29,33) D(34,35) D(44,45) D(54,55)
INSTRUMENTAL TRAINING TRIAL BLOCKS
B(59,61) D(58,60)
Figure 2. Mean Response Amplitude in Mm. as a Function
of Training Trial Blocks for the T2 Condition
cn
N>
MEAN RESPONSE AMPLITUDE I N MM
14.0
13.0 -
12.0 -
11.0 -
10.0 -
9.0
8.0 -
7.0
Treatment Conditions
PI (60 B+ - 5 B+)
o—— o
P2 (20 B+ - 5 B+)
A — — A
P3 ( 2 B+ - 5 B+)
• — •
P4
( o B+ - 5 B+)
A—— A
P5 ( o B+ - 0 B+) □ — — □
D(14,15) D(24,25) B(29,33) D(34,35) D(44,45) D(54,55)
INSTRUMENTAL TRAINING TRIAL BLOCKS
B(59,61) D(58,60)
Figure 3. Mean Response Amplitude in Mm. as a Function of Training
Trial Blocks for T4 and T2 Conditions Combined
64
was reinforced during instrumental training. For in
stance, the notation "60 B+ - 5 B+" for PI indicates that
60 B-marble pairings occurred, and that performing R in
the presence of B resulted in marble ejection on each of
five instrumental training trials.
Inspection of Figure 3 would indicate that mean R
amplitude increased for all four experimental conditions
subjected to various numbers of B-marble pairings (PI, P2,
P3, and P4) from block D(24,25) to block 0(34,35); this
was not the case for P5, the latter actually declining in
mean R intensity. It is plausible that the noted eleva
tions for the PI through P4 conditions on the D(34,35)
block were due to the presentation of B-marble events on
trial(s) 29 and/or 33. It is not likely that such incre
ments were simply attributable to either a change in stim
ulus conditions per se or continued training, since the
P5 subgroup yielded a decrement in R intensity. It will
be recalled that P5 Ss never received marbles during
pairing or instrumental training. Mean response amplitude
also increased on blocks B(29,33) and B(59,61), which
involved presentation of B, as compared to immediately
preceding D-trial blocks; this statement applies particu
larly to data for conditions other than P5.
Such findings are compatible with an incentive
motivation interpretation: if one assumes that at least
some degree of marble expectancy had developed in the
65
situation, then presentation of B should theoretically
have elicited an anticipatory verbal 'Hnarble" response
along with characteristic internal stimulation (s ), lead-
9
ing to increased incentive motivation (K), and finally to
elevated response amplitude to B presentations. Then,
through the mechanism(s) of stimulus generalization and/or
perseverative drive enhancement, the observed intensifica
tion of R following a D event (an event never previously
associated with marble delivery) is not unexpected. Similar
elevations may be seen in Figure 1 for block D(54,55) as
compared to block D(58,60), but the comparable data in
Figure 2 reveal that mean P5 amplitude increases also;
hence the findings are not entirely consistent.
In order to evaluate statistically the trends of
mean R amplitude for the various treatment conditions at
selected times during instrumental training, a series of
Type III analyses of variance (Lindquist, 1953) were per
formed. Again, only the last four replicates were employed
for analytical purposes. The first analysis considered
instrumental training D-trial block as a within-subjects
variable, the D blocks being: D(14,15), D(24,25), D(34,35),
0(44,45), D(54,55), and D(58,60). The first between-
subjects factor was that of pairing condition (PI, P2, P3
and P4), while the second may be referred to as the test
reward condition; that is, during the test phase Ss in the
particular subgroup were to receive presentations of B
66
followed by R, and either nondelivery of marbles (frustra
tion) , or delivery of marbles following R (control). The
test reward-by-pairing interaction was a between-subjects
effect, while all other interactions were within-subjects
effects. Table A in the Appendix presents a summary of the
analysis which considered T4 Ss only. The main effects
were nonsignificant (p >0.05), the same being true for all
interaction terms. Table B in the Appendix presents the
results of an identical analysis except that only T2 Ss
were used. The only significant main effect was that
attributable to D-trial block (0.01 > p > 0.005), and indi
cated that mean R amplitude was not the same for all D-trial
blocks (disregarding pairing condition and test reward con
dition) . The means were 11.03, 11.42, 12.52. 10.89, 11.77
and 12.16 for consecutive blocks with the underlined values
indicating preceding B trials. Analysis of the differences
between means^ suggested that the following pairwise
*The typical analysis consisted of a two-tailed £
test using the formula:
MX - M2
** (2) MSerror(w)/bcn
where Mj_, * the two mean R values considered
MSerror(w\ " t^e within-subjects error term from
the Type III analysis
b - number of levels of the first between-subjects
variable
c * number of levels of the second between-subjects
variable
n ■ » number of observations per cell.
67
comparisons between D-trial blocks were not the same in
mean R amplitude: D(14,15) and D(34,35), D(34,35) and
D(44,45), and D(44,45) and D(58,60). These data lend
support to the hypothesis that interspersed B events during
training did enhance mean reactivity to a D event which
followed closely in time, and, additionally, that mean R
intensity appeared to return to levels which were observed
prior to the introduction of B events. While the two
analyses of variance do not prove that training differ
ences did not exist (or develop) for the two treatment
conditions (pairing and test reward), the lack of signifi
cant interactions, in particular, provide no data which
contradict such assertions. If it can be assumed that
the mean R amplitudes were essentially homogeneous for the
subgroups, then it cannot be argued that differences noted
during test were attributable to differences which existed
during the instrumental training series.
As will be apparent when the analysis of test trials
is presented, it was necessary to select pretest estimates
of R amplitude against which to compare test performance.
The mean of the R value associated with the last B in
training and the R value elicited by the first test B was
calculated for each S, and these means served as the pre
test estimates. The first test B was not a trial on which
frustration effects would be detected since R was made
before nondelivery of marble was apparent; the second test
68
B provided the first occasion for potential observation
of response facilitation. Another possibility would have
been to use the mean of the R amplitudes associated with
the two B presentations which appeared about midway through
the instrumental training sequence. Two further Type III
analyses of variance were performed, one for T4 Ss and
the other for T2 Ss, which considered B-trial block as
the within-subjects variable, that is, block B(29,33) vs.
block B(59,61). The between-subjects variables included
test reward condition and pairing condition, as previously
defined. Tables C and D in the Appendix summarize the
findings: all main effects and interactions were non
significant (p > 0.05). There appeared to be a general
tendency for mean response values to be higher on the
B(59,61) block, although the observation did not prove
to be statistically reliable. While not unequivocal,
it would seem safe to assume that the various treatment
subgroups did not differ in mean R amplitude on the two
B-trial blocks, and that test performance was probably
not due to any existing differences in R intensity prior
to test.
Pretest estimates of mean D amplitude were also
calculated. The mean of the response amplitudes associated
with the last two D events prior to test was the comparison
value used for each :S. An alternate estimate could have
been the mean R amplitude for trials 54 and 55. If mean
69
R intensity for the various subgroups interacted with
the particular trial block estimate used, then any obtained
results in the test phase might well be due to this factor
alone. Thus it was considered desirable to perform two
further Type III analyses, with the within-subjects vari
able being that of D-trial block, and test reward condition
and pairing condition serving as the two between-subjects
factors. Table E in the Appendix outlines the results for
T4 Ss, and Table F presents the summary for T2 Ss. Only
one main effect was significant (0.005 > p > 0.001), that
being attributable to D-trial block for T4 Ss. The mean
response amplitudes for the D(54,55) and D(58,60) blocks,
disregarding the variables of test reward condition and
pairing condition, were 9.83 and 11.27, respectively.
The increase in mean amplitude may have been in part the
result of perseverative incentive motivational effects
from the B presentation on trial 59. The corresponding
mean values for T2 Ss were 11.76 and 12.16, so at least
the directional relationship was consistent. It was con
cluded that no statistical evidence was found which would
support the contention that the treatment subgroup means
differed in ways which would probably bias the test
analysis.
R Speed
The relationships between the training-trial-block
70
variable (as defined previously), and mean response speed
for the various pairing conditions are portrayed in
Figures 4, 5 and 6. Only T4 Ss are considered in Figure 4,
T2 Ss in Figure 5, and results for T4 and T2 Ss combined
appear in Figure 6. As was the case for the amplitude
response measure, comparisons are for the last four repli
cates of the experiment. In general, it would appear that
mean response speeds increased over trial blocks with the
obvious exception of D-trial blocks immediately following
B-trial blocks— on these occasions mean response speeds
were generally lower than those R speeds which appeared
for D-trial blocks just prior to the introduction of B
events. Apparently, the alterations are in part due to a
change in stimulus conditions alone, but there is a sug
gestion that the modifications are more extreme for chil
dren who received marbles as a result of the instrumental
response to B.
Statistical analysis proceeded in an identical
fashion to that indicated for R amplitude; that is, there
is an equivalent analysis in terms of response speed for
each R amplitude analysis. Tables G and H in the Appendix
present summaries of Type III analyses used to examine the
nature of the relation between mean response speed and
D block in instrumental training, for T4 and T2 Ss,
respectively. In both cases, the main effect attributable
to D-trial block was significant (p < 0.001). A series of
MEAN RESPONSE SPEED I N 1/SEC.
1.30 Treatment Conditions
1.20 -
1.10
1.00 -
0.90
0.80
0.70
0.60
Pi (60 B+ - 5 B+)
o—— O
P2 (20 B+ - 5 B+) A A
P3 ( 2
B+ - 5 B+)
• — — •
P4
( o
B+ - 5 B+) A— — A
P5 ( o
B+ - 0 B+) □ — — a
D(14,15) D(24,25) B(29,33) D(34,35) D(44,45) D(54,55)
INSTRUMENTAL TRAINING TRIAL BLOCKS
B(59,61) D(58,60)
Figure 4. Mean Response Speed in 1/Sec. as a Function
of Training Trial Blocks for the T4 Condition
MEAN RESPONSE SPEED I N 1/SEC.
1.20 H
Treatment Conditions
1.10 -
1.00 -
0.90 -
0.80 -
0.70 -
0.60 -
0.50
1 1 1 1 1 1 1 ------------------------ 1 —
D(14,15) D(24,25) B(29,33) D(34,35) D(44,45) D(54,55) B(59,61) D(58,60)
INSTRUMENTAL TRAINING TRIAL BLOCKS
Figure 5. Mean Response Speed in 1/Sec. as a Function
of Training Trial Blocks for the T2 Condition
t o
MEAN RESPONSE SPEED I N 1/SEC.
1.30 -
1.20 -
1.10 -
1.00 .
0.90 -
0.80 -
0.70 -
0.60
Treatment Conditions
PI (60 B+ - 5 B+)
o— — o
P2 (20 B+ - 5 B+)
A A
P3
( 2 B+ - 5 B+) • — — •
P4
( o
B+ - 5 B+) ▲ ▲
P5 ( o B+ - 0 B+) □ — — □
1 1 1 1 1 1 1 1
D(14,15) D(24,25) B(29,33) D(34,35) D(44,45) D(54,55) B(59,61) D(58,60)
INSTRUMENTAL TRAINING TRIAL BLOCKS
Figure 6. Mean Response Speed in 1/Sec. as a Function of Training
Trial Blocks for T4 and T2 Conditions Combined
"vl
u>
►>o
t tests was performed (see footnote 1) on the mean response
speeds associated with D-trial blocks (disregarding the
between-subjects variables). The results indicate that on
the average Ss responded faster on either block D(44,45)
for T4 Ss, or block D(54,55) for T2 Ss, than on preceding
blocks. This increase is probably the result of practice
alone. Further, it would appear that mean response speed
declined from block D(54,55) to block D(58,60) for both
T4 and T2 Ss. These data suggest that the introduction
of a B event produces a systematic disruption in mean
response speed to a D which follows closely in time.
In conjunction with the results of amplitude analysis,
a reasonable alternate explanation to the incentive-
motivation account already stated might be as follows:
the presentation of a D event (one not accompanied by
marble delivery) closely following a B event (one asso
ciated with marble delivery) is "frustrating.” In other
words, the presentation of a stimulus never associated
with marble ejection is frustrating in the sense that the
D is indicative of the fact that no marble is possible
for the trial— an expected reward is not omitted in this
case; rather, the possibility of reward is negated. If
both operations result in increased drive effects, the
elevation in response amplitude is deducible, and the
noted decrease in speed might be the result of conflicting
75
response tendencies, that is, whether or not to perform R.
From an incentive-motivation theory, an increase in mean
response speed might be expected.
Turning now to comparisons of both B-trial esti
mates and D-trial estimates, treated in an identical manner
as that already described for R amplitude, Tables I and J
in the Appendix refer to the former estimates, while the
latter are presented in Tables K and L. The main effect
of pairing condition was significant (0.05 > p > 0.025)
for T4 Ss, the respective means being 0.911, 0.704, 1.117,
and 0.724 for Pi, P2, P3, and P4. Such was not the case
for T2 Ss; rather, the main effect associated with B-trial
block was significant (0.01 > p > 0.005); mean response
speed was 0.797 for block B(29,33) and 0.918 for block B
(59,61). The only significant result in Tables K and L
was that due to D-trial block (0.025 > p > 0.01) for T4 Ss.
Mean response speed declined from block D(54,55) to block
D(58,60), the respective means being 0.970 and 0.846.
This finding again lends support to the proposition that
mean D speed following a B trial is lower than mean D
values which were observed just prior to B. In conclusion,
there were no data which would suggest that the particular
estimate used would have an important influence on the
results of test-trial analysis.
76
Verbal Responses
Informal observations by E would support the
assertion that Ss acquired stable verbal discriminations
with regard to B and D stimuli. During preliminary train
ing and/or pairing, each £ made one or two errors in
naming stimuli. This error commission appeared early in
the experiment, and in only one or two instances did such
mistakes occur after pairing had been completed. However,
even for these latter isolated cases, the total number of
errors was extremely small. These data are important
due to the fact that such a discrimination potentially
was one basis for the hypothesized differential reward
expectancies.
Throughout the course of the pairing phase, E
recorded the occurrence of verbal responses which could
be classified as anticipatory. For example, if S said
“marble” after the presentation of B, but before marble
ejection, it could be argued that the relationship between
the occurrence of B and marble delivery had been estab
lished. In terms of theory language, it might be inferred
that conditioning of a verbal "marble” anticipatory
response had been accomplished. A total of seven children
in the Pi condition and two children in the P2 condition
displayed such verbal responses, but this was not true
for any member of P3, P4, or P5. It is tempting to con
clude that differential anticipatory habit strengths
77
existed for the various pairing conditions; however, it
must be pointed out that Ss in PI and P2 had a greater
number of occasions on which to potentially exhibit such
verbalizations. This fact alone might be responsible for
the differences noted. In any event there does seem to
be a suggestion that pairing had been effective in the
sense that an expectancy of marble delivery following a
B had probably developed for Pi and P2.
Tests of Hypothesis 1
It will be recalled that Hypothesis 1 was stated
as follows: if children have been trained according to
the plan given in Table 1, then the resulting increments
in R intensity on B test trials (after the first test B)
will have the ordering: Pi > P2 > P3 > P4. For the Ss
in T2, the second test B alone provided the basis for
testing the hypothesis; on the other hand, the second,
third, and fourth test B presentations for T4 Ss were
trials on which frustrative nonreward effects would be
expected.
Second B Trial in Test
Initially, a Type III analysis was performed in
which the within-subjects variable was that of training
vs. test condition. The training condition referred to
mean R amplitudes which were calculated for each S as
78
follows: the R Intensity associated with the last B in
instrumental training was added to the R amplitude dis
played on the first test trial, and the resulting sum was
divided by two. As stated before, the first test B was
not an occasion on which frustration effects would be
detected, since nonreward had not yet occurred at the time
of responding. The computed means served as the raw data
for further analysis, and provided a stable estimate of
response amplitude just preceding differential reward.
Data for the test condition were simply the R amplitudes
measured for the second test trial. Inclusion of the
training vs. test condition is analytically equivalent
to considering difference scores; either technique is
required to extract the theoretical effects of frustration.
Since TA and T2 Ss had received identical treatments up to
and including the second test trial, the relevant data
were combined. As defined previously, the first between-
subjects variable was test reward condition, either marble
ejection (control) or no marble (frustration) in test, and
the second was pairing condition (PI, P2, P3, and PA).
The first-order interaction between the latter two vari
ables was a between-subjects effect, all other interactions
being classified as within-subjects effects. Table M in
the Appendix summarizes the findings of the three-way
analysis of variance. The test reward condition main
effect was significant (p < 0.001), as was the case for
79
the main effect of training vs. test (p < 0.001). Even
more interesting from a theoretical standpoint was the
significant (p < 0.001) test reward-by-training vs. test
interaction. Table 4 presents the mean response amplitudes
when the pairing condition variable is ignored. Clearly
TABLE 4
MEAN RESPONSE AMPLITUDE IN MM. DURING LATE
INSTRUMENTAL TRAINING AND ON THE
SECOND TEST TRIAL3
Reward Condition
in Test (B Trials)
Training vs. Test Condition
Training Second Test Trial
Marble Ejection
(Control) 12.12 12.02
No Marble
(Frustration) 14.44 16.85
aT4 and T2 Ss combined
there was minimal change in mean R intensity from training
to the second test trial for control subjects, but a mean
change of 2.41 was observed for subjects which had marble
delivery omitted.
The noted enhancement in R amplitude, along with
the lack of same for the control condition, is the usual
frustration effect (FE) found when expected reward is no
longer present. The main effect of pairing condition was
80
not significant, the same being true for all interaction
terms which included the pairing-condition variable.
Hypothesis 1 would receive support had the triple interac
tion been significant (assuming predicted directional
relationships among the means); hence one can only con
clude that an FE was demonstrated, but the FE was not
different for the various pairing conditions.
Third B Trial in Test
Table N in the Appendix presents the results of
a nearly identical analysis performed for T4 Ss only,
which differed from that just described in the following
way: R characteristics were determined for the third B
in test, the remaining portions of the analysis being
unchanged. The main effect of test reward condition was
significant (p < 0.001), indicating a difference between
mean R amplitude for marble delivery (control) as compared
to the marble-omission condition (disregarding both pair
ing and training vs. test distinctions). The means were
11.69 for controls, and 16.06 for the omission condition,
respectively. The main effect attributable to training
vs. test condition was also significant (p < 0.001), as
was the case for the training vs. test-by-test reward
interaction (p < 0.001). Table 5 indicates mean response
intensity as a function of the indicated treatments.
81
TABLE 5
MEAN RESPONSE AMPLITUDE IN MM. FOR T4 SUBJECTS
DURING LATE INSTRUMENTAL TRAINING AND
ON THE THIRD TEST TRIAL
Reward Condition
in Test (B Trials)
Training vs. Test Condition
Training Third Test Trial
Marble Ejection
(Control) 11.88 11.50
No Marble
(Frustration) 14.02 18.10
The same conclusion is tenable for the third test
trial as was reached for the second test trial; that is,
an FE was noted, but did not appear to be reliably related
to the pairing treatment. However, there was a suggestion
of relationship in that the triple interaction was of
marginal significance (0.10 > p > 0.05). If the differ
ence between differences (test minus training and no
marble minus marble) in mean amplitude is calculated for
each pairing condition, the following mean values result:
6.40, 3.50, 0.80, and 7.10 for the Pi, P2, P3, and P4
subgroups. The trend is according to hypothesis except
for the reversal displayed for P4.
82
Fourth B Trial in Test
The results of the analysis of the fourth test
trial for T4 Ss (see Table 0 in the Appendix) reveal four
significant effects, including all three main effects,
and the interaction of training vs. test with test reward.
Mean response intensities for Pi, P2, P3, and P4 subgroups
were 12.10, 15.88, 13.25, and 13.22, when other variables
are ignored. Table 6 presents the means associated with
the significant interaction term. Once more an FE was
noted which was apparently independent of pairing con
dition.
TABLE 6
MEAN RESPONSE AMPLITUDE IN MM. FOR T4 SUBJECTS
DURING LATE INSTRUMENTAL TRAINING AND
ON THE FOURTH TEST TRIAL
Reward Condition
Training vs. Test Condition
in Test (B Trials) Training Fourth Test Trial
Marble Ejection
(Control) 11.88 11.55
No Marble
(Finis trat ion) 14.02 17.00
An Alternate Approach: Dunnettfs Test
The preceding three-way analyses provide one method
for testing Hypothesis 1. A possibly more powerful alter
83
native is available, however; one that has been described
as the Dunnett method for comparing k means with a control
condition (see Winer, 1962, pp. 89-92). The Dunnett
technique enables a joint probability statement to be
attached to a collection of k comparisons, using a single
control. Inquiry is not directed to all possible differ
ences between means; only to those between the k treat
ments and the control. The usual analysis of variance
effectively does consider all possible differences between
subgroup means, which may lead to a less powerful test in
certain instances; for example, when several means have
the same intermediate value while two extreme means appear
to be quite divergent. The Dunnett approach is performed
independently of the outcome of the usual F test.
Difference scores were formed for the pairing
conditions (PI through P5) in the following manner: for
each S the mean R intensity estimate for late training
was subtracted from that R value appearing on the second
test trial, limiting the determinations to children sub
jected to the no-marble condition during test. By
assumption, Pi, P2, P3, and P4 subgroups should have
yielded mean increments in R amplitude attributable to
frustration, but to varying degrees. The P5 Ss should
have represented a nonfrustrated control, since marble
delivery had never been associated with B, D, or as a
result of R.
84
Analysis of the mean difference scores indicated
that Pi and P5 means were different (Jt ■ 3.00, df ■ 40,
two-tailed), while the P2-P5, P3-P5, and P4-P5 contrasts
were not; the joint summary statement had an associated
probability of less than 0.05. The respective mean diff
erences were 6.05, 3.35, 2.90, 3.35, and 0.15 for Pi,
P2, P3, P4, and P5 conditions. Such results suggest that
a reliable enhancement in R amplitude (an FE) was demon
strated for Pi Ss, but not for P2, P3, and P4 Ss, when
compared with a nonfrustrated control (P5). This finding
would certainly provide confirmatory evidence for Hypothe
sis 1, in that the subgroup (PI) which had the greatest
number of B-marble pairings (at least 40 more than any
other subgroup) displayed an FE. At the same time,
smaller increments in mean amplitude were noted for P2,
P3, and P4, but these elevations were not reliable from
a statistical standpoint. Perhaps the pairing procedure
produced a relatively high reward expectancy for PI, the
number of pairings for P2, P3, and P4 resulting in smaller
but equivalent expectancies.
Dunnett*s approach was then applied to the data
for the third and fourth test trials in an identical
fashion. Table 7 presents mean difference scores and
results of t tests for the various B test trials.
Test trial three results indicated that Pi, P2,
and P4 means were larger than the P5 mean, but P3 was not
85
TABLE 7
MEAN R-AMPLITUDE DIFFERENCE SCORES IN MM. AND
t TEST OUTCOMES FOR SUBJECTS RECEIVING
NO MARBLES IN TEST: SECOND THROUGH
FOURTH TEST TRIALS
Pairing
Condition
Test Trial Number
2 (T4 and T2 Ss) 3 (T4 Ss) 4 (T4 Ss)
PI 6.053 4.803 2.00
P2 3.35 4.62a 2.50
P3 2.90 0.80 2.20
P4 3.35 7.003 5.20
P5 0.15 -1.30 -1.10
aDunnett's 1 : was significant (p < 0.05) when the
designated subgroup was compared to P5.
different from P5. The appearance of the large mean
difference for P4 on the third and fourth test trials is
difficult to explain. By the fourth test trial no indi
vidual pairing subgroup was significantly different from
the P5 control, but elevations in mean R intensity were
still present for all conditions except P5. On the basis
86
of test trials three and four, no support for Hypothesis 1
was evident.
It should be noted that Dunnett's analysis using
the second B test trial was probably the single best test
of Hypothesis 1, since: (1) the sample size was largest
(T4 and T2 data were combined), and (2) conditioned frus
tration effects were minimal. On the other hand, the
majority of statistical analyses suggested that the vari
able of pairing condition was either not related to the
magnitude of the FE observed, or the relationship was not
according to prediction. Further discussion of the
findings appears in Chapter VI.
Tests of Hypothesis 2
Hypothesis 2 stated: if children have been trained
according to the plan given in Table 1, and if frustrative
events (nonrewarded B events) are followed closely in time
by a stimulus-response event which has never been rewarded
(the D event following the final test B), then the rela
tive enhancement of R intensity should have the following
order: Pi > P2 > P3 > P4. The primary test of Hypothesis
2 had reference to the first D test trial; for T4 Ss the
fifth test trial fulfilled the requirement, and the third
test trial was applicable for T2 Ss. The second D trial
in test was included in order to determine if generalized
or conditioned frustration effects continued past the
87
initial D event. The basic techniques used in the
evaluation of the first experimental hypothesis were used
again.
Analysis of T4 Data
The first Type III analysis of variance considered
an estimate of D performance late in training, as well as
the R intensity noted on the fifth test trial. As stated
previously, the estimate of pretest R intensity to D for
each involved the mean of the R amplitudes associated
with the final two D presentations in instrumental train
ing. This distinction defined the training vs. test
condition and served as the within-subjects variable.
As before, the between-subjects variables consisted of
test reward condition (whether or not marbles had been
delivered for B test trials), and the pairing condition
(PI, P2, P3, and P4). Table P in the Appendix summarizes
the results. The only significant (p < 0.001) main effect
was that of training vs. test, and the interaction of
training vs. test and test reward was the sole significant
(p < 0.001) interaction term. Table 8 presents the mean
R amplitudes associated with the interaction.
For control Ss the mean response intensity change
from training to test was 0.53, while the comparable
difference was 3.32 for Ss who had received no marbles
on B test trials. The results may be interpreted in terms
88
TABLE 8
MEAN RESPONSE AMPLITUDE IN MM. FOR
T4 SUBJECTS DURING LATE TRAINING
AND ON THE FIRST D IN TEST
Test Reward
Condition
(B Trials)
Training vs. Test Condition
Training First D Test Trial
Marble Ejection
(Control)
11.62 12.15
No Marble
(Frustration)
11.88 15.20
a frustration account: through the operation of general
ized and/or conditioned frustration, a response enhance
ment was observed for Ss who had received nonrewards a
short time before. The increment was noted on D trials
(never rewarded in the past), while the frustrative non
rewards occurred on B trials. The FE was not statisti
cally different for subgroups exposed to the various
numbers of pairing events, thus lending no support to
Hypothesis 2.
The related analysis of variance for the sixth test
trial (see Table Q in the Appendix) considered the same
variables mentioned above. Two significant (0.025 > p >
0.01) interactions appeared, namely, training vs. test
89
condition with both pairing condition and test reward
condition. The latter interaction was once more attribut
able to perseverative frustration, the means for training
and test being 11.62 and 11.05 vs. 11.88 and 13.10 for
control and non-marble conditions, respectively. The mean
R intensities for the former interaction effect are indi
cated in Table 9. It is interesting to note that only Ss
TABLE 9
MEAN RESPONSE AMPLITUDE IN MM. FOR T4
SUBJECTS DURING LATE TRAINING
AND ON THE SECOND D IN TEST
Pairing
Condition
Training vs. Test Condition
Training Second D Test Trial
PI 11.10 9.90
P2 13.10 13.00
P3 11.70 11.50
P4 11.10 13.90
subjected to P4 conditions continued to display facili
tated R amplitude on the sixth test trial, the other sub
groups essentially returning to pretest levels.
90
Analysis of T2 Data
The analyses for T2 Ss paralleled exactly those
performed for the T4 condition. Table R in the Appendix
presents the analysis concerned with the third test trial
(the first D in test for T2 Ss). Once more, the only
result of theoretical importance was the significant
(0.05 > p > 0.025) interaction between the training vs.
test condition and the test reward condition. Table 10
TABLE 10
MEAN RESPONSE AMPLITUDE IN MM. FOR T2
SUBJECTS DURING LATE TRAINING
AND ON THE FIRST D IN TEST
Test Reward
Condition
Training vs. Test Condition
(B Trials) Training First D Test Trial
Marble Ejection
(Control) 11.88 12.35
No Marble
(Frustration) 13.05 15.80
shows the mean R intensities during late training and for
the third test trial. As was the case for the T4 analysis,
an FE occurred which did not seem to be related to pairing
condition.
91
Finally, an analysis was done which examined mean
R intensity on the fourth test trial for T2 Ss (Table S
in the Appendix) . No significant effects were found, so
it is likely that frustration effects had dissipated.
This was a reasonable result, since T2 Ss experienced
two less frustrating events than did T4 Ss, the latter
still exhibiting mean response enhancement by the second
D trial in test.
The Dunnett Technique as Related
to Hypothesis 2
As described for the alternate tests of Hypothesis
1, difference scores were computed (test minus training)
separately for the D trials in test, and retaining the
T4-T2 distinction. Table 11 presents the mean R-amplitude
scores associated with the series of Dunnett's tests for
Ss who had received no marbles on B test trials.
The following conclusion applies equally for T4
and T2, as well as for both D test trials: according to
Dunnett's analyses, the pairing subgroups were not differ-
ent in mean R-amplitude change from training to test.
The apparently contradictory findings (compared to those
obtained by means of the Type III analyses) may be partly
due to the fact that quite different control groups were
used. Even though not significant, P4 difference scores
were higher than corresponding values for the other pairing
92
subgroups.
TABLE 11
MEAN R-AMPLITUDE DIFFERENCE SCORES IN MM. FOR
D TEST TRIALS: BASED UPON ONLY SUBJECTS
RECEIVING NO MARBLES DURING TEST
Number of B
Trials in Test
(T4 vs. T2)
Pairing
Condition
Test Trial
First D Second D
PI 3.40 0.20
P2 3.40 0.80
4 P3 2.10 -0.10
P4 4.40 4.20
P5 2.80 1.60
PI 2.30 1.50
P2 2.80 1.40
2 P3 2.50 -0.50
P4 3.40 2.00
P5 -0.90 0.10
Tests of Hypothesis 3
The third experimental hypothesis indicated: if
children have been trained according to the plan given in
Table 1, then resistance to extinction for the pairing
subgroups will have the ordering: PI < P2 < P3 < P4.
93
The extinction series consisted of repeated B presenta
tions never being followed by marble ejection.
Number of Subjects Reaching
the Extinction Criterion
A simple frequency count was made of the number of
Ss who voluntarily terminated the experiment (the extinc
tion criterion). The respective frequencies associated
with subgroups Pi, P2, P3, P4, and P5 were 5, 5, 5, 3,
and 2. The associated proportions for the stated frequen
cies were 0.25, 0.25, 0.25, 0.15, and 0.20. If an alter
nate criterion were adopted (response latency of five
seconds or more), the proportions generally increase
(0.30, 0.30, 0.25, 0.15, and 0.40), but not in a way
which would be suggestive of a differential relationship
to pairing condition. In retrospect, it is obvious that
nonrewarded trials were not continued a sufficient time
for most Ss to cease responding. No statistical analyses
were performed due to the very small sample sizes
involved.
Response Characteristics
during Extinction
Both response amplitude and response speed during
the extinction phase were examined. Since Ss did not
always have equivalent numbers of extinction trials, it
was decided to consider fifths of extinction. This process
involved determining the number of extinction trials on
which S responded, and in turn, recording response data
on trials which appeared on consecutive approximate fifths
of the total number of trials. For example, if £ stopped
on extinction trial 67, then extinction data points in
cluded trials 13, 26, 39, 52, and 66. As a reference
point, the previously determined estimate of late training
amplitude and speed were employed; that is, the mean
response value for the last B in training and the first
test B. Figures 7 and 8 depict mean amplitude and mean
speed, respectively, as a function of late training per
formance and extinction fifths.
In order to evaluate the response trends, a Type I
analysis of variance (Lindquist, 1953) was performed for
each response measure. The between-subjects variable was
that of pairing condition (Pi, P2, P3, P4, and P5).
Within-subjects effects included trial block (late training
and fifths of extinction) and the interaction of pairing
condition and trial block. A preliminary check on mean
amplitude and speed confirmed that trends were highly
similar for Ss reaching the extinction criterion and for
those Ss who continued to respond until the end of the
session; thus the relevant data were combined. Tables T
and U in the Appendix indicate the results of the anal
yses, the first for amplitude data and the second for
speed. No significant effects were discovered for
MEAN RESPONSE AMPLITUDE I N MM
Treatment Conditions
16.0 -
15.0 ■
14.0 -
13.0 -
12.0 -
11.0
PI (60 B+ - 5 B+) o—— O
P2 (20 B+ - 5 B+) A A
P3 ( 2 B + - 5 B+)
• — — •
P4 ( 0 B+ - 5 B+) A— ▲
P5 ( 0 B+ - 0 B+) a — — □
▲
i --------------- 1 --------------- 1 --------------- 1 --------------- 1 --------------- 1 --------
Tr 1/5 2/5 3/5 4/5 5/5
TRAINING < ------------ EXTINCTION FIFTHS---------------- ►
Figure 7. Mean Response Amplitude in Mm. as a Function ^
of Late Training and Extinction Fifths t-n
MEAN RESPONSE SPEED I N 1/SEC.
Treatment Conditions
1.70 -
1.50 ■
1.30 -
1.10 ■
0.90 -
0.70
PI (60 B+ - 5 B+)
o— — o
P2 (20 B+ - 5 B+) A A
P3 ( 2 B+ - 5 B+)
• — — •
P4 ( 0 B + - 5 B+)
▲ A
P5 (0B+- 0 B+)
□--— □
1 1 1 1 1 1 —
Tr 1/5 2/5 3/5 4/5 5/5
TRAINING ^ ------------ EXTINCTION FIFTHS ►
Figure 8. Mean Response Speed in 1/Sec. as a Function
of Late Training and Extinction Fifths
no
O'
97
amplitude measures, but several findings of interest
emerged as the result of speed analysis. First, the main
effect of pairing condition was significant (0.025 > P >
0.01), the associated means (disregarding trial block)
being 1.174, 1.006, 1.256, 1.105, and 0.853 for Pi, P2, P3,
P4, and P5. This result would suggest that P3 Ss responded
more rapidly on the average, and P5 Ss more slowly, than
did members of the other pairing subgroups. Of more
interest was the main effect attributable to trial block
(p < 0.001); apparently mean speed (ignoring pairing con
dition) was not the same for all trial blocks. The means
for successive blocks were: 0.913, 1.179, 1.176, 1.106,
2
1.143, and 1.106. Related _ t tests were performed on the
means just stated. It was found that mean late-training
response speed was significantly slower (p < 0.05) than
that mean speed for each extinction fifth, and that
extinction fifths did not differ from each other. From
2
The typical t^ test (two-tailed) was computed
using the formula:
Mi " M2
t » ... —
V (2) MSerror(w)/N
where
Mi, M2 * two mean R speeds
MSPT. rnr/u\ ■ the within-subjects error term from
the Type I analysis
N ■ the number of subjects.
98
the viewpoint of frustration theory, one would expect a
decrease in speed at some point in extinction due to
assumed conflicting response tendencies. The consistent
elevation of mean speeds over extinction fifths would
indicate that the increment is not an isolated event.
The analysis of variance provides no support for a
hypothesis of differential change in mean speeds for the
various pairing treatments as a function of trial block.
However, inspection of Figure 8 shows that mean speeds for
P5 were consistently slower than for the other pairing
subgroups, and additionally, there was no obvious eleva
tion in mean speed as compared to final training. The
failure to support such observations by statistical eval
uation is probably due to the relative insensitivity of
the analysis of variance when one of k groups is different
from all others, which are essentially equal.
Supplementary Analysis of Test Trials
For each analysis of variance mentioned in the
evaluation of test trial amplitude data, an equivalent
analysis was performed on response speed. Only those
analyses which had significant results are discussed.
Table V in the Appendix is concerned with mean response
speeds on the second test trial. As in the case of ampli
tude, data were combined for the T4-T2 distinction: the
99
outcome of the Type III analysis indicated that only the
main effect of pairing condition was significant
(0.05 > P > 0.01), the Pi through P4 means being 0.968,
0.805, 1.058, and 0.812 (ignoring test reward condition
and training vs. test) .
Similar analyses for the third and fourth B events
in test (see Tables W and X in the Appendix) once again
showed over-all differences in mean speeds for the pairing
subgroups. For the third test B, mean speeds were 1.047,
0.839, 1.215, and 0.852 for subgroups Pi to P4, while the
fourth test B data indicated analogous mean speeds of
1.094, 0.812, 1.260, and 0.934. Also, on both test trials
three and four, mean response speed increased from
training to test; the respective mean values (disregarding
other variables) were 0.890 and 1.086 for the third test B,
and 0.890 and 1.160 for the fourth test B. From such data
one could infer that frustrative nonreward produces at
least an initial increment in instrumental response speed.
One more analysis demonstrated significant out
comes, that is, the Type III analysis of variance, for
T4 Ss, which considered mean response speeds on the first
test D (see Table Y in the Appendix). Of primary interest
was the reliable (0.05 > p > 0.025) interaction of test
reward condition and training vs. test. Table 12 presents
the mean speeds associated with the significant interaction
term.
100
TABLE 12
MEAN RESPONSE SPEED IN 1/SEC. FOR T4 SUBJECTS
DURING LATE INSTRUMENTAL TRAINING AND
ON THE FIRST D TEST TRIAL
Reward Condition
in Test (B Trials)
Training vs. Test Condition
Training First D Test Trial
Marble Ejection
(Contro1) 0.954 0.787
No Marble
(Frustration) 0.762 0.870
It would seem that while the mean speed decreased for con
trol Ss, the opposite was true for the no-marble condition.
The difference between mean scores (test minus training)
was -0.167 for controls and 0.108 for Ss who received no
marbles on B test trials.
For nonrewarded trials in both test and extinction,
E recorded verbal responses made by Ss which were indica
tive of possible "awareness" of the B-marble relationship.
Typically, such an occasion consisted of the onset of B,
followed by R, and shortly thereafter a verbal response
of marble. Of course, no marbles were forthcoming on such
trials. A count was determined of the number of children
within each pairing subgroup who emitted such verbaliza
101
tions. For subgroups Pi through P5, the frequencies were
3, 1, 1, 4, and 0, respectively. Because of extremely
small values, no statistical analysis was attempted.
It is notable that P4 had the greatest observed frequency
which may be associated with the unexpected large increase
in mean response amplitude present on certain test trials
(see Table 7).
CHAPTER VI
CONCLUSIONS AND DISCUSSION
Conclusions
The results of statistical analyses, as well as
the more informal examinations of data, suggest the
following conclusions:
1. All Ss did acquire a stable verbal discrimina
tion between B and D, and there appeared to be some
evidence (the anticipatory verbal marble responses of
certain children in Pi and P2 observed during pairing
and instrumental training) that B did come to elicit
fractional anticipatory goal responses (rg's). Second,
supplementary analyses did not contradict the assumption
of relatively homogeneous performance for all pairing
subgroups during instrumental training; thus it is
improbable that the noted differences in test and/or
extinction were the result of pre-existing discrepancies.
2. With respect to Hypothesis 1, the data indi
cated that a reliable increment in response amplitude
attributable to frustrative nonreward (an FE) was present
102
103
on the second, third, and fourth B test trials; the FE
was not different for the various pairing subgroups if
examined by the analysis of variance method. When an
alternate technique was used (Dunnett's test), data con
sistent with Hypothesis 1 were found on the second B test
trial, but similar evaluations for the third and fourth
B test events yielded no confirmatory evidence. In fact,
the large enhancement in mean amplitude observed for P4
was contrary to prediction. Over-all, the assumption of
a positive relationship between reward expectancy strength
and the magnitude of frustration received only marginal
support.
3. Tests of Hypothesis 2 were essentially negative
in that no differential R amplitude performance for the
various pairing subgroups was observed on either the
first or second D in test. However, an FE was apparent
which was presumably due to perseverative generalized or
conditioned frustration (when the pairing variable was
disregarded). In general, there were no data which
supported Hypothesis 2.
4. Hypothesis 3 predicted that resistance to
extinction was negatively related to reward expectancy
strength. When a count was made of Ss reaching the extinc
tion criterion, it was evident that most children did not
cease responding in the time allowed. The pairing variable
104
was not obviously related to the number of Ss extinguish
ing, but since the total sample size involved was very
small, no statistical tests were performed. Examination
of instrumental response amplitude and speed during
extinction once more revealed no information which would
be consistent with an hypothesis of differential behavior
as a function of reward expectancy strength. The only
finding of interest was the observation of an over-all
enhancement of speed during extinction as compared to an
estimate of late training speed. Since the majority of
children did not meet the extinction criterion, perhaps
the most reasonable decision would be to suspend judgment
on Hypothesis 3.
5. Supplementary analyses of behavior on the test
trials indicated that the general result of frustrative
nonreward was an increase in mean response speed. An
Amsel-type frustration theory would predict such facili-
tative effects on early frustration trials prior to the
development of conditioned frustration, the latter leading
to a decrement in speed to a stimulus indicating forth
coming frustrative nonreward.
Discussion
The most unexpected result was the large increment
in mean amplitude, from training to test, for P4 Ss.
This was particularly evident on the third and fourth B
105
events in test, although a similar (but smaller) elevation
was detected on the first and second D trials in test.
It would seem to be paradoxical that a group which should
have had minimal expectancy of reward did display an FE
at least as large as that noted for PI Ste, a subgroup
which theoretically had typically greater reward expectan
cies. The obvious way in which P4 differed from Pi, P2,
and P3 subgroups was that P4 Ss had never received marbles
during the pairing phase, while the former children had
at least two B-marble pairings. For Pi, the presentation
of B had invariably been accompanied by marble delivery,
but P2 and P3 experienced both nonrewarded and rewarded
B events throughout the pairing phase. If frustrative
nonreward were suddenly introduced after all Ss had been
subjected to the same instrumental training, several
possibilities are apparent. The modification of reward
conditions would perhaps be more noticeable to PI and P4
Ss than for P2 and P3 conditions. To the extent that
children recognized the pairing phase and the instrumental
training-and-test phase as involving distinctly different
associations with reward, performance of Pi and P4 would
be disrupted relatively more than that for P2 and P3.
For PI, the greater modification would be due to a con
trasting consistent history of reward, while P4 had
experienced a nonrewarded series (pairing), followed by
a session of reward (instrumental training), and finally
106
nonreward once again (test). On the other hand, P2 and
P3 had been both nonrewarded and rewarded during the two
major phases, so perhaps this equivalence might have re
sulted in a relatively smaller FE than would be expected
on the basis of expectancy manipulations. It is also
plausible that P2, P3, and P4 Ss perceived the deletion
of reward during test as a change back to conditions which
held during pairing; for P2 and P3 this would presumably
be less aversive since these Ss had been partially rewarded
during the earlier phase. Thus, if either or both factors
were operative, the FE for P4 (and possibly Pi) should be
relatively greater than originally predicted from the
standpoint of reward expectancies.
The P4 data indicated that four of the five Ss in
that portion of the subgroup which received four nonre
warded B test trials, showed particularly large changes
from training to test. The mean amplitudes for nonrewarded
T4 and T2 children in P4 were quite divergent on the
second B in test; that is, a definitely less marked eleva
tion was found for T2 than for T4. This result is
peculiar, since both T4 and T2 had been treated in exactly
the same manner prior to the third test trial. With the
small number of children involved it may be that a dis
proportionate number of "sensitives" were represented in
T4; that is, Ss who yielded greater FEs on the basis of
individual differences. Of course, the random assignment
107
procedures used should have minimized such an eventuality;
nevertheless, the possibility remains. It is unfortunate
in a way that the design was such that T2 and T4 were
exposed to different stimulus events on the third and
fourth test trials. If identical treatments had been
delivered, and the differences already cited for the
second test trial continued or increased, then a stronger
case for the "individual difference" argument would be
present.
The speed data gathered during test and extinction
revealed increase in mean speed as a result of frustra
tion, the elevations persisting throughout the extinction
series. The rise in mean speed on early nonrewarded
trials was probably attributable to a perseverative drive
enhancement (due to frustration) from one trial to the
next. The brief intertrial interval (two seconds) would
certainly increase the probability of such perseverative
effects. Amsel's theory also predicts that frustration
should produce an approach-avoidance type of conflict
after a number of nonrewards: re_s should elicit
o o
approach (R), while at the same moment r^-Sf, being
aversive, should lead to avoidance of the response leading
to frustrative nonreward. The avoidance responses should
become more probable with successive nonrewards until
finally approach responses essentially cease altogether
(experimental extinction). No such decline in speed was
108
noted, even for Ss who met the established extinction
criterion. It is understandable for the expected reduc
tion not to have been observed for children who did not
extinguish, but it is difficult to account for the lack
of same for Ss who did stop responding.
Of course, the amplitude data generally supported
Amsel's position regarding intensification following
frustrative nonreward in that such an effect was mani
fested consistently on the second, third, and fourth B
test trials. A similar result was found on the first and
second D events in test, even though D had never been
associated with reward. In the past (for example, Brown,
1961) a good deal of attention has been devoted to an
alternate explanation of the response increment following
nonreward in terms of associative factors rather than
appealing to a motivational (frustration) explanation.
With reference to the specific instrumental response, R,
of the present experiment, if children had been rewarded
in the past for turning handles (or similar motor
responses) more vigorously, then the enhancement noted
may well be due to learning alone. However, considering
that R intensification was exhibited on D test occasions,
the associative account is rendered less likely than for
the case of B test trials. Prior to test, D had never
been presented in conjunction with reward and continued
to be associated with nonreward during test. With respect
109
to B, reward had been delivered on at least several trials
during pairing and instrumental training (except for P5)
but was suddenly omitted on test trials. It is conceiv
able that the latter reward alteration might have activated
a different habit, for example, one which involved per
formance of a more intense R, but it is difficult to enter
tain an associative explanation when the stimulus-reward
relationship was unchanged, as was true for D test
instances.
CHAPTER VII
SUMMARY
Amsel's extension of Hull-Spence behavior theory
proposes that frustration be defined as an implicit reac
tion determined by nonreward after several previous
rewards have been delivered. Since frustration is
aversive, and presumably contributes an increment to
generalized drive strength, enhanced post-frustration
instrumental responding is predicted. The expected
intensification has been noted in numerous animal experi
ments and is termed the frustration effect (FE). Amsel
has stated additional assumptions regarding the operation
of frustration; the following selected implications were
experimentally tested with children: (1) if frustrative
nonreward is introduced into an instrumental reward situ
ation, then the stronger the expectative reward responses
(rg's), the larger is the FE; (2) if Sl-R, a frustrated
stimulus-responsc event, is followed in close temporal
succession by S2-R, a second stimulus paired with the
same instrumental response but never rewarded in the past,
then the magnitude of the facilitative effect attributable
110
Ill
to perseverative frustration on S2-R is a positive func
tion of reward expectancy strength associated with Sl-R;
and (3) the strength of reward expectancy is inversely
related to resistance to extinction in the instrumental
reward setting.
Initially, children were instructed that in order
to win a prize a certain number of marbles had to be
accumulated in a transparent tube, which was mounted on
the front of a stimulus-response unit. At the same time,
the subjects were required to name the intensity of a
light (either bright, B, or dim, D), which appeared inter
mittently in a stimulus window. After determining that a
child could verbally distinguish between the two light
intensities, different numbers of B-marble pairings were
introduced. A representative pairing consisted of the
onset of B, followed by marble ejection into the tube
about one second later, at which time B was terminated.
The children in subgroup Pi received 60 pairings, while
members of P2, P3, P4, and P5 had 20, 2, 0, and 0 pairings,
respectively. All pairing subgroups had the same random
sequence of 60 B and 30 D presentations, and D events were
typically never associated with marble delivery. In the
next phase, children turned a response handle in the
presence of B or D; a series of 55 D and 5 B events then
transpired. Only handle-turning responses (Rs) to B pro
duced marble ejection for PI through P4, but no marbles
112
were forthcoming for P5. By means of the indicated
manipulations it was assumed that reward expectancy
strengths (elicited by B) were greatest for PI, less for
P2, and so on, being least for P5.
The test phase began unannounced. One quarter of
the children within each pairing subgroup (Pi through P4)
received the sequence BBBBDD, marbles being delivered as
a result of R-performance to each B, while a second
quarter was administered the same series but R never led
to marble ejection. The third and fourth quarters were
exposed to BBDD, with analogous reward conditions to
those which held for the first and second quarters.
Children in P5 were divided into two equal sections:
one half of the subjects received BBBBDD, and the other
half BBDD. No marbles were delivered for P5 members.
Finally, repeated nonrewarded B events were presented for
about 185 trials, or until the subject voluntarily ceased
responding.
Analysis of the test trials indicated a reliable
enhancement in mean R amplitude after the first B occur
rence for subjects in the nonreward condition. However,
the noted intensification was apparently independent of
the previous number of pairings except on the second B
test trial. The latter finding, which was consistent with
the first hypothesis, indicated that mean R facilitation
113
for Pi was significantly greater than P5 change, while
contrasting each of P2, P3, and P4 with P5, revealed no
significant differences. The pattern of results on sub
sequent test trials was altered in that P4 enhancement
in R was even larger than that observed for Pi. The data
for D test trials were indicative of an over-all persever
ative FE, since mean R amplitude was reliably greater for
children who had previous nonrewarded B test trials than
for children experiencing continued marble presentations
on B occasions in test. No differential response charac
teristics were found during extinction as a function of
pairing condition. But, a sufficient number of extinction
trials was not administered for the majority of subjects
to cease responding voluntarily. Supplementary analyses
of R speed showed that the FE (as defined by amplitude
change) was sometimes accompanied by elevated mean speed
for frustrated children.
It was concluded that the first hypothesis received
only marginal support on the basis of the outcome of the
second test trial analysis, but, since further data were
essentially negative, the status of the reward expectancy-
frustration relationship remains unclear. A consideration
of pretest conditions would indicate that the unexpected
R alteration for P4 was possibly not due to pairing
manipulations alone; several alternate explanations were
114
suggested to account for the substantial P4 increment.
The second hypothesis was not confirmed by the analyses
of D trials, and the extinction data provided no empirical
evidence supportive of the third hypothesis.
r e fe ren ces
REFERENCES
Amsel, A. The role of frustrative nonreward in non-
continuous reward situations. Psychological Bulletin,
1958, 55, 102-119.
_______. Frustrative nonreward in partial reinforcement
and discrimination learning: some recent history
and a theoretical extension. Psychological Review,
1962, 69, 306-328.
Amsel, A., Emhart, C.B., and Galbrecht, C.R. Magnitude
of frustration effect and strength of antedating goal
factors. Psychological Reports. 1961, 8, 183-186.
Amsel, A., and Hancock, W. Motivational properties of
frustration: III. Relation of frustration effect
to antedating goal factors. Journal of Experimental
Psychology, 1957, 53, 126-131.
Amsel, A., and Roussel, J. Motivational properties of
frustration: I. Effect on a running response of
the addition of frustration to the motivational
complex. Journal of Experimental Psychology, 1952,
43, 363-368.
Amsel, A., and Ward, J.S. Frustration and persistence:
resistance to discrimination following prior
experience with the discriminanda. Psychological
Monographs. 1965, 79 (4, Whole No. 597;.
Barrett, R.J., Peyser, C.S., and McHose, J.H. After
effects of complete and incomplete reduction in
reward magnitude. Psychonomic Science, 1965, 3,
277-278. -------------------
Brown, J.S. The motivation of behavior. New York:
McGraw-Hill Book Co., 1961.
Brown, J.S., and Farber, I.E. Emotions conceptualized as
intervening variables— with suggestions toward a
theory of frustration. Psychological Bulletin, 1951.
48, 465-495.
116
117
Clifford, T., and Schindelheim, R.H. The frustration
effect as a function of runway length. Psychonomic
Science, 1968, 10, 109-110.
Davenport, J.W., Flaherty, C.F., and Dyrud, J.P, Temporal
persistence of frustration effects in monkeys and
rats. Psychonomic Science, 1966, 6, 411-412.
Deaux, E.B., and Patten, R.L. Measurement of the
anticipatory goal response in instrumental runway
conditioning. Psychonomic Science, 1964, 1, 357-358.
Deese, J. The extinction of a discrimination without
performance of a choice response. Journa1 of Com-
parative and Physiological Psychology, 1951, 44,
362-366.
Elmes, D.G. The role of frustration in the extinction
of a running response. Psychonomic Science, 1964,
1, 345-346.
Endsley, R.C. Effortfulness and blocking at different
distances from the goal as determinants of response
speed and amplitude. Journal of Experimental Child
Psychology, 1966, 3, 18-30.
Ford, L.H., Jr. Reaction to failure as a function of
expectancy for success. Journal of Abnormal and
Social Psychology, 1963, 67, 340-348.
Graham, T.M. The role of frustrative nonreward in
instrumental escape conditioning. Unpublished
doctoral dissertation, University of Southern
California, 1963.
Grusec, T., and Bower, G. Response effort and the
frustration hypothesis. Journal of Comparative and
Physiological Psychology, 1965, 60, l28-l3o.
Haner, C.F., and Brown, P.A. Clarification of the
instigation to action concept in the frustration-
aggression hypothesis. Journal of Abnormal and
Social Psychology. 1955, 51, 204-206.
Holton, R.B. Amplitude of an instrumental response
following cessation of reward. Child Development,
1961, 32, 107-116.
118
Hull, C.L. Principles of behavior. New York: Appleton-
Century-Crofts, Inc., 1943.
_______. A behavior system. New York: John Wiley and
Sons, 1952.
Robasigawa, A. Observation of failure in another person
as a determinant of amplitude and speed of a simple
motor response. Journal of Personality and Social
Psychology. 1965, 1, 626-630.
Lawson, R. Frustration: the development of a scientific
concept. New York: Macmillan Co., 1965.
Lawson, R. , and Marx, M.H. Frustration: theory and
experiment. Genetic Psychology Monographs. 1958,
57, 393-464.
Lindquist, E.F. Design and analysis of experiments in
psychology and education. Boston: Houghton Mifflin
Co., 1953.
Longstreth, L.E. The relationship between expectations
and frustration in children. Child Development.
1960, 31, 667-671.
_______. Incentive stimuli as determinants of instru
mental response strength in children. Journal of
Comparative and Physiological Psychology. 1962,
55, 398-401.
_______. Unconditioned and conditioned frustration in
retardates . American Psychological Association
Proceedings. 1965, 1-2.
_______. Frustration and secondary reinforcement concepts
as applied to human conditioning and extinction.
Psychological Monographs. 1966, 80 (11, Whole No.
619).
Marx, M.H., and Murphy, W.W. Resistance to extinction
as a function of the presentation of a motivating
cue in the startbox. Journal of Comparative and
Physiological Psychology, 1961, 54, 207-210.
119
Marzocco, F.N. Frustration effect as a function of
drive-level, habit strength and distribution of
trials during extinction. Unpublished doctoral
dissertation, State University of Iowa, 1950.
McHose, J.H., and Ludvigson, H.W. Frustration effect
as a function of drive. Psychological Reports,
1964, 14, 371-374.
Moltz, H. Latent extinction and the fractional antici
patory response mechanism. Psychological Review,
1957, 64, 229-241.
Peckham, R.H., Jr., and Amsel, A. Within-subject
demonstration of a relationship between frustration
and magnitude of reward in a differential magnitude
of reward discrimination. Journal of Experimental
Psychology. 1967, 73, 187-195.
_____. Magnitude of reward and the frustration effect
In a within-subiects design. Psychonomic Science,
1964, 1, 285-286.
Roussel, J. Frustration effect as a function of repeated
non-reinforcements and as a function of the con
sistency of reinforcement prior to introduction of
ron-reinforcement. Unpublished master's thesis,
Tulane University, 1952.
Ryan, T.J. The effects of nonreinforcement and
incentive value on response speed. Child Develop
ment, 1965, 36, 1067-1081.
Ryan, T.J., and Watson, P. Frustrative nonreward theory
applied to children's behavior. Psychological
Bulletin. 1968, 69, 111-125.
Screven, C.G. The effects of frustration on response
strength. Journal of Comparative and Physiological
Psycho logy ,""1954. 47. TO-15?:------------- -----
Screven, C.G., and Cummings, L. The effect of nonreward
and interference on variation in amplitude of an
instrumental response. Journal of Comparative
and Physiological Psychology. 1955. 48. 299-304.
120
Seward, J.P., Pereboom, A.C., Butler, B., and Jones, R.B.
The role of prefeeding in an apparent frustration
effect. Journal of Experimental Psychology, 1957,
54, 445-450.
Shapiro, M.M. Salivary conditioning in dogs during
fixed— interval reinforcement contingent upon lever
pressing. Journal of the Experimental Analysis of
Behavior. 1961, 4, 361-364.
_______. Salivation and lever pressing relationships.
Journal of Comparative and Physiological Psychology,
1962, 55, 567-571.
Skinner, B.F. Are theories of learning necessary?
Psychological Review, 1950, 57, 193-216.
Spence, K.W. Behavior theory and conditioning. New
Haven: Yale University Press, 1956.
Stein, L. Classical conditioning of the consummatory
response as a determinant of instrumental perform
ance . Journal of Comparative and Physiological
Psychology, 1957, 50, 269-278.
Tacker, R.S., and Way, J. Motivational properties of
nonreward. Psychonomic Science, 1968, 10, 103-104.
Wagner, A.R. The role of reinforcement and nonreinforce
ment in an apparent frustration effect. Journal of
Experimental Psychology. 1959, 57, 130-136.
Whiteley, J.H., and Ryan, T.J. The effects of direct and
vicarious nonreward upon instrumental performance.
Psychonomic Science, 1967, 7, 351-352.
Winer, B.J, Statistical principles in experimental
design. New York: McGraw-Hill Book Co., 1962.
APPENDIX
122
TABLE A
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS DURING
INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ects 31
Pairing (B) 3 14.362 0.32
Test Reward
Condition (C) 1 7.720 0.17
B x C 3 44.164 0.98
Error(b)
24 45.252
Within-Subjects 160
Training Trial
Block (A) 5 8.870 1.59
A x B
15 5.373 0.96
A x C
5 2.939 0.53
A x B x C
15 5.886 1.06
Error(w)
120 5.575
............... ■ ■ V- ■ • *
123
TABLE B
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T2 SUBJECTS DURING
INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ects 31
Pairing (B)
Test Reward
3 7.120 0.13
Condition (C) 1 97.755 1.80
B x C
3 29.259 0.54
Error(b)
Within-Subjects
Training Trial
24
160
54.161
3.26a Block (A) 5 12.980
A x B
15 1.853 0.47
A x C
5 2.612 0.66
A x B x C
15 5.164 1.30
Error(w)
120 3.977
i.rrr ------1 ------ 1
a(0.01 > P > 0.005)
124
TABLE C
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS ON TWO B-TRIAL
BLOCKS DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subjects 31
Pairing (B) 3 14.983 0.77
Test Reward
Condition (C) 1 18.598 0.96
B x C 3 11.348 0.59
Error(b) 24 19.353
Within-Subjects
32
B-trial
Block (A) 1 11.816 2.76
A x B
3 0.796 0.19
A x C
1 2.848 0.67
A x B x C
3 1.076 0.25
Error(w)
24 4.275
----- 4
125
TABLE D
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T2 SUBJECTS ON TWO B-TRIAL
BLOCKS DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ects 31
Pairing (B) 3 7.594 0.38
Test Reward
Condition (C) 1 39.062 1.95
B x C 3 12.010 0.60
Error(b) 24 20.036
Within-Subj ects 32
B-trial
Block (A) 1 11.391 2.53
A x B
3 6.380 1.42
A x C 1 4.516 1.00
A x B x C
3 2.609 0.58
Error(w)
24 4.505
—..
126
TABLE E
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS ON TWO D-TRIAL
BLOCKS DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subjects 31
Pairing (B) 3 15.807 1.07
Test Reward
Condition (C) - 1 0.766 0.05
B x C 3 7.432 0.50
Error(b) 24 14.807
Within-Subj ects 32
D-trial
Block (A) 1 33.062 12.40*
A x B
3 0.792 0.30
A x C 1 1.000 0.37
A x B x C
3 5.521 2.07
Error(w) 24 2.667
--sat--1 ---1
a(0.005 > P > 0.001)
127
TABLE F
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T2 SUBJECTS ON TWO D-TRIAL
BLOCKS DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ect s 31
Pairing (B) 3 1.764 0.07
Test Reward
Condition (C) 1 26.910 1.11
B x C 3 6.535 0.27
Error(b) 24 24.259
Within-Subj ects 32
D-trial
Block (A) 1 2.441 0.75
A x B
3 1.212 0.37
A x C
1 2.066 0.63
A x B x C 3 3.670 1.12
Error(w) 24 3.270
--------1
128
TABLE G
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 SUBJECTS DURING
INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ect s 31
Pairing (B)
3 0.3307 0.70
Test Reward
Condition (C) 1 0.6440 1.37
B x C
3 0.3827 0.81
Error(b)
24 0.4711
Within-Subjects
160
Training Trial
a
Block (A) 5 0.2132 5.54
A x B
15 0.0347 0.90
A x C
5 0.0333 0.86
A x B x C
15 0.0360 0.94
Error(w)
120 0.0385
J
a(p < 0.001)
129
TABLE H
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T2 SUBJECTS DURING
INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ects 31
Pairing (B) 3 0.0899 0.26
Test Reward
Condition (C) 1 0.1190 0.34
B x C 3 0.0812 0.23
Error(b)
24 0.3488
Within-Subjects 160
Training Trial
Block (A) 5 0.2799 4.67a
A x B 15 0.0534 0.89
A x C
5 0.0718 1.20
A x B x C 15 0.0391 0.65
Error(w)
120 0.0599
m --------1
3(p < 0.001)
130
TABLE I
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 SUBJECTS ON TWO B-TRIAL BLOCKS
DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ects 31
Pairing (B) 3 0.5925 3.65a
Test Reward
Condition (C) 1 0.0390 0.24
B x C 3 0.0507 0.31
Error(b) 24 0.1623
Within-Subjects 32
B-trial
Block (A) 1 0.0169 0.44
A x B
3 0.0100 0.26
A x C
1 0.0001 -
A x B x C
3 0.0474 1.22
Error(w)
24 0.0388
......... .
a(0.05 > P > 0.025)
131
TABLE J
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T2 SUBJECTS ON TWO B-TRIAL BLOCKS
DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subjects 31
Pairing (B)
3 0.0722 0.41
Test Reward
Condition (C)
1 0.0013 0.01
B x C
3 0.1443 0.82
Error(b) 24 0.1752
Within-Subjects 32
B-trial
Block (A) 1 0.2364 7.48a
A x B
3 0.0751 2.38
A x C
1 0.1114 3.53b
A x B x C
3 0.0182 0.58
Error(w) 24 0.0316
s- - - - - - - - r r m - - - - - - 1
a(0.01 > p > 0.005)
b
(0.10 > p > 0.05)
132
TABLE K
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 SUBJECTS ON TWO D-TRIAL BLOCKS
DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subj ects 31
Pairing (B) 3 0.1271 0.68
Test Reward
Condition (C) 1 0.3922 2.09
B x C 3 0.1011 0.54
Error(b) 24 0.1880
Within-Subjects 32
D-trial
Block (A) 1 0.2488 6.14a
A x B 3 0.0281 0.69
A x C 1 0.0863 2.13
A x B x C 3 0.0214 0.53
Error(w) 24 0.0405
■ " .i
a(0.025 > p > 0.01)
133
TABLE L
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T2 SUBJECTS ON TWO D-TRIAL BLOCKS
DURING INSTRUMENTAL TRAINING
Source of Variance df Mean Square F Ratio
Between-Subjects 31
Pairing (B) 3 0.1462 0.70
Test Reward
Condition (C) 1 0.1425 0.68
B x C 3 0.0202 0.10
Error(b) 24 0.2090
Within-Subjects 32
D-trial
Block (A) 1 0.3364 2.56
A x B 3 0.1348 1.03
A x C 1 0.1764 1.34
A x B x C 3 0.0993 0.76
Error (w)
24 0.1312
134
TABLE M
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 AND T2 SUBJECTS COMBINED
DURING LATE TRAINING AND ON THE
SECOND TEST TRIAL
Source of Variance df Mean Square F Ratio
Between-Subjects 79
Pairing (B) 3 0.422 0.02
Test Reward
Condition (C) 1 746.062 30.573
B x C 3 4.514 0.18
Error(b) 72 24.409
Within-Subj ects 80
Training vs.
Test (A) 1 145.351 22.63a
A x B 3 2.852 0.44
A x C 1 161.002 25.083
A x B tf-C 3 8.257 1.29
Error(w) 72 6.423
--=r:---: --1
a(p < 0.001)
135
TABLE N
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE THIRD B IN TEST
Source of Variance df Mean Square F Ratio
Between-Subjects 39
Pairing (B) 3 43.758 1.95
Test Reward
Condition (C) 1 382.811 17.063
B x C 3 6.904 0.31
Error(b) 32 22.436
Within-Subj ects 40
Training vs.
Test (A) 1 68.450 15.93a
A x B 3 7.892 1.84
>
X
o
1 99.012 23.04a
A x B x C 3 10.440 2.43b
Error(w) 32 4.298
a(p < 0.001)
b(0.10 > p > 0.05)
136
TABLE 0
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE FOURTH B IN TEST
Source of Variance df Mean Square F Ratio
Between-Subjects 39
Pairing (B) 3 51.254 3. II3
Test Reward
Condition (C) 1 288.800
b
17.53
B x C 3 8.925 0.54
Error(b) 32 16.473
Within-Subjects 40
Training vs.
Test (A) 1 35.112 6.96C
A x B 3 5.688 1.13
A x C 1 54.450 10,79d
A x B x C 3 3.440 0.68
Error(w) 32 5.048
a(0.05 > p > 0.025) C(0.025 > P > 0.01)
b(p < 0.001) d(0.005 > P > 0.001)
137
TABLE P
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE FIRST D IN TEST
Source of Variance df Mean Square F Ratio
Between-Subj ects 39
Pairing (B)
3 22.379 0.83
Test Reward
Condition (C)
1 54.450 2.02
B x C
3 9.950 0.37
Error(b) 32 26.975
Within-Subj ects 40
Training vs.
Test (A) 1 74.112 27,45a
A x B 3 3.146 1.17
A x C 1 39.200 14.52a
A x B x C 3 3.365 1.25
Error(w) 32 2.700
=aai .1
a ( p < 0 . 0 0 1 )
138
TABLE Q
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE SECOND D IN TEST
Source of Variance df Mean Square F Ratio
Between-Subjects
39
• s .
Pairing (B)
3 24.879 1.03
Test Reward
Condition (C) 1 26.450 1.10
B x C
3 16.083 0.67
Error(b) 32 24.041
Within-Subjects 40
Training vs.
Test (A) 1 2.112 0.59
A x B 3 14.846 4. lla
A x C 1 16.200 4.49
A x B x C 3 1.632 0.45
Error(w) 32 3.609
—... -- i
a(0.025 > p > 0.01)
139
TABLE R
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T2 SUBJECTS DURING LATE TRAINING
AND ON THE FIRST D IN TEST
_ -------- L ----------
Source of Variance df Mean Square F Ratio
Between-Subj ects 39
Pairing (B) 3 13.111 0.58
Test Reward
Condition (C) 1 106.953 4.76a
B x C 3 6.553 0.29
Error(b) 32 22.489
Within-Subjects 40
Training vs.
Test (A) 1 52.003 9.21b
A x B 3 0.836 0.15
A x C 1 25.878 4.58a
A x B x C 3 0.445 0.08
Error(w) 32 5.645
S(0.05 > P > 0.025)
b(0.01 > P > 0.005)
140
TABLE S
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE FOR T2 SUBJECTS DURING LATE TRAINING
AND ON THE SECOND D IN TEST
Source of Variance df Mean Square F Ratio
Between-Subj ects 39
Pairing (B) 3 13.011 0.49
Test Reward
Condition (C) 1 48.828 1.83
B x C 3 2.845 0.11
Error(b) 32 26.667
Within-Subjects 40
Training vs.
Test (A) 1 10.153 2.44
A x B 3 0.603 0.15
A x C 1 3.003 0.72
A x B x C
3 5.819 1.40
Error(w)
32 4.154
141
TABLE T
TYPE I ANALYSIS OF VARIANCE OF MEAN RESPONSE
AMPLITUDE DURING LATE TRAINING AND ON FIFTHS
OF EXTINCTION
Source of Variance df Mean Square F Ratio
Between-Subjects 89
Pairing (B) 4 54.724 0.80
Error(b) 85 68.324
Within-Subjects 450
Trial Block (A) 5 5.199 0.61
A x B 20 4.392 0.52
Error(w) 425 8.492
142
TABLE U
TYPE I ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED DURING LATE TRAINING AND ON
FIFTHS OF EXTINCTION
Source of Variance df Mean Square F Ratio
Between-Subjects 89
Pairing (B) 4 2.0741 3.18a
Error(b) 85 0.6528
Within-Subjects 450
Trial Block (A) 5 0.8769 9.68b
A x B 20 0.1345 1.48
Error(w) 425 0.0906
3(0.025 > p > 0.01)
b (P < 0.001)
143
TABLE V
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 AND T2 SUBJECTS COMBINED
DURING LATE TRAINING AND ON THE
SECOND TEST TRIAL
Source of Variance df Mean Square F Ratio
Between-Subjects
79
Pairing (B) 3 0.6143 3.28a
Test Reward
Condition (C) 1 0.0837 0.45
B x C 3 0.1306 0.70
Error(b) 72 0.1873
Within-Subj ects 80
Training vs.
Test (A) 1 0.0360 0.52
A x B 3 0.0256 0.37
A x C 1 0.1850 2.67
A x B x C
3 0.0271 0.39
Error(w) 72 0.0694
-------- r = r r = ------=rr— 1
a(0.05 > p > 0.01)
144
TABLE W
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE THIRD B IN TEST
Source of Variance df Mean Square F Ratio
Between-Subjects 39
Pairing (B) 3 0.6390 4.04a
Test Reward
Condition (C) 1 0.1288 0.82
B x C 3 0.0333 0.21
Error(b)
32 0.1580
Within-Subjects 40
Training vs.
b
Test (A) 1 0.7664 13.03
A x B 3 0.0008 -
A x C 1 0.1337 2.27
A x B x C 3 0.1308 2.22
Error(w)
32 0.0588
rj.-vi , ,l i
3(0.025 > p > 0.01)
b
(0.005 > p > 0.001)
145
TABLE X
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE FOURTH B IN TEST
Source of Variance df Mean Square F Ratio
Between-Subjects
39
Pairing (B)
3 0.7549
_ _, a
5.01
Test Reward
Condition (C)
1 0.0904 0.60
B x C
3 0.0397 0.26
Error(b)
32 0.1506
Within-Subjects 40
Training vs.
Test (A) 1 1.4499 20.77b
A x B 3 0.0395 0.57
A x C 1 0.0945 1.35
A x B x C
3 0.1646 2.36c
Error(w)
32 0.0698
J - l . - l . . ! .'.'I'. 1' J - . - 4
a(0.05 > P > 0.001)
b(p < 0.001)
C(0.10 > p > 0.05)
146
TABLE Y
TYPE III ANALYSIS OF VARIANCE OF MEAN RESPONSE
SPEED FOR T4 SUBJECTS DURING LATE TRAINING
AND ON THE FIRST D IN TEST
Source of Variance df Mean Square F Ratio
Between-Subjects
39
Pairing (B)
3 0.4394 2.26a
Test Reward
Condition (C)
1 0.0594 0.31
B x C 3 0.0214 0.11
Error(b) 32 0.1946
Within-Subj ects 40
Training vs.
Test (A) 1 0.0174 0.22
A x B 3 0.2551 3.23b
A x C 1 0.3781 4.78b
A x B x C 3 0.0259 0.33
Error(w) 32 0.0791
... . _ „ | 4
3(0.10 > P > 0.05)
b(0.05 > P > 0.025)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Sex-Role Preferences Of Early Adolescents In Relation To Adjustment
PDF
Human Performance As A Function Of The Joint Effects Of Drive And Incentive Motivation
PDF
Dominance And Contiguity As Interactive Determinants Of Autonomic Conditioning
PDF
Human Gsr Classical Conditioning And Awareness Of The Cs-Ucs Relation
PDF
Instrumental Escape Behavior Applied To Remedial Reading
PDF
Creativity In Children: A Study Of The Relationship Between Temperament Factors And Aptitude Factors Involved In The Creative Ability Of Seventh Grade Children With Suggestions For A Theory Of C...
PDF
Recognition Accuracy In The Method Of Single Stimuli: A Test Of An Operational Definition Of The Distinctiveness Of Stimuli
PDF
Familism, Suburbanization, And Residential Mobility In A Metropolis
PDF
Middle-Class Marital Roles - Ideal And Perceived In Relation To Adjustment In Marriage
PDF
The Relation Of Evaluative Attitudes To Traits Of Introversion And Extraversion
PDF
Classical Discrimination Conditioning As A Function Of Probability Of Reinforcement
PDF
Frustration Versus Secondary Reinforcement In Children Under Continuous And Partial Primary Reinforcement
PDF
A Critique Of The Concept Ethnocentrism As Set Forth In Selected Social Science Literature
PDF
A Four-Year Follow-Up Of Educationally Disadvantaged Preschool Children, Analyzing Home Environment Variables Facilitating Achievement
PDF
Social Factors Related To Dentistry As A Career
PDF
The Visual Fusion Threshold (Vft) Test As A Measure Of Perceptual Efficiency In Kindergarten And First Grade, And As A Possible Predictor Of Later Reading Retardation
PDF
Autonomic and cognitive indices of semantic conditioning and generalization
PDF
Inhibitory-excitatory interaction of conditioned autonomic responses through interoceptive and exteroceptive conditioning
PDF
Habituation Of The Multiple Unit Discharge Response To White Noise Stimulation In The Unanesthetized Rabbit
PDF
Verbal And Motor Mediation In Normal Children Of Two Ages And Retarded Individuals
Asset Metadata
Creator
Gilbert, Dunham Harding
(author)
Core Title
Reward Expectancy Strength As Related To The Magnitude Of Frustration In Children
Degree
Doctor of Philosophy
Degree Program
Psychology
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
OAI-PMH Harvest,psychology, experimental
Language
English
Contributor
Digitized by ProQuest
(provenance)
Advisor
Longstreth, Langdon E. (
committee chair
), Grings, William W. (
committee member
), McDonagh, Edward C. (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c18-656962
Unique identifier
UC11362330
Identifier
6906493.pdf (filename),usctheses-c18-656962 (legacy record id)
Legacy Identifier
6906493
Dmrecord
656962
Document Type
Dissertation
Rights
Gilbert, Dunham Harding
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA
Tags
psychology, experimental