vfi—w—v—v ____._—‘—

 

 

.. . .
.01 l ‘ t
.. . ‘ .

It“'t0\

,\.
.cuc

'0§I

‘Il.".
. .

|

.99.

l'0-"|A

4|40ton'

'Q..
. .

. 0
'0'0\'. .“

'..\c

.

.

...

a

.4“.

i

."" I ‘1

\Q'

.mm
.‘ .
‘5—

“J

Q.-I

:

.'! "‘ 3 .‘l.....b.’«f"‘f9'\n n :1...O‘O.u_ Q *NCN 0.x-

‘§:‘.' "’-0‘°". ‘.’ ~o-o

 

..--.—-:4 v -.a- ._ ‘fi‘<. «aqu‘g.-.._.,,,‘ . 9.‘ 3 ,......,_,

usmrsow EFFECT as A macaw .

F REWARD hummus.

 

- the vegan am

 

mam STATE [mm ‘ ’
mum Gum! K01:
1921 * .

‘ ““-I.ro-o

‘._.." _-.. .. -.— con...- .a ‘.~~.‘..~~”n‘

.‘R

'I
. o‘.
o 0 I '
-
. .
‘9
- . -- -
O ‘-
' -'- .‘
. . . .
"C C
.
. . ..-
-o
-o
‘
.
Q . j
- . o
n - .
. . .
I o
. ‘o
- - . ..
_ -
. . . . .
a q . -
- - . ..
I . O - .
.. . u
. . ,
- . -.
- ‘
.
', . .
0 . .
a V
' C
. . - '
. - .
.. O" -
. - Q
. . -
. . _
. C .
. - a .
. . '
.. ’
- a
. .
. .
-- s - . o
_ - . V
. . -
'. ‘ . . _
u ' ‘ .
. - 0 ' ‘ I
_ .
. ...
.
. .
- - .
- . . .
‘
‘ - - ,
.. ‘.
- . .
.0 .
_. .' . .
. -
. . - .
~- 0 .
o .. - o . ..
NO -‘ d I
‘ - ‘ . . .
. ‘ - - -
. . . .
.-, . .
. . .
.~-
. x , ,
- e .
- r O
-. .
, - o.
_ \ .
o — ' Iv
- .
- ---’¢-. cu
- -.n .I.
.
... ...
‘-‘ o o
- x? --
. . . .
- . . a ..
- g - s
- .\-¢
- - . .. .
u -‘ J- . l‘
_ .-'-__
. . . _ .
. I O I I 0.-
. - - -
- . a . I.
. . .
. . . . ‘
- 4 a u a
. .
'o - V O
o .
-
. -
. -
. . .
I ‘7 ‘
. .
v 0 .
‘ .
- - . . ‘
. . . .
. , . .
C
.
- -.
. . .3
o
. .
, ,
a
.
I "
. .
.
o
.
~
.
. . ..
. .
I, I
.
- .
, ’
A .‘C 4'
I
. .
. .
..
.
. .
u
.
.—
o o‘ a .
‘ ~ ;
.
_ 1‘ o. ‘
.I . .
. . _ I
. 4' - 0' '
’4 -.
/ .
1
'.l
0‘- l

 

 

 

 

ABSTRACT

FRUSTRATION EFFECT AS A FUNCTION
OF REWARD MAGNITUDE

BY
Linda Gilbert Kok

The frustration effect is defined as an increase in
response vigor immediately following non-reward after an
expectancy of reward has been develOped. Response vigor,
usually running speed, is measured following trials in which
reward has been eliminated and compared to running speed
following reward. Frustration effect, according to one
model, Amsel's theory of frustrative non-reward, has been
studied using a double alley apparatus which has met with
considerable criticism. The first criticism suggests that
demotivational factors due to the intake of food on reward
trials may be, in part, responsible for the superior running
speeds following non-reward. The second criticism stems
from possible inadequacies of tests of the effects of magni-
tude of reward and the effects of the degree of incomplete
reduction of reward (to some non-zero level), as these tests
are conducted in the double runway. An apparatus (panel-
push) was used to eliminate some of the problems associated

with the double runway. It was postulated that an increase

Linda Gilbert Kok

in response strength would occur immediately following the
withdrawal of reinforcement and, further, that the size of
this effect would be directly related to the magnitude of
reward used in training.

Five groups of rats were trained with different
amounts of reinforcement, either 20 mg, 45 mg, 97 mg, 190 mg,
or 500 mg food pellets. Following 45 consistently rewarded
trials the second stage of the experiment began. In this
stage, which lasted for four days, the first three trials
of each day were rewarded and the last five were non-rewarded.
This design yielded four trials following reward and four
trials following non-reward for these four days. The third
’stage of the experiment was regular extinction with a total
of 25 non-rewarded trials.

The results showed a significant decline in response
force within the first four trials following the initial
withdrawal of reward. This effect was found to be related
to magnitude of reward only in a limited (non-significant)
fashion, and then only in terms of the drop from the re-
sponse force level at the end of acquisition to its level
at the end of extinction.

These results were interpreted as not supporting
Amsel's theory of frustrative non-reward and were discussed
from the viewpoint of elicitation theory. This discussion
included a comparison of the panel-push apparatus with other
types of apparatus previously used to test frustration

theory.

FRUSTRATION EFFECT AS A FUNCTION

OF REWARD MAGNITUDE

BY

Linda Gilbert Kok

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1971

ACKNOWLEDGMENT

I sincerely wish to thank the members of my
committee: Dr. M. Ray Denny, Dr. S. Ratner, Dr. Gordon
WOod, and Dr. Robert Raisler for their guidance in the
preparation of this thesis with special appreciation to
Dr. Denny, my committee chairman, for his help at every

stage of the investigation.

ii

TABLE OF CONTENTS

Chapter Page

INTRODUCT ION O O O O O O O O O O O O O I l
Amsel's Theory of Frustrative Non-reward. . . . 1
Criticism of Amsel's Experimental Procedure. . . 3
Double Runway Apparatus . . . . . . . . . 5

Demotivational Criticism . . . . . . . . 5
Criticism Related to Studies of Effect of
Reward Magnitude . . . . . . . . . . 8
Operant Response. . . . . . . . . . . . 13
Single Runway. . . . . . . . . . . . . 15
Panel Pushing. . . . . . . . . . . . . 16
The Present EXperiment. . . . . . . . . . 17
Hypotheses. . . . . . . . . . . . . . 18

METHOD 0 O O O O O O O O O O 0 O O O O 2 1
Apparatus . . . . . . . . . . . . . . 21
Subjects . . . . . . . . . . . . . . 24
Overview of Method . . . . . . . . . . . 24
Preliminary Training . . . . . . . . . . 25

RESULTS. . . . . . . . . . . . . . . . 28

DISCUSSION. . . . . . . . . . . . . . . 49

REFERENCES. . . . . . . . . . . . . . . 55

APPENDIX . . . . . . . . . . . . . . . 59

iii

Table

1.

LIST OF TABLES

Page

Analysis of Variance on Effects of Reward
Magnitude, Days of Training, and the
Interaction of These, with Response Force
as the Measure, Stage I . . . . . . . 30

Analysis of Variance on Effects of Reward
Magnitude, Days of Training, and the
Interaction of These, with Response
Latency as the Measure, Stage I . . . . 33

Analysis of Variance for Effects of Reward
Magnitude, Blocks of Trials (Following
Reward and Following Non-reward) and the
Interaction of Reward Magnitude and
Blocks, with Response Force as the Measure,
Day 1, Test Stage . . . . . . . . . 37

Analysis of Variance for Effects of Reward
Magnitude, Blocks of Trials (Following
Reward and Following Non-reward) and the
Interaction of Reward Magnitude and
Blocks, with Response Latency as the Mea—
sure, Day 1, Test Stage . . . . . . . 42

Analysis of Variance for Effects of Reward
Magnitude, Stage of Experiment, and the
Interaction of These, with Response Force
as the Measure (End of Acquisition and End
of Extinction) . . . . . . . . . . 43

Analysis of Variance for Effect of Reward
Magnitude, Stage of Experiment
(Acquisition and Extinction), and the
Interaction of These Effects with
Response Latency as the Measure . . . . 47

Analysis of Variance on Effects of Reward
Magnitude, Blocks of Trials, and Magnitude
by Blocks Interaction with Response Force
as the Measure, Day 2, Stage II . . . . 59

iv

Table Page

B. Analysis of Variance on Effects of Reward
Magnitude, Blocks of Trials, and Magnitude
by Blocks Interaction with Response Force
as the Measure, Day 3, Stage II . . . . . 61

C. Analysis of Variance on Effects of Reward
Magnitude, Blocks of Trials and Blocks by
Magnitude Interactions with Response Force
as the Measure, Day 4, Stage II . . . . . 63

D. Analysis of Variance for Reward Magnitude
Blocks of Trials and Magnitude by Blocks
Interaction with Response Latency as the
Measure, Day 2, Stage II . . . . . . . 65

E. Analysis of Variance for Effects of Reward
Magnitude, Blocks of Trials and Magnitude
by Blocks Interaction with Response Latency
as the Measure, Day 3, Stage II . . . . . 67

F. Analysis of Variance for Effects of Reward
Magnitude, Blocks of Trials and Blocks by
Magnitude Interaction with Response Latency
as the Measure, Day 4, Stage II . . . . . 69

LIST OF FIGURES

Figure
1. Drawing of Apparatus Discussed in Text . . .

2. Mean Response Force in Training by Groups Over
Days 0 O O O O O O O O O O O O O

3. Mean Response Latency in Acquisition by Groups
over Days 0 O O O O O O O I O O O

4. Mean Response Force on First Day of Testing
by Trial 0 O O O O O O O I O O O O

5. Mean Response Force for Each Group for First
Day of Test Stage by Trial Blocks -. . . .

6. Mean Response Latency for Each Group for Each
Trial on Day 1 of Test Stage . . . . . .

7. Mean Response Latency for Each Group for
First Day of Test Stage by Trial Blocks . .

8. Mean Response Force for Each Group on Last Day
of Acquisition, First Day of Extinction
(Following Test Sessions), and Last Day of
Extinction . . . . . . . . . . . .

9. Mean Response Latency for Each Group on the
Last Day of Acquisition, First Day of
Extinction, and Last Day of Extinction. . .

A. Mean Response Force for Each Group for Second
Day of Test Stage by Trial Blocks . . . .

B. Mean Response Force for Each Group for Third
Day of Test Stage by Trial Blocks . . . .

C. Mean Response Force for Each Group for Third
Day of Test Stage by Trial Blocks . . . .

vi

Page

23

29

32

34

36

39

40

44

46

60

62

64

Figure Page

D. Mean Response Latency for Each Group for the
Second Day of Test Stage by Trial Blocks . . 66

E. Mean Response Latency for Each Group for the
Third Day of the Test Stage by Trial Blocks . 68

F. Mean Response Latency for Each group for
Fourth Day of the test stage by Trial Blocks. 70

vii

INTRODUCTION

Numerous experiments over the past two decades have
been concerned with Amsel's theory of frustrative non-reward.
Three major types of apparatus and several experimental
designs have been used to examine the effects of non-reward
after an expectancy of reward has been established. These
different methods have produced a set of results which are
not entirely consistent with each other.

The present experiment was an attempt to examine
the effect of frustrative non-reward on the force or ampli-
tude of a response as a function of the magnitude of reward
used during original training. Latency of the response was

similarly studied.

Amsel's Theory of Frustrative Non-reward
Amsel defined primary frustration as an increase in

response vigor immediately following non-reward after an
expectancy of reward (rg) has been developed by consistent
reinforcement for a number of trials.

The addition of frustration to the motivational com-

plex increases the strength of an immediately follow-

ing, hunger-motivated response, presumably through an

increase in generalized drive strength (Amsel and Ward,
1954).

In his review of the literature of frustrative non-
reward, Amsel (1958, 1962) cited earlier allusions to
"frustration effects" by Skinner, Hull, and others. Skinner
(1950) observed that "when we fail to reinforce a response
that has been reinforced . . . we set up an emotional re-
sponse--perhaps what is often meant by frustration. . . .”
Also, Hull (1952) stated that "abrupt cessation of a cus-
tomary reinforcement will, at first, result in a slight rise
in SER (excitatory potential or response tendency) due to
emotion."

Similarly, Miller and Stevenson (1936) explained
the agitated behavior of rats during extinction, which
carried over from one trial to the next, by an emotional-
motivational interpretation. This state, according to Bower
(1962) is due to the discrepancy between expected and ob-
tained rewards.

One of the basic assumptions underlying Amsel's
frustrative non-reward hypothesis is that the strength of
the anticipatory goal response (rg) will directly influence
the amount of the increase in response vigor following non-
reward (Amsel, 1962). This concept, from Hullian theory,
was viewed as related to frustration.

Not rewarding a R will elicit frustration to the degree
that earlier rewards of that R have lead to the con-
ditioning of r to the cues in an instrumental response
sequence (Hull, 1952).

Amsel dealt with another aspect of frustration

which he referred to as conditioned frustration (rf-sf).

Conditioned frustration develops more slowly. This process
may be conceptualized as resulting from the classical con-
ditioning of the effects of non-reward to the cues associ-
ated with responding. This, Amsel calls the fractional
anticipatory frustration reaction (rf) which is analogous
to fractional anticipatory goal reaction from Hull-Spence
theory (Amsel, 1962).

Presumably, this rf-sf effect has been demonstrated
most efficiently by Goodrich (1959). The initial inhibitory
effect of partial reinforcement schedules in contrast to
continuous (100%) reward schedules reversed itself after a
number of trails. What resulted was that as asymptote was
reached, the partial (50%) reward group ran faster than the
100% reward group. Amsel interpreted these results as evi-
dence that the heightened motivational level associated with
non-reward is gradually conditioned to the cues of respond-
ing. Goodrich's results will be described in detail later.

Criticism of Amsel's Experimental
Procedure

 

 

A great deal of research has been stimulated by
the theoretical assertions of the frustrative non-reward
hypothesis. Some of the reports of this research have in-
cluded varying degrees of criticism of the experimental
apparatus in which Amsel first examined the frustration
effect. These criticisms are primarily directed at two

aspects of the theory.

First, Seward, Pereboom, Butler, and Jones (1957)
suggested that demotivational factors, inherent in the
double runway apparatus and design which involved the intake
of food on rewarded trials, may be in part responsible for
the faster running speeds following non-reward.

Second, some of the studies of the effect of reward
magnitude on the amount of increase in response vigor follow-
ing non-reward produced results which may be inconsistent
with Amsel's assumption concerning the role of the antici-
patory goal response (McHose & Ludvigson, 1965; McHose &
Gavelek, 1969). Further, examination of the effects of
reduction of reward magnitude to some non-zero level pro-
duced results which indicated that different amounts of
reward in the two goal boxes of the double runway apparatus
may lead to the development of dual expectations for the
two goal boxes. Thus, utiliZation of different magnitudes
of reward in the two goal boxes produced results which may
not be entirely attributable to a frustration effect (Daly,
1968; McHose, 1970).

The first of these criticisms necessitates an
examination of the type of apparatus used in the present
research. The types of apparatus that have been used in—
clude the double runway, single runway, Operant response
apparatus, and a panel-push device in which both latency
and force of response can be measured.

The second criticism will be elaborated and dis-

cussed within the discussion of the double runway apparatus

since it is in this apparatus that most of the work re-

lated to frustration has been done.

Double Runway Apparatus

 

Demotivational Criticism

The double runway is the most frequently used
apparatus in the study of frustration since it was the one
in which Amsel's theory was first tested (Amsel & Roussel,
1952). It consists of two consecutive runways. Each is
divided into three segments: start box (SB), alley, (A),
and goal box (GB). The goal box of the first runway serves
as the start box for the second runway. Running speed in
each alley is measured by a series of photo cells.

Subjects are rewarded in both goal boxes during the
acquisition stage of the experiment. This continues for
30-130 trials in order to build an expectation of reinforce-
ment. Following these consistently rewarded trials, reward
in the first goal box (GB-l) is eliminated (Peckham &
Amsel, 1967; Krippner, Endsley, & Tacker, 1967) or in some
studies reduced (Bower, 1962; McHose & Ludvigson, 1965).
Periodically, non-rewarded trials are interspersed among
the trials. This is usually referred to as the "post-
shift" stage of the experiment.

The running speed in the second alley (A—2) follow-
ing the non-rewarded trials in GB-l is measured and compared

to A-2 speed following reward in GB-l, from both the

acquisition stage and the post-shift rewarded trials. It
has been observed that the running speed for A-2 following
non-reward is faster than A-2 running speed following reward
in GB-l. It is this increase in running speed following
non-reward that Amsel has termed the frustration effect
(FE).

The first criticism of this interpretation of the
increased running speed came from Seward 33.21' (1957).
They concluded that at least a portion of the increase was
due not to the motivational prOperties of frustration, but
to the demotivational properties associated with the intake
of food in 63-1 on rewarded trials. In their experiment
A-l speeds were calculated. On some trials subjects re—
ceived either one-half or one gram of food just prior to
entering the first start box. On other trials subjects
were placed directly into SB-l from their waiting cages
with no prefeeding. Following prefeeding the A-l running
speeds were found to be significantly slower. They did not
interpret this as necessarily resulting from decreased
hunger, but perhaps from chewing and/or swallowing during
the initial stages of the alley or, more likely, from a
reduction in "the drive to eat."

To this, Amsel (1958) replied that the large magni-
tude of reward used in the Seward gt_al. (1957) study was
responsible for the slower running speeds. He had con-

sidered this possibility in his earliest studies of FE and

had ruled against this interpretation since the magnitude
of reward used in his study was small, 100-125 milligrams
(Amsel & Roussel, 1953).

Wagner (1959) in an attempt to clarify the situ-
ation, designed an experiment utilizing a control group
which had never received reward in GB-l. Such a group
would not develOp an anticipatory goal response and would
not experience frustration following non-reward. Nor would
their A-2 running speeds be affected by demotivational
factors. Therefore, Wagner suggested, if the increase in
A-2 running speed was "entirely due to demotivational
factors" of reward intake, then following non-rewarded
trials the control group should run about equally as fast
as the experimental group on its trials following non-
reward. The results showed that the experimental group
ran faster after non-reward than the control group which
had never experienced reward in GB-l. He concluded that
FE is "independent of any demotivational effects of feed-
ing" (Wagner, 1959).

This extreme conclusion, however, does not neces-
sarily follow from his data. Although Wagner has probably
shown that demotivational factors are not the only effects
involved, he has not ruled out the possibility that demoti-
vation may be partially responsible. The criticism by
Seward et a1. did not reject such a possibility; they

merely stated that the demotivational factors could not

be eliminated as a partial explanation of the increase in
A-2 running speed following non-reward as studied in the

double runway.

Criticism Related to Studies of
Effect of Reward Magnitude

 

 

A set of studies grew from attempts to examine the
effect of magnitude of reward and the effect of incomplete
reward reduction on the amount of FE. From reports on
these studies came new criticisms of Amsel's experimental
procedure.

As was stated earlier, if the magnitude of FE is
directly related to the degree of anticipatory goal response,
as it postulated in Amsel's theory, then magnitude of FE
should also be directly related to reward magnitude used
in training (Peckham & Amsel, 1967).

Results supporting Amsel were obtained in two
studies by Peckham and Amsel (1964, 1967). These studies
showed that A-2 speed increased more following non-reward
for eight pellets in GB-l than following non-reward for
two pellets in 63-1. They concluded that this offered
support for a direct relationship between FE and magnitude
of training reward.

Krippner, Endsley, and Tacker (1967) obtained
similar results but found that large GB-l rewards depressed
A-Z speeds following rewarded trials as compared with small
GB-l rewards following rewarded trials. Thus they sug-

gested that large rewards may have a tendency to elicit

staying in GB-l rather than approaching GB-2 (as in keep—

ing with the demotivational hypothesis). They interpreted
their results as not completely supporting the FE hypothesis,
and concluded that the discrepancy may be due to their use
of a between subjects design as opposed to Amsel's within
subject design.

McHose, Meyer, and Maxwell (1969) utilized a within
subject design and obtained results "directly opposite to
the predictions derivable from the model" (Amsel's frus-
trative non—reward theory), namely, that the small reward
produced at least as large an FE as the large reward. They
suggested that a comparison between the within subject and
between subjects designs was necessary. McHose and Gavelek
(1969) conducted such a study and found that performance
following non-reward did not depend on reinforcement magni-
tude, irrespective of design. They interpreted this as
damaging to Amsel's theory since the vigor of frustration
response and the resultant increment in drive is assumed
to vary directly with reward expectancy (reward magnitude).
They concluded that the results of Peckham and Amsel's
study (1967) which are in apparent support of the frus-
tration theory, could (along with their own results) mean
that the double alley tests of hypothesis are inadequate.
The procedural inadequacies in question are described in

the following studies.

10

Another variable concerned with frustrative non-
reward theory that has been studied is the effect of reduc-
ing the reward in GB-l to some non-zero level. Bower (1962)
was one of the first to examine this variable. He hypothe-
sized that if non-reward is frustrating and if this is due
to the discrepancy between expected reward and obtained
reward, perhaps the amount of FE would be related to the
relative size of reduction of reward to a non-zero amount.
He tested this using varying numbers of pellets in GB-l
in training and examined the effect of reducing these to
the same amount of increase in running speed in A-2 follow-
ing reward decrease was about proportional to the reduction
in reward. He concluded that this finding may support
either Amsel's theory of frustration or Spence's theory of
incentive motivation which stresses the value of G-2 rela-
tive to G-1 and does not require the introduction of the
concept of frustration. One incidental finding in this
study was that rewards larger than eight pellets produced
decrements in the speed of the next response.

In replicating Bower's procedures, McHose and
Ludvigson (1965) found no support for the hypotheses that
incomplete reward reduction produces a graded frustration
effect. They interpreted their finding as indicating "that
these A-2 speeds are simply function of absolute reward
magnitude in GB-l, and that no frustration-drive increment
from incomplete reward reduction need to be involved'(p.

495). They concluded that no existing theory, either that

ll

of Amsel, Bower, or Seward §E_al,, can satisfactorily
explain the results of their study. Similar results were
obtained by Barrett, Peyser, and McHose (1965). They found
that complete reward reduction facilitated A-2 running
speed, but incomplete reward reduction did not. These two
studies both found that post-shift speeds in A-2 were in-
versely related to absolute (post-shift) GB-l reward magni—
tudes, but unrelated to expected reward magnitudes (i.e.,
magnitudes used in original training).

Daly (1968) also studied the role of incomplete
reward reduction using a design similar to that of Bower.
Goal box two contained six pellets throughout the study.
The number of pellets in GB-l was shifted in a between
subjects design from 15 to 6, 15 to l, and 15 to 0; and
from 6 to 1, 6 to 0; or from 1 to 0. She also included
control groups, 15 to 15, 6 to 6, l to 1, and 0 to 0. The
observed increases in A-2 running speeds following the
shifts were complicated by the effects of the size of
reward used in training. The A-2 speeds following reward
were slower than the A—l speeds for the lS-pellet group.
This was seen as due to an apparent inhibitory effect of
six pellets in GB-2. Receiving six pellets in GB-2 after
having received 15 in GB-l produced an inhibitory effect
on A—2 running speed. Receiving six pellets in GB-2 after
having received one in GB-l produced an excitatory effect.

These effects increased over trials and occurred in the

12

alley closer to the frustrating (smaller) reward. An im-
portant implication of all this is that groups which
receive a smaller reward in GB-2 than in GB-l have experi-
enced frustration throughout acquisition and are, there-
fore, not an adequate test of FE.

At about the same time Meyer and McHose (1968) com-
pleted a study of the effects of reward increases in GB-l
of the double runway. They concluded that

the fact that both reward increases and some reward

decreases produced facilitation, but that no changes
in reward have produced debilitation in A2, warrants
the conclusion . . . that magnitude of change per se
is responsible for the facilitative effects obtained
in the present study and in the studies shifting to

non-reward in GB—l (p. 166).

Finally, McHose (1970) examined the relative effects
of differential magnitudes of reward in GB-l and GB-Z and
concluded that the results were uninterpretable as a "frus-
tration phenomenon" but were consistent with a discrimi-
nation contrast hypothesis as was also suggested by Daly.

One last variable studied in relation to frustration
in the double runway is the role of number of previously
rewarded trials in A-l (Stimmel & Adams, 1969). Their pro-
cedure was different from those of the experiments previously
discussed since these experimenters substituted regular
extinction for the test period where randomly interspersed
rewarded trials had been given. They found that although

there was an increase in running speeds during the six days

of non-reinforcement in GB—l, it did not occur until after

13

five non-reinforced trials in GB-l. They interpreted
these results as not supporting a frustration interpre-

tation.

Operant Response

 

In the area of operant responding, Notterman and
Mintz (1965) have examined the effect of frustration on
the force of a bar-pressing response. Their study was
quite extensive and indicated that response force increases
during extinction. Notterman and Mintz discussed three
major interpretations of the mechanisms causing the rise in
response force during extinction trials.

The first explanation is a biological one. It is
at least conceivable, according to Notterman and Mintz,
"that the behavioral phenomenon of exerting additional
force following non—reinforcement is a biological trait
resulting from the process of natural selection" because
exerting this extra force raises the probability of success
after an initial failure. They see this as most closely
associated with Amsel's conception of a "primary aversive
motivational condition."

The second is a learning interpretation. Its major
prOponent was Schoenfeld (1950). He conceived of response
force as made up of various subclasses of responses, that
is, the various levels of response force may be viewed as

subclasses of the response. He hypothesized that a:

l4

. . . wide sampling of response magnitudes or sub-
classes is reinforced during the early stages of
acquisition and that each of these subclasses must
then be separately extinguished. It is for this
reason that higher forces as well as lower forces
are observed during extinction.
If Schoenfeld was correct, the distribution of
forces during extinction should be similar to the distri-
bution during acquisition. This was tested by Goldberg
(1959) using the apparatus of Notterman and Mintz. It was
found that although the two distributions were very similar,
there was a greater frequency of higher force responses
during extinction. This was observed for each of the sub-
jects in the experiment.
Observing this, Notterman and Mintz developed a
third interpretation of the data which they call "inci-
dental directional reinforcement," which they explained
as follows:
Biological organisms are exposed from birth to
reinforcement contingencies in which they successively
more vigorous responding tends to produce reinforce-
ment. Thus, the animal learns, incidentally, that an
increase in response vigor tends to increase the
probability of reinforcement.

It is this interpretation which they feel underlies their

experimental results, but they do not consider theirs to

be a critical test.

In a lever-pressing experiment with rats, Cole and
Van Fleet (1970) examined the effects of interoceptive and
exteroceptive cues associated with differential reward

schedules in a "Skinner box analogue of the double runway."

They found no frustration effect for the latency data.

15

However, they did find a significant effect due to the
reward--non—reward (R-NR) blocks of trials for the other
measure, cumulative response time on a second lever. In
other words, following trials where lever l was not re-
warded, the subjects responded more efficiently, i.e.,
spent less time on lever 2. They found no significant
interaction between reward schedules and the R-NR effect.
To the extent that they obtained an R-NR effect, their

results seem to be consistent with a frustration hypothesis.

Single Runway

 

Goodrich (1959) incidentally examined the con-
ditioned frustration effect in relation to segments of the
runway apparatus. Using a partial (50%) reinforcement
group and a continuous (100%) reinforcement group, he studied
the running speed in the start box, alley, and goal box. He
found that in the early trials, the 100% group ran faster
in the start box and alley than did the 50% reinforcement
group. However, this phenomenon was reversed on later
trials (a cross-over effect). This is in accordance with
the postulates of Amsel's theory. Conditioned frustration
(rf-—sf) is assumed to gain control of the response later
in a partial reward schedule and this tends to make the
running speed at asymptote faster for the partial reinforce-
ment group than for the continuous reinforcement group.

The early-trial 50% reinforcement decrement persists for a

greater number of trials as performance is measured closer

16

to the goal region of the runway. In one of the two experi-
ments discussed by Goodrich, this early trial partial
reinforcement decrement at the goal region persisted for

60 trials. Goodrich ruled out the possibility of the
relatively slower performance of the 50% group in the goal
box region being due to subjects slowing down as they
approached the empty food cup on non-rewarded trials by
showing that goal Speed was the same on rewarded and non-
rewarded trials.

Goodrich concluded that although his results are in
accord with the basic idea of conditioned frustration theory,
the early appearance of cross-over suggests that other
factors may be involved. In Amsel's study (Amsel, 1962),
the faster speeds of the partial reinforcement group did
not show up until some 20 or 30 trials had been run. In
the Goodrich study the facilitation of running speed for
the 50% group developed as early as trials 9-12. This
earlier effect could be due to the narrow width of the
Goodrich runway which may have discouraged competing
responses, but Goodrich concluded that there exists no

clear and adequate explanation of his results.

Panel Pushing

 

Robert Davis (1957) measured response force and
latency as a function of drive level. During extinction,
he observed a non-significant increase in response force
for the second and third trials in the high drive group.

This increase was attributed to frustration.

 

17

The Present Experiment

 

A panel push apparatus very similar to the one
Davis used, was used in the present research in order to
test the generality of Amsel's hypothesis and to elimi-
nate many of the complications already mentioned that are
involved in this area of research. It consisted of a
chamber with a door which when raised provided access to
a panel. The food cup was located just behind this panel.
Measures of both response force and latency were possible
on separate trials without handling the subject. Lowering
and raising the door defined the end and start of a trial,
respectively.

By measuring response force and latency the
generality of Amsel's hypothesis of increased response
region following non-reward could be examined. In the
earlier studies cited above, response vigdr was examined
by measuring running speed (Amsel & Roussel, 1952; McHose
& Ludvigson, 1965), and by measuring bar press response
force (Notterman & Mintz, 1965). Thus, in the panel push
apparatus the effect of non-reward on two new measures of
response region could be studied.

In the double runway apparatus the subject remained
in GB-2 until the door into A—2 was opened. This was neces-
sary to eliminate the possiblefinfluence of handling the
subject on what could be an emotional response. In the
panel push apparatus also, the subject was not handled

between trials.

18

If the lower A-2 running speed in the double runway
following a large reward in 6B-l was due to an increased
tendency to remain in 6B-l, utilization of the panel push
apparatus would eliminate this interfering response tendency,
since approaching the panel on the next trial is not in-
compatible with the tendency to remain in the goal area, on
rewarded trials.

Use of the panel push apparatus also eliminated the
possibility of the develOpment of dual goal box expectancies
which Daly and MCHose (1968) see as controlling A-2 running
speed. In order to eliminate this problem each subject
was exposed to only one size of pellet in the panel push
apparatus, therefore making the subjects' eXpectancy level
more clearly definable.

Thus, the present experiment, through utilization
of the panel push apparatus eliminated the demotivational
influences and the influences of dual goal box expectancies
while studying the generality of Amsel's hypothesis to two

new measures of response region.

Hypotheses

 

1. It was hypothesized that response vigor as measured
by the force of striking the panel would increase
immediately following the first non-rewarded trial.
A positive result on this test would support Amsel's

theory of frustrative non-reward.

19

The second hypothesis was that response force
should decrease following the initial increase
when non-reward is continued. This result would
be in agreement with that of Davis (1957) in the
panel push apparatus. This decrease would not
directly contradict nor would it support Amsel's

theory.

The amount of immediate effect of non-reward on
response force was hypothesized to vary directly
with magnitude of reward used in training. This
would imply that the reward magnitude was posi-
tively related to magnitude of rg, the anticipatory
goal response, which Amsel viewed as controlling

the amount of frustration aroused by non-reward.

The amount of ultimate decrease in response force
during extinction was also hypothesized to be
directly related to magnitude of training reward,
This would be similar to the effect, most frequently
attributed to Armus (1959), in which amount of
reward in a runway situation is inversely related

to the number of trials to extinction.

Finally, it was hypothesized that the magnitude of
reward would have a direct influence on response
force and latency during the acquisition of the

panel push response. More explicitly, it was

20

predicted that as the magnitude of reward was
increased, response force would increase and

response latency would decrease.

METHOD

Apparatus

 

A panel push apparatus was used in which it was
possible to measure response latency and force. It was
constructed of one-half—inch plywood and its dimensions
were 24 in. long, 12 in. wide, and 12 in. deep. A guil-
lotine door at one end of the chamber, when raised, gave
access to a 4 in. by 2 in. panel door hinged at the top.

A thin metal rod, hung at the top of the panel, was twisted
so that it extended to the back of the panel in one
direction and to the top of the apparatus in the other
direction. The rod was designed so that the lower half
rested on the panel and rode back on the panel as this

was pushed open. The upper half came forward toward the
chamber with any movement of the panel. By means of this
rod, force applied to the panel was transmitted to a screw
which was attached to the side of a light wooden wheel
which rotated on its center. The force of the response,
pushing on the panel, could be measured by the displace-
ment of the wheel, measured in degrees. Because the wheel
offered little resistance to the metal rod, the initial
movement of the panel caused it to turn out of the range

of further movements of the rod. That is, the distance

21

 

22

that the wheel was displaced depended upon the force of
the initial push on the panel and not merely upon the
distance through which the panel was moved.

Response latency was measured by a Standard
Electric Timer, connected in series through two micro?
switches. The first switch was located at the tOp of the
guillotine door. When the door was fully raised this
switch was closed. This started the timer and defined the
beginning of a trial. The second switch closed when a
brass strip was displaced. The strip was located behind
the panel so that when the panel was one-half opened the
switch was Opened. Thus, the timer started when the guil—
lotine door was raised and stOpped when the panel was
pushed open.

Located behind the panel was a three sided enclo-
sure. The floor of this held the small plastic cup in
which the reward pellets were placed. The sides of this
small enclosure prevented escape and discouraged explora-
tory behavior. The base of the microswitch, behind the
panel, prevented the panel from swinging completely Open,
helping to prevent escape.

The interior of the chamber was unpainted as was
the area behind the guillotine door. The top of the
chamber was clear plexiglass which could be raised to
give entrance to the box. The apparatus is shown in

Figure l.

 

 

 

23

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

FESSOCKL‘C\\QQ\VJ\\D\
\
\
;£‘ ; “:' .\ _Wheel
-1, \ \ \l\ K\:\\ \
\ \
W E
K ‘7 \\+ Metal rod to whee1
\ >_
t :21.
1t :,,’—'Panel switch
\ \,
K \
‘\ Panel
'\ x .
x
g
3 \
'V\\Y\\\\\\ }
Guillotine Guillotine

Guillotine door H5
\-

door switch

S:§§§*—‘ door frame

 

 

PlexiglasI—r‘
door

 

3
o

I/7rZIIIIq<;;\

 

 

a.
Wheel
‘1 Metal rod
r to wheel

Panel switch

Panel

‘ Food tray

 

 

 

 

 

xx 3}

Cross-sectional side view

Figure 1. Drawing of apparatus discussed in text.

24

The experiment was conducted in a white room which
was dimly lighted and relatively quiet. The rats were main-

tained in the same room in which they were run.

Subjects

Forty, male, albino rats from Spartan Research
Animals, approximately 100-175 days old at the beginning
of the experiment were used as S5. They had been previously

used for a study which had not called for food deprivation.

Overview of Method

 

Four subjects from the same group were housed in
each cage. A 23—hour food deprivation cycle was maintained
throughout the experiment. Following each day's trials
for all groups, the SS were given food in the home cage.
Following the l-hour free feedings, the remainder of the
food was taken from the cage. Observation assured that
l-hour was enough time for all the animals to eat to
satiation. This schedule was maintained to allow the
smaller reward groups to maintain a level of deprivation
approximately the same as the larger reward groups. Water
was continuously available in the home cage.

The 40 subjects were divided into five independent
groups with eight subjects in each group. On each train-
ing trial and on each rewarded trial during testing, each
subject received one pellet of the size designated for his
group. The sizes of pellets used were 20 mg, 45 mg,

97 mg, 190 mg, and 500 mg. These sizes were chosen since

25

they are the standard pellet sizes available from Noyes.
One pellet of various sizes was used rather than various
numbers of the same size pellets since this eliminated

the necessity of a subject repeatedly returning to the
panel for the remainder of his reward. Usually, a subject
would take the pellet back to the chamber to eat it. By
using one pellet the subject would have to approach the
panel only once, which permitted the effect of the abso-
lute size of the reward pellet to be more clearly measur-

able.

PreliminarygTraining

 

The SS were placed on food deprivation for approxi-
mately three weeks prior to the first day of shaping. They
received four blocks of mouse food (approximately 12 grams)
in their home cages each day. After the first day of
shaping, the SS were placed on the 23-hour food deprivation
cycle which was maintained throughout the experiment.

Prior to the first shaping day, the §$ received
four days of hand habituating and habituation to the small
waiting cages. This consisted of E lifting the S, holding
it for a few seconds and returning it to the waiting cage
with a small chunk of food. The day before shaping began
each §_was placed in the experimental chamber with the
guillotine door open and given a small amount of food, of

the appropriate size, in the chamber near the panel.

 

26

Shaping the panel push response lasted three days
with five trials per day. The first day the panel was
prOpped fully Open with the appropriate pellet in the food
cup for three trials and one-half open for two trials. The
second day the panel was one-half Open for the first three
trials and one-fourth Open on the last two. On the third
day the door was propped one-fourth Open for the first
three trials and fully closed for the last two.

Acquisition sessions, five trials per day for nine
days, were given seven days a week for all groups. The SS
were placed in the chamber from the waiting cages. Thirty
seconds after gs were placed in the chamber the guillotine
door was raised and the first trial begun. A trial con-
sisted of the approach to the panel which provided access
to the food cup. When the SS had withdrawn from the panel
far enough for the guillotine door to be closed the trial
was completed, and the inter-trial interval of 30 seconds
was begun.

The second stage of the experiment was identical to
the first, except that eight trials per day were given; the
first three rewarded and the last five non-rewarded. Thus
providing four trials prior to the withdrawal and four
trials following the withdrawal of reward. Stage II lasted
for four days.

The extinction stage (Stage III) was conducted

similarly to the first two stages except that reward was

27

completely eliminated. Extinction lasted five days with

five trials per day.

RESULTS

At the end of the acquisition stage of the experi-
ment, the two groups which received the largest reward,
G-SOO and G-l90, showed the greatest response force. The
other three groups showed the Opposite effect. The small-
est of these, G-20, showed the greatest response force,
G-45 showed the next greatest and 6-97 showed the lowest
reSponse force. The daily mean response force for each
group for each day of acquisition is presented graphically
in Figure 2. The analysis of variance on these data indi-
cated that the effect of reward magnitude on response force
was significant (F = 4.52; df = 4, 35; p < .01), although
no direct relationship between reward magnitude and response
force was found. The results Of this ANOVA, given in
Table 1, further indicated that the effect on response
force of days of training was not significant (F < 1),
and no significant interaction effect between days Of
training and reward magnitude was Observed (F < l).

Latency of response in acquisition appeared to be
related to magnitude of training reward. The longest
response latencies were observed for the two smallest

reward groups, G-20 and G-45, but the response latencies

28

Mean Response Force (in degrees)

29

55 j

 

 

 

 

 

 

 

° 0 500
A» A.l90
D D 97
’ 0 45
A ‘ l 20
50 - A\/\
A A
o 500
A
45 - \ ¢ 190
A
//// \\\\o
\o/
o
40 — //,/
D \ ‘
. g 45
D \./
A
O / O
' 97
D n
35 -w 0/ /
\D
30'-
25 I I T I I I l I
1 2 3 4 5 6 8 9
Days

Figure 2.
over days.

Mean response force in training by groups

30

 

vmmH.mmN.H 0mm mmm.HmN.Nmm HOHHm

 

m: Av mmns.e~m mm mes.mme.mm meme ma

opsuwcmmfi pumsmm

ms av Hom.Hmm m Hm.mba.v mama

nmmo.ona.w mm hm.mma.mmm Houum

Ho.v ewmm.s mmms.mmm.em e mm.mem.nva mesuacmms cumsmm
m monmsqm Eoommnm monosqm cowumaum>
m cow: mo momumma mo mEOm mo mousom

 

.H momum .musmoma map mm OOHOM Oncommmn nuflz .mmmnu mo coauomnmuca emu
paw .mcflcflmuu no what .mpsuflcmmfi pneSmH mo muommwm do OOGMHHM> mo mflmhamad .H canoe

31

of the other three groups were approximately equal. The
mean response latencies for each group for each day of
acquisition are shown in Figure 3. The analysis of vari-
ance on these data indicated that the Effect of reward
magnitude on response latency was significant (F = 5.28;
df = 4, 35; p < .01).

As may be seen in Figure 3, response latencies for
all groups grew shorter with days of training. The effect
of days on response latency was found to be significant
(F = 10.43; df = 8, 280; p < .01). Although the effect
of the interaction between days of training and reward magni-
tude was not significant, the relationship was in the pre-
dicted direction. The response latencies Of the three
largest reward groups reached asymptote more quickly than
G-45 and G-20, further, the asymptote for the three large
reward groups was lower. The results of the analysis of
variance for these response latency data are given in
Table 2.

Each of the four test sessions (Stage II) consisted
of three reinforced trials followed by five non—reinforced
trials. Figure 4 shows the mean response force of each
group for each trial of the first day of the test stage.
Contrary to the prediction, a decline in response force
was observed in four of the five groups from trial four
to trial five, the trial immediately following the first

non-rewarded trial. However, this immediate decrement was

Mean Response Latency (in seconds)

15‘-

 

0 500

14 " A A 190

 

 

. . 45
13 —« ‘-—-‘ 20

12-—

11 fi

A

.\ .

3- ._,/\ .
2— “Kg;\ \;/*<\;

1- VK

 

 

"j
d
d
—
.1
_
—
fl
—1

Days

Figure 3. Mean response latency in acquisition by
groups over days.

33

 

 

 

ommm.¢nm omm Hmmn.mmm.wn HOHHm

ma seam.a mema.emm mm mmmm.emm.ea mama ma

Opspflcmma pnm3mm

Ho.v mnmv.oa mumm.mom.m m wovm.aom.mm mumn

mhmv.mmm mm momm.v>m.ma Houum

Ho.v wmm~.m mmom.mno.m w mam.hom.m OOODHQOOE pum3mm
m moumsqw EOOOOHm mmumswm cOHDMflHm>
m new: no momummo mo meow mo condom

 

.H mmmum .OHOmMOE may we mucouma mmcommmm npw3 .mmmsw mo GOHDOMHODGH on»
was .mcwswmuu mo mast .OOOHHGOOE UHMBOH mo mpommmm co mommauo> mo mHmMHmcd .N magma

Response Force (in degrees)

50“

45‘1

 

;/
D ‘ ‘/////A

 

 

 

  

 

O
40 ~ ‘
0 A
\\\\\\ A ' e
A
o A
e o
o
35 -
\ A
e
D
\D .
D
30 ‘
D
\\\\\\D
25
T’ 5 I l I I l I
1 3 4 S 6 7 8
R R R N N N N N
Trials

Figure 4.

Mean response force on first

day of testing by trial.

 

35

tOO small to be entirely attributed to the effects Of
non-reward.

The eight trials were divided into two blocks for
further analysis. Block I consisted of those trials which
followed reward (trials one through four) and Block II con-
sisted of those trials which followed non-reward (trials
five through eight). Figure 5 shows the mean response
force for each group for the two trial blocks on the first
day Of testing.

A decrease in mean response force comparing Block I
to Block II was observed for all groups. An analysis of
variance on these data (Table 3) found that the effect of
blocks was highly significant (F = 18.87; df = l, 35;

p < .01). That is, the mean response force for each group
was lower following withdrawal Of reward, i.e., for trials
five through eight of each test session.

The effect due to magnitude of training reward on
response force was found to be non-significant (F = 2.07;
df s 4, 35; p > .10), although the greatest drOp from
Block I to Block II was observed in the mean response force
of G-500, the largest reward group. A direct relationship
between size of training reward and amount of response
force decrement was not Observed.

Results for the other three days of testing were
very consistent with those of the first day which are given
above. The mean response force for Block II was lower than

the mean response force of Block I for each group on each

36

 

 

50
W 0500
A190
O 97
a, 45— e 45
O ‘ 20
LI
0"
[um
43 40-
(DH
now
00)
Qa'U
8.:
“:1 35‘ O
Q
(U
(D
2
30-
25

 

 

I I
Block I Block II

(Trial 1-4) (Trial 5-8)

Figure 5. Mean response force for each group for first
day of test stage by trial blocks.

37

 

 

 

 

II II m>¢.wmm mm Nmm.mmm.o~ uouum
m: Ho. moa.omm v mmv.o¢v.a mHOHHu MO mxooan
m9 OODDHOOOE Onm3wm
Ho.v hm.ma mam.mmo.aa H Nam.mmo.HH mHOfiHu mo meOHm
II II omm.ham.a mm Nmm.mom.mm Houum
ms no.m mos.meh.m s mmm.meo.ma messficmms cumsmm
Q mmumsqm EOOmmnm wwumsqm cowumfium>
m coo: mo momummo mo mEOm mo condom

 

.mmmum pmmu .H hop .OHDmOOE may no mouom Oncommmu SDHS .mxooan UGO OOODHOOOE
Oumzmu mo OOHDOOHODOH on“ was Apnmsmulcoc mcwonHOm UGO OHOBOH mcflonHOmv

mamauu mo mxooab .mOsuficmmE Onm3wu mo mpommwm MOM mocwwnm> mo wHOMHmcd

.m OHQOB

38

day Of testing. These data are presented graphically in
the Appendix along with the results of the ANOVA for
each day.

The response latency data for the test stage were
similar to those of response force. On the trial immedi-
ately following withdrawal of reward for the first test
day, the average latency over all groups stayed at approxi-
mately the same level as for the first four trials. Indi-
vidual groups showed slight declines or rises, but these
tended to average out when the groups were combined. The
mean response latencies for each group for each trial of
the first day of the test stage are shown graphically in
Figure 6. As can be seen from this figure, the mean
response latencies lengthened considerably within the
four trials following withdrawal of reward.

The eight trials were divided into two blocks, as
was done for the response force data analysis. An increase
in mean response latency between Block I (following rewarded
trials) and Block II (following non-rewarded trials) was
observed for all groups. The effect of blocks on response
latency was found to be significant (F = 13.82; df = l,

35; p < .01). The mean response latencies for each group
for Block I and Block II on the first day of testing are
shown in Figure 7. The results of the ANOVA for the effects
of blocks of trials, reward magnitude, and the interaction

between blocks and reward magnitude used in training are

Mean Response Latency (in seconds)

39

 

 

 

 

0 (>500
30‘. A A 190
D E) 97
O O 45
‘-————‘~ 20
25.1
20 -

lsJ

 

 

 

10 A
5 —d
O
.A/I/z’ ‘\\\\\ ff”,,z A
‘ : I ‘_.4' '
A .‘ . A .f ‘7'
:' ‘A'b‘ «Cs; 0 .
0 I I ‘jl I l 1 l
l 2 3 4 5 6 7 8
R R R N N N N N

Figure 6. Mean response latency for each group for each
trial on day l of test stage.

40

 

 

 

 

10'- O-———o 500
A---—-4 190
D——D 97
o e—————O 45
3__ A A 20 ‘
m
0
c
3
ma~ 5“
AIn
O
a):
0:0
:20
833 A
m 4“
GIG
06::
c
m
m
2 2__ O
A
0 I I
Block I Block II

(Trial 1-4) (Trial 5-8)

Figure 7. Mean response latency for each group for
first day of test stage by trial blocks.

41

given in Table 4, the only significant effect was due to
blocks.

The relationship between magnitude of training
reward and amount Of increase in response latency from
Block I to Block II can also be seen in Figure 7. The
greatest increase in response latency from Block I to II
was Observed for the largest reward group, the next
largest from the next largest reward group. Although a
perfect direct relationship was not Observed, the results
tentatively suggest a relationship. The effect due to
reward magnitude used in training was found to be non-
significant in the test stage (F = 1.72; df = 4, 35;

p > .10).

The results observed for the other three test
sessions were consistent with the results Of the first.
Mean response latency increased from Block I to Block II
for each day for each group. These data are presented
graphically along with their respective ANOVA results in
the Appendix.

A decrease in response force was Observed from the
last day of extinction for all five groups. The effect
of stage of experiment was found to be highly significant
(F = 30.29; df = 1,35; p < .01). The ANOVA for these data
are presented in Table 5, and the mean response force for
each group on the last day of acquisition, the first day
of extinction, and the last day of extinction are presented

graphically in Figure 8.

 

42

 

 

 

II II mmh.vmm mm Houum
m: me.~ «mm.omm w mmm.mmm.m mamanu mo mxooan an
OOODHOOOE pum3mm
Ho.v mm.ma mmm.nme.e H mmm.e~e.e mamas» mo mxooam
II II mm.oo¢ mm mmo.mmorva uounm
ma ~5.H om.omm a mma.men.~ soundness euazmm
a mmHmOOm EOOmmHm mmnmdqm OOHDOHHO>
m com: mo mmmummo mo mEOm mo wousom
.mmmum
ummu .H amp .wusmmma may mm mocmuma oncommou cuHB .mxooan Odo OOODHOOOE
pumzou mo cOfluomHOpcfl on» was AOHm3mHIcOO OOHBOHHOM Ocm pum3mu OOH3OHHOMV

macaw“

mo mxooan .mpsuficmmfi pumzmu mo muommmm How mocmwum> mo mflmwamcd

.w OHQOB

43

 

 

mwm.mm~.m mm mhma.mmm.maa Honnm
ma mhmm.a om.ae~.e e oo.mem.ma pamsflnmmxo
mo mmmum ma
OOODHOOOE pumzmm
Ho.v Hmmm.om mmam.nmh.mm a mmam.hm>.mm ucmfifiummxm mo oomum
mmom.vmw.~ mm memH.HHm.mm HOHHM
ms momm.a mmm.mmo.m e ooom.mva.ma OOODHGOOE pum3mm
m monmsqm Bovomum mmnmswm OOHDOHHO>
m com: mo momumoa mo mEOm mo condom

 

.AOOHDOOHDNO_MO One was cowuwmwnqom
mo UOOV OHOmOOE may no OOHOM wmcommmu ODAB .mmmnu mo coauomnmgcw on» was
.ucmafiuomxm no woman .mpsuwcmmfi OHMBOH mo muommmm How OOOOHHO> mo mammamcd .m OHQOB

 

44

 

 

 

50 1
O o 500
A____.A 190
D D 97
o e—————e 45
45- 4 ‘—-‘ 20
A
23
o
m
’6» 40-
m
P e
c
-H
8
H 35 - n
o
[u
m
In 0
c
o 30“
a:
c
m
m
E
25 — D
D 0
2° I I I
Last Day First Day Last Day
acquisition extinction extinction

Figure 8.

Mean response force for each group on last day
Of acquisition, first day of extinction (follow-
ing test sessions), and last day of extinction.

45

Although the reward magnitude by stage of experi-
ment (in last day of acquisitiOn on last day of extinction)
interaction was not significant (F = 1.28) it was Observed
that the larger the magnitude of training reward, the
greater the decrement in mean response force during
extinction. This relationship was observed for all but
one group. The smallest group, G-20, showed a response
force decrement in extinction larger than that of G-45
and G-97. The decreases in mean response force for each
group were as follows: G-500, 22.9; G-190, 18.1; G-97,
10.3; G-45, 6.3; and G-ZO, 13.0. However, as noted above,
these differences among groups are not significant.

Response latencies for all groups increased con-
siderably comparing acquisition with extinction, as pre-
dicted. At the end of acquisition, prior to any experi-
ence with non-reward, all five groups showed mean response
latencies between .5 seconds and 2 seconds, by the first
day of extinction the response latencies were between 5
seconds and 23 seconds, and on the last day of extinction
they were between 8 seconds and 46 seconds. These data
are presented graphically in Figure 9.

An analysis of variance indicated that the effect
of the stage Of experiment (i.e., increase in response
latency from end of acquisition to the end of extinction)
was highly significant (F = 267.63; df = l, 35). The

ANOVA for these data are presented in Table 6.

50
45
4o

75

'O

S

U 35

(D

U}

C

'31 30

>1

0

C.

3

m 25

Q

Q)

U)

g 20

In

(D

a:

g 15

0)

Z
10
5
0

Figure 9.

d

 

46

O-————o 500
A-——-—A 190
O—————O 97

e——-——e 45
A—A 20 O

D.

 

A
El
/
67’
7 I I
Last Day First Day Last Day
acquisition extinction extinction

Mean response latency for each group on the
last day of acquisition, first day of ‘
extinction, and last day of extinction.

47

 

 

mamv.nmo.a mm whmm.amo.mm Houum

Ho.v mm.mH oama.omm.ma e mmmum AA

OOODHOOOE pumsmm

Ho.v mm.nm~ mmmm.>mo.amm H mmom.wmo.amm ucmawuomxw

mo mmmum

maoa.ao~.ma mm Hmmm.mmo.hmw HOHHm

.m.c omn.a mmmo.vee.ma e mmma.eam.ee menusamma eumzmm
m mmumsqm EOOOOHm mmumsqm coaumflum>
m com: mo monummn mo meow mo condom

 

.musmmma map mm Mocmpma oncommmh
zufl3 muommmm Ommsu mo cofluomumucfi way was .Acofluocfluxo Ocm coauwmfisqomv
pcmafiummxm mo mmmuw .OOODHOOOE Onmzmu mo poommm Mom mocmflum> mo mammHmcd .m magma

48

The interaction between reward magnitude used in
training and stage Of experiment (end Of acquisition or
end of extinction) was also found to be significant as
may also be seen in Table 6. Although direct relationship
between amount of reward and change in response latency
was not observed the largest increase in response latency
during extinction was observed for the largest reward

magnitude group, G-500.

DISCUSSION

A decrease in response force following withdrawal
of reward was Observed rather than the predicted increase
in response force. This decrement was small at first, but
it continued during the four trials following non-reward
during each day of the test stage. These results are
Obviously not in agreement with Amsel's theory of frus-
trative non-reward. It is possible, however, that a
reasonable interpretation of the diverse effects is avail-
able in a theory proposed by Denny, namely, elicitation
theory (Denny, 1967, 1971, 1971; Denny & Aditman, 1955).

In elicitation theory it is postulated that
extinction occurs because of effective competition by com-
peting responses (RC) in the same or similar stimulus
situations. The effectiveness Of the competition between
S-RO and S—RC is a function of the degree of incompati-
bility Of R0 and Rc' A further postulate of elicitation
theory is that an elicitation hierarchy describes the
hierarchy eliciting value of all the stimuli in a stimulus
situation. The response of the stimulus-response tendency
which is momentarily strongest is the one that occurs

(Denny, 1970).

49

50

Adelman and Maatsch (1955) studied the role that
the type of RC elicited by non-reward played in the ex-
tinction process. They allowed three separate groups to
extinguish under one of the following conditions:

a) allowed to jump out of the goal box onto a two-
inch ledge for twenty seconds;

b) allowed to recoil from the goal box for a period
of twenty seconds and retrace;

c) stayed in the goal box for twenty seconds--
regular extinction.

The results indicated that extinction occurs most
rapidly when the response elicited by non-reward was
directly incompatible with R0. That is, the recoil group
extinguished fastest since recoil was directly incompatible
with approach; and the jump out group which was allowed to
escape from the goal box and spent the 20 seconds in another
stimulus situation, extinguished the slowest. In fact,
they hardly extinguished at all within 100 massed extinction
trials. It was concluded that the jumping response was
directly elicited inside the goal box and conditioned to
these cues. The jump-out response was not incompatible
with approach and was assumed to be an extension of this
response chain.

The panel push apparatus, as previously described,
does not allow g to excape from the goal area behind the
panel by continuing on in the same forward direction. The
situation is like regular extinction or the "recoil"

condition in the Adelman and Maatsch study. The available

responses when reward is eliminated are limited to remaining

51

near the empty food cup or recoiling from it. Recoiling,
the most likely response, according to elicitation theory,
is directly incompatible with the R0 of approaching the
goal. We would expect, with repeated non-rewards, that as
RC is frequently and consistently elicited, it becomes
conditioned to S and interferes with R0, and as non-reward
continues, and R0 is interfered with, the force of Ro
should decrease.

An increase in response latency was Observed follow-
ing withdrawal of reinforcement. This increase occurred
within the first four trials following onset of non-reward
in the test stage and continued through extinction. This
increase in response latency of RO also could be attributed
to the effective competition Of RC. As non-reward is
repeated RC would be more consistently elicited and Ro
would become less likely to occur. Therefore, the latency
of RO would tend to increase just as force of RO would tend
to decrease.

The assumption that incentive size, since it is
assumed to partially control the strength of rg, will be
directly related to the amount Of frustration effect,
found only limited support in the present study. In part,
this may be due to the type of reward magnitude used, size
Of pellet rather than number of pellets. The results were
in the predicted direction (the larger reward groups showed
larger decreases in response force in extinction) except

for the smallest group (G-20). The eliciting value of one

52

20 mg pellet may be so small as to yield a response
strength for R0 which could easily be outweighed by com-
peting responses. This could result in faster extinction
and greater response force decrement for this group as
Opposed to the larger reward magnitude groups. Except for
the 20 mg group, the relationship observed between the
amount of decrease in response strength and the reward
magnitude used during acquisition was in keeping with the
predictions based on the Armus (1959) effect.

As expected latency of responding decreased signifi-
cantly with days of original training, as the criterion
response (R0) effectively competes with other responses,
latency decreasing more quickly the larger rewards, although
the three largest reward groups were almost exactly equal in
performance. This is perhaps not unexpected in light of
research which indicated that the effect of incentive size
has little effect on the speed of acquisition of simple
responses unless the incentive size is extreme.

Let us also see if this theory can shed light on
some of the results in the double runway. This apparatus
permits S to make responses following non-reward which are
not necessarily incompatible with approach, being somewhat
similar to the jump-out condition of the Adelman and
Maatsch (1956) study. Here, it was found that a jump-out
response elicited by non-reward, when an anticipation Of
reward had been developed, was learned faster than a jump-

Out response reinforced with familiar food. Thus, a

53

combination in which elicitation Of the response of
running away from an empty goal box is joined with the
elicitation Of approach to food in GB—2, would predict
a strong response in A-2, as generally is found in the
double alley.

The results obtained by Notterman and Mintz (1965),
namely, that frustration increases the amplitude of the
bar press response, can also be seen as consistent with
the elicitation position. In the operant situation, the
food tray is separate from the bar, and the instrumental
bar-pressing response represents an available alternative
response, an elicited, tangential RC occurs when the food
tray is empty. Such a directly elicited response could
increase the force with which the bar is pressed during
the first part of extinction. Although this directly
elicited component of the total Rc complex is incompatible
with approach to the food tray it is quite compatible with
the operant bar-press and the two summate. In the present
study, on the other hand, panel push is an integral part
of approach to food, closely associated in time with a
directly elicited withdrawal response. Thus, it is con-
gruent for frustration to increase the force of the operant
in the Notterman and Mintz study and decrease the force Of
the Operant in the present study where the operant is a
part of the approach response to food.

This interpretation is also compatible with Good—

rich's finding in the single runway apparatus that the

 

54

faster running speeds for partial reward groups were not
Observed for many trials in the area closest to the goal
as compared with the 100% reward group. Competing responses
which are elicited by the empty goal cup would compete
more effectively with that part of the running response
chain which is closest to the goal cup.

The results of the present experiment were inter-

preted as not supporting Amsel's theory of frustrative non-

_.__.___5

reward. A decrease in response force was Observed rather
than the increase which would be predicted from Amsel's
theory. These results and the results of earlier studies
were examined in terms of the competing response view Of
extinction (of elicitation theory). The analysis Of the
effects of the magnitude of reward used in training only
partially supported the hypothesis that this variable is
related to response force decrement in extinction (in—

versely related, as is the case in the present experiment).

REFERENCES

REFERENCES

Adelman, H. M., & Maatsch, J. L. Resistance to extinction
as a function of the type of response elicited by
frustration. Journal of Experimental Psychology,
1955, £9, 61-65.

 

Adelman, H. M., & Maatsch, J. L. Learning and extinction
based upon frustration, food reward, and explora-
tory tendency. Journal of Experimental Psychology,
1956, sg, 311-315.

 

Amsel, A. The role of frustrative non-reward and non-
continuous reward situations. Psychological

 

Amsel, A. Frustrative nonreward in partial reinforcement
and discrimination learning: Some recent history
and a theoretical extension. Psychological Review,
1962, 62, 306-328.

 

Amsel, A., & Hancock, W. Motivational properties of frus-
tration: III. Relation of frustration effect to
antedating goal factors. Journal of Experimental
Psychology, 1957, §§, 126-131.

 

 

Amsel, A., & Penich, E. The influence of early experience
on the frustration effect. Journal of Experimental

 

Psychology. 1962, g;, 167-176.

 

Amsel, A., & Roussel, J. Motivational properties of frus—
tration: I. Effect on a running response of the
addition of frustration to the motivational com-
plex. Journal of Experimental Psychologx: 1952,
5;, 363-368.

 

Amsel, A., & Ward, J. Motivational prOperties of frus-
‘ tration: II. Frustration drive stimulus and
frustration reduction. Journal Of Experimental
Psychology, 1954, fig, 37-47.

 

 

Armus, H. L. Effect of magnitude Of reinforcement on
acquisition and extinction of a running response.
Journal of Experimental Psychology, 1959, 28,
61—63. ‘

 

55

 

56

Barrett, R. J., Peyser, C. S., & McHose, J. H. Effects of
complete and incomplete reward reduction on a subse-
quent response. Psychonomic Science, 1965, 3,
277-278.

 

Bower, G. The influence of graded reductions in reward
and prior frustration events upon the magnitude
of the frustration effect. Journal of Comparative
Physiological Psychology: 1962, ééj 582-587.

 

Cole, M., & Van Fleet, F. The frustration effect as a
function Of interoceptive and exteroceptive cues
in a Skinner box analogue of the double runway.
Psychonomic Science, 1970, g9, 33-35. a

 

Daly, H. B. Excitatory and inhibitory effects of complete
and incomplete reward reduction in the double run-
way. Journal Of Experimental Psychology, 1968,
lg, 430—438.

 

Davis, R. H. The effect of drive on latency, amplitude and
activity level. Journal Of Experimental Psychology,
1957, 53, 310-315.

Denny, M. R. A learning model. In W. C. Corning and S. C.
Ratner (Eds.), Chemistry of Learning, New York:
Plenum Press, 1967.

 

Denny, M. R. Relaxation theory and experiments. In F. R.
Brush (Ed.), Aversive conditioning and learning.
New York: Academic Press, 1971, in press.

Denny, M. R. A theory of experimental extinction and its
relations to a general theory. In H. H. Kendler
and J. T. Spence (Eds.), Essays in neobehaviorism:
A memorial volume to Kenneth W. Spence. New York:
Appleton-Century-Crofts, 1971, in press.

 

Denny, M. R., & Adelman, H. M. Elicitation theory: I.
An analysis of two typical learning situations.
Psychological Review, 1955, 63, 290-296.

Goldberg, I. A. Relations of response variability in
conditioning and extinction. Unpublished doctoral
dissertation, Columbia University, 1959.

Goodrich, K. P. Performance in different segments Of an
instrumental response chain as a function of rein-
forcement schedule. Journal of Experimental Psy-
chology, 1959, 21, 57-63.

57

Hull, C. L. A behavior system. New Haven: Yale Uni—
versity Press, 1952.

 

Krippner, R. A., Endsley, R. C., & Tucker, R. S. Magni-
tude of G-l reward and the frustration effect in
a between subjeCts design. Psychonomic Science,
1967, 2, 385-386.

McHose, J. H. Relative reinforcement effects: S /S2 and
Sl/S paradigms in instrumental conditioning.
PschOlogical Review, 1970, 77, 135-146.

McHose, J. H., & Ludvigson, H. W. Role of reward magni-
tude and incomplete reduction of reward magnitude 4
in the frustration effect. Journal of Experimental
Psychology, 1965, 19, 490-495.

 

McHose, J. H., Meyer, P. A., & Maxwell, R. R. Frustration
effect as a function of training magnitude in a
within-S design. Psychonomic Science, 1969, ii,
137-138.

 

Miller, N. E., & Stevenson, S. S. Agitated behavior Of
rats during experimental extinction and a curve of
spontaneous recovery. Journal of Comparative
Psychology, 1936, 2;, 205-231.

 

Notterman, J. M., & Mintz, D. E. Dynamics of response.
New York: John Wiley & Sons, Inc., 1965.

Peckham, R., & Amsel, A. Magnitude reward and the frus-
tration effect in a within-subjects design.
Psychonomic Science, 1964, A, 285-286.

Peckham, R. H., & Amsel, A. Within-subjects demonstration
of a relationship between frustration and magnitude
of reward discrimination. Journal of Experimental
Psychology, 1967, 1;, 187-195.

Schoenfeld, W. N. On the difference in resistance to
extinction following regular and periodic rein-
forcement. Notes, Conference on the Experimental
Analysis Of Behavior, Indiana University, 1950,
no. 20. (Mimeographed.)

Seward, J. R., Pereboom, A. C., Butler, B., & Jones, R. B.
The role of prefeeding in an apparent frustration
effect. Journal of Experimental Psychology, 1957,
5g, 445-450.

 

Skinner, B. F. Are theories of learning necessary?
Psychological Review, 1950, 51, 193-216.

 

58

Stimmel, D., & Adams, P. Magnitude of frustration effect
as a function of number of previously rewarded
trials. Psyphonomic Science, 1969, 1Q, 31-32.

Wagner, A. R. The role of reinforcement and nonreinforce-
ment in an "apparent frustration effect." Journal
of Experimental PsychOlOgY, 1959, g1, 130-136.

APPENDIX

59

 

 

mwmn.hvm mm mnmv.o>m.m~ Hounm

me we. mmmm.ass s ocmo.mms.a mesuflcmme

ma mxooam

Ho.v NH.mH mmao.nmn.aa a mmao.hmh.aa mxooam

mNHo.mmh.H mm mnmv.mon.am HOHHm

.m.c HMHO.N mnmw.oom.w v mm.emv.ma OOODAOOOE pumBmm
m mmumsqm EOOmmHm moumsqm OOADOHHO>
m com: mo moonmmo mo mEOm mo monsom

 

.HH mmmum

.N MOO .OHOmMOE mnu mm OOHOM mmcommmu nuw3 coauomumucfi mxooan an OOODMOOOE
Ono .mamwhu mo mxooan .OOODAOOOE Ohm3mu mo muommmm so mocmwnm> mo mHmMHmc< .¢ OHQOB

Mean Response Force
(in degrees)

Figure A.

501

45‘-

40“

35“

30-‘

 

25

60

O-————O 500

 

 

 

A A 190
97
A e O 45
A___A20
O
A
e
A
D\\\\\\\\\\\
D
If I
Block I Block II
(trial 1-4) (trial 5-8)

Mean reSponse force for each group for
second day of test stage by trial blocks.

 

61

mHmh.mmm mm mNHm.w¢H.mN HOHHm

 

 

m: mm. mmee.mm~ v omsm.omm mesuflcmme
ma mxooam
Ho.v Hm.m mmam.hmo.n H mmam.>mo.n mxooam
mamm.mom.a mm mmam.vma.mm Honum
ma A~.H mmeo.mmo.~ s omsa.ema.m menusqmms eumzmm

m mmumsqm Eovwwum mmumsqm soaumfium>

m coo: mo mmmummo mo mEOm mo condom

.HH ommum

rm mop .Ousmmmfi map mm mouom oncommmu ODH3 cowuomumucfl mxooan an OOODHOOOE
UGO .mamflnu mo mxooan .mpsuflcmmfi Ouw3mu mo muommmo co mocmfluw> wo mammamcd .m OHQOB

62

 

 

45-1 0—0500
A—————A 190
m D—————D 97
o e-———-e 45
8,. 40- g\ A———A 20
m
07m
'35
8.4 35"
8c: 2
5::
g 30..
o
z
25 | I
Block I Block II

(trial 1-4) (trial 5-8)

Figure B. Mean response force for each group for third
day of test stage by trial blocks.

63

 

 

s~.~so mm omsm.mmm.m~ gonna

.m.c ms.m mmso.aam.~ s omsa.eeo.oa oosuflooos

an mxooam

Ho.v mo.Hm me.omm.o~ H m¢.omm.o~ mxooam

sooo.~mm.a mm mma.mmm.os House

.m.o mmn.m moom.omm.m a mmo.oom.ea oosuflomos ouosom
m mmumsqm EOpoonm mmumswm moamfluw>
m coo: mo mooumoo mo mEsm mo mousom

 

.e mop .musmmoa may mm mouom mmcommon rues mOOHuomuouOH opsuflcomfi an mxooan
Odo mamas» mo meOHQ .mpspflcmma chosen mo mpoommo co mocmwnm> mo mHmMHmcfl

.HH omoum

.o magma

64

 

 

 

 

 

501 o o 500
A A 190
CL————CI 97
e e 45
o
8
4 —
:9... ° 0
U)
an
ma)
3:.
an» 354 ‘
a“ 0
ms: ‘
'3
fi 304 8
m
2
251
20- D
I F
Block I Block II

(trial 1-4) (trial 5-8)

Figure C. Mean response force for each group for third
day of test stage by trial blocks.

 

65

 

 

 

mmmm.mmm.a mm mmm.mmv.mv Houum

Ho.v Hmvnrva mmv.HmH.mH v mmmm.mom.mn OOODHOOOE

pumBOH >3 mxooam

Ho.v ommh.m oomm.mmm.aa H omnm.m~m.aa Mooam

moww.mvm.a mm ~mma.wma.¢m HOHHm

.m.c mmom. mmme.vmh e onem.>ma.m opsuflcmma pumzmm
m moumdqm EOOooum mmumdqm coaumflum>
m coo: mo mwmnmon mo mEOm mo mousom

 

.HH mmmum .m hop rousmmms man no MOOODMH omOOQmOH nuflz coauomuoucfl meOHn
wn mospflcmoe was mamfluu mo meOHQ OOODHOOOE pumzmn How OOOMAHO> mo mammaoam .o manna

Mean Response Latency

Figure D.

(in seconds)

66

 

 

O—————o 500
A—————A 190
U—-———D 97
11.. ‘ O-————O 45
o A————_A 20
10'-
9—1
3..
e
7 _
6 _ D
5-
4-— A
e
3..
2—
A'
O
l- U
0 I' I
Block I Block II

(trial 1-4) (trial 5-8)

Mean response latency for each group for the
second day of test stage by trial blocks.

67

 

 

meem.mnm.~ mm mOOH.mno.mm nomum
Ho.v mo.H oHMH.mm¢.~ e momm.omm.m oosuHaooa
an mxoOHm
Ho.v mm.HH ammo.Hom.om H mmmo.Hom.o~ mHOOHm
nmwm.mwo.m mm Hmme.moo.~m uouum
ma mm. amm.OHm.N s meem.Hso.oH mosoHaoos ogosom
m moumsqm EOOooum moussvm OOHHOHHO>
m coo: mo momumma mo mEOm mo OOHOOm

 

.HH omoum .m moo
.ousmmoa may no honoumH ochOmou OUH3 GOHHOOHOHOH mHOOHn ma OOODHOOOE Odo

mHmHHp mo mMOOHn .OUOHHOOOE pum3mn mo mpoommm Mom OOOOHHM> mo mHmemcfi .m OHQMB

 

 

 

20- o O 500 °
A-————A 190
D__D 97
A———__A 20
15-
>‘ e
o
O
m
4..)
(Ur-s
ham
'0
3:8
c:8 10- A
8..
m
a):
5:1
a
m
m
2
A
54 D
I
0 I I
Block I Block II

(trial 1-4) (trial 5-8)

Figure E. Mean response latency for each group for the
third day of the test stage by trial blocks.

69

 

 

vmmm.OHH.m mm hmhh.mom.mOH Houum
ms we. omHm.mHm.H v mmmo.vhm.m OOODHOOOE
an mxoon
Ho.v «N.HH mHHm.~hm.em H mHHm.~hm.vm mMOOHm
homo.mom.m mm mmmm.h>m.mmH Houum
mc mvmm. mHmv.mmn v bmmw.mmH.m OOOuHcmmE pum3mm
a monmsvm Booomum mouosvm OOHumHHm>
m com: mo mooumoo mo Sam mo oousom

 

.HH mooum .4 moo
.musmooe man we mucoumH oncommmu cuH3 OOHDOOHODOH OOOHHOOOE an mMOOHQ Odo

mHmHHu mo mxooHn .OOODHOOOE OHO3OH mo muoommm How mucmHHm> mo mHmmHmcd .m OHQOB

Mean Response Latency
(in seconds)

Figure F.

70

 

 

 

20-
.1.~ 0-———O 500
7,‘ A—A 190
o O———.O 97
' e-————4I 45
A A 20
A
15‘q
10 “I
D
’1/
5 _
/
.1
0 I I
Block I Block II

(trial 1-4) (trial 5-8)

Mean response latency for each group for fourth
day of the test stage by trial blocks.

 

MIIIIIIIIILIIIfiIflIIIIWIIIII'Es