TEEE EIFECT 01"“ {EYE 3.3.55}? 5:. EEC» 8&3 E5323 3235212351}. GP
SEMPEE END CGEEE’EEI’. ' .ESJE‘: E. .33: ”3.955 EN
JAPANESE QUAIL {23'} l} CUEEEEE‘. JAEJTIEM;

cs:

. . f 1
“team :9: Ehe agree 9E E‘EE. A

iiiiii

MW . .
kx‘umud smug ukhtswui

NI EEEZEE ”. . .L‘U..A

Fz'ffi
Us

 
    
 
 

3:293”
LI BEBE - Y
MiChig. *3 State
University

ABSTRACT
THE EFFECT OF OVERLEARNING ON THE REVERSAL OF SIMPLE AND COMPLEX
VISUAL DISCRIMINATIONS IN JAPANESE QUAIL
(COTURNIX COIURNIX JAPONIC?)
33’

Jennifer G. Fidura

The overlearning reversal effect (ORE) was first denon—
strated in a black—white discrindnation with rats. Since that time
there have been numerous attenpts to find facilitation of reversal
learning following overtraining on an original discrimination. The
ORE is contrary to the assumptions of the traditional S-R position
that a) habit strength is directly related to the number of reinforce-
ments, b) speed of extinction is inversely related to habit strength,
and c) reversal learning is simply a matter of extinguishing one
habit and learning a new one.

The most successful explanations of ORE have concentrated on
modifying the third assumption. However, any theoretical approach to
the phenomenon of ORE must include hypotheses to account for the many
failures to find the effect experimentally. As a general rule ORE is
found in visual discrindnations with rats, but not found in spatial
discriminations. Lovejoy and Mackintosh have hypothesized that if
the §_continues to attend to the relevant dimension during that per—
iod of reversal while it is learning to make the new choice response,

reversal will be relatively rapid. On the other hand, if the §_stcps

Jennifer G. Fidura
attending to the relevant dinension during reversal, reversal learn-
ing will be more difficult.

The deterndning factor in the occurrence of ORE, therefore,
is the saliency of the relevant dimension. In a discrimination of a
dimension of low saliency the overtraining will increase the proba—
bility that the §_will continue to attend to the relevant dimension

and, consequently, facilitate reversal.

Eqprfientl

 

This experiment assessed the effects of overtraining on the
reversal of two different discriminations with Japanese quail.
TWenty-four Japanese quail learned either a simple or complex color
or form discrimination in a commercial Operant chamber. One half of
the §s received overtraining to a predetermined criterion on the
relevant dimension. All §s were then given the reversed discrimina—
tion.

The ORE which was expected following overtraining on the
more difficult form discrimination was not found. Two interpreta—
tions were offered for these results: first, the criterion for
original learning (15 consecutive correct trials) was so difficult

that all §s received.scne amount of overtraining, second, the over-

training procedure itself was not sufficient to facilitate reversal.

Experiment II

a

 

This experiment was designed to assess the validity of the

two interpretations offered for the results of Experiment I. Eight

Jennifer G. Fidura
Japanese quail were run using a procedure identical to that used
above except that the criterion for original learning was eight con-
secutive correct trials and none of the §s received overtraining.
The reversal scores were compared to the appropriate scores for
those §s which received no overtraining from Experiment I. Again,
no ORE was found.

The results indicated that the criterion for learning in
Experiment I was not so difficult that all §f received overtraining
which.ndght have facilitated reversal. The data did suggest that
the overtraining procedures employed in the two experiments did not
significantly increase selective attention to the relevant dimension

and, consequently, the facilitiation of reversal.

Approved .zgiki /<E:1fb/E%Eé‘~/";;/

Date hxé’ 011/1 / 77$
._// f ’

 

THE EFFECT OF OVERLEARNING ON THE REVERSAL OF SIMPLE.AND COMPLEX
VISUAL DISCRIMINATIONS IN JAPANESE QUAIL
(COTURNIX COTURNIX JAPONICA)

By .X

{r I.
Jennifer G5 Fidura

A.THESIS
Submitted to
Michigan State University

in partial fulfillnent of the requirements
for the degree of

MASTER OF ARTS

Department of Psychology

1970

.ACKNOWLEDGMENTS

The author wishes to thank the Herbers of her thesis come
Hattee, Dr. Stanley C. Ratner and Dr. Glenn I. Hatton, for their
assistance and encouragenent both in the preparation of this manu—
script and throughout the years she spent in the Psychology Depart-
nent at Michigan State University. .A very special thanks is due
Dr. M. Ray Denny for his guidance as an academic advisor and for
his understanding as a personal friend. The many opportunities
afforded the author by Dr. Denny and his faith in her ability is
deeply appreciated.

The assistance of Mr. Robert A. Dittrich, who helped run
the subjects, is also gratefully acknowledged. There were tines
when it would have been impossible to complete the experinent on
schedule without his valuable assistance. And finally, a special
debt of gratitude is owed to my husband, Frederick; his technical
and theoretical advice, his understanding, encouragement and patience
made the completion of this work possible. It is to Fred that this

work is dedicated.

ii

TABLE OF CONTENTS

Page
ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . ii
TABLE OF CONTENTS . . . . . . . . . . . . . . . . . . . iii

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . iv

Chapter
I. EXPERIMENTS AND THEORIES . . . . . . . . . . . 1
II. EXPERIMENT I . . . . . . . . . . . . . . . . . 17
Method . . . . . . . . . . . . . . . . . . . . 22
Results . . . . . . . . . . . . . . . . . . . . 25
Discussion . . . . . . . . . . . . . . . . . . 27
III. EXPERIMENT II . . . . . . . . . . . . . . . . . 32
Method . . . . . . . . . . . . . . . . . . . . 33
Results . . . . . . . . . . . . . . . . . . . . 33
Discussion . . . . . . . . . . . . . . . . . . 3H

IV. SUMMARY OF RESULTS . . . . . . . . . . . . . . 37

REFERENCES . . . . . . . . . . . . . . . . . . . . . . 38

iii

LIST OF TABLES

Table

1. Means and standard deviations of total and error
trials to the first criterion trial for each
djmr‘Sim O O O O O O O O O O O O O O O O O O

2. Means and standard deviations of total trials to
criterion on complex discriminations . .

3. Means and standard deviations of total trials to
criterion in original learning . . . . . . . .

H. Means and standard deviations of total trials to
criterion in overlearning .

5. Means and standard deviations of total trials to
criterion in the reversal of the discrimination
with color relevant . . . . . . . . . . . . .

6. Means and standard deviations of total trials to
criterion in the reversal of the discrimination
withformrelevant

7. Summary table of analysis of variance among mean
total trials to criterion as presented in
Table 5 . . . . . . . . .

8. Summary of analysis of variance among mean total
trials to criterion as presented in Table 6 .

9. Means and standard deviations of total trials to
criterion in original learning . . . . . . . .

10. Means and standard deviations of total trials to
criterion in reversal learning with color
relevant . . . . . . . . . . . . . . . . . . .

ll. Means and standard deviations of total trials to

criterion in reversal learning with form
re levarlt O I O O C O O O C O O O I O O O O O 0

iv

Page

20

20

26

M)
O)

28

28

29

35

35

36

CHAPTER I
Experiments and Theories

Riley (1968) suggests that the phenomenon of overlearning
reversal effect (ORE) has attracted more attention and, consequently,
more research effort in the last ten years than any other facet of
discrimination learning. Adthough prior to 1953 Harlow (1949) had
described the appearance of learning sets and Lawrence (19u9, 1950)
had formulated an explanation for the speed of reversal based on the
acquired distinctiveness of cues, there was no direct empirical evi-
dence for the facilitation of reversal following additional training
trials. Reid (1953) was the first to demonstrate ORE using a simple
brightness discrimination with rats. He trained three groups of
animals to a criterion of 18 in 20 correct trials on a black—white
discrimination. One group received no further training, a second
group received 50 additional trials, and the third received 150 over-
training trials. When the relative positiveness of the cues was re—
versed, the group which had received 150 additional trials reversed
their choice response more rapidly than either of the other two
groups. Thus it appeared that the overlearning of a simple brightness
discrimination facilitates the learning of the reversal of the orig—
inal discrimination. This effect has become known as the overlearn-
ing reversal effect (ORE).

The effect of overlearning on reversal perfornence which was

feund by Reid (1953) is not only contrary to the direct relationship
between the number of trials and the difficulty of reversal found
experimentally by MCCulloch and Pratt (193”), Spence (19u5), and many
others, but also to the traditional S—R position as a whole (Riley,
1968). Three assumptions made by the traditional S—R theorists are
directly involved: a) habit strength is directly related to the
number of reinfOrcements, b) Speed of extinction is inversely related
to habit strength, and c) reversal learning is sinply a.natter of
extinguishing one habit and learning a new one in its place (Lovejoy,
1966; Fidura, 1966). Since these assumptions would not allow for the
prediction of ORE it would seem obvious that at least one of thenxis
in error.

Riley (1968) calls ORE one of the ”most explained phenomena"
in psychology today. However, prior to any analysis of the various
explanations and interpretations which have been given for this ef-
fect, a brief review of the literature should be given. Two experi—
ments by Lawrence (19u9, 1950), while not a direct attempt to study
this effect, are relevant in terms of later attenpts to explain ORE.
Lawrence (1950) was the first to recognize the importance of reversal
learning and the role that the "acquired distinctiveness of cues”
nught play (Mackintosh, 1965). The concept of the acquired distinc—
tiveness of cues which he used to explain the results of a reversal
study was formulated in an earlier study (Lawrence, 19u9). In this
first study he used one of three stimulus dinensions in a simultaneous
discrimination study with rats. The three dimensions were black vs.

white cues, rough vs. smooth floors, and large vs. small goal boxes.

The rats were then given transfer tests using two independent dimen-
sions in a successive discriminatim; the relevant dimension was
correlated with reinforcement and the irrelevant dimension was ran—
domly paired with the relevant dimension and not correlated with re—
inforcement. For the positive transfer group the relevant dimension
in the successive discrimination was the same stimulus dimension
learned in the previous simultaneous discrimination . The negative
transfer group had the dimension learned in the simultaneous dis-
crimination as the irrelevant dimension in the transfer test, and the
omtrol group had two entirely new sets of stimuli. The results show
that the positive transfer group did significantly better in the
transfer test than either the negative transfer group or the control .
The results also show that the negative transfer group and the con—
trol did not differ significantly.

Lawrence (1949) uses the concept of acquired distinctiveness
of cues to explain the positive transfer, _i_.e. , a mediating process
was established during the first discrimination problem which tended
to enhance the distinctiveness of the relevant dimension. In other
words, at the outset of the transfer test the previously learned
dimension, which for the positive transfer group was still the rele—
vant dimension, was more distinctive than any new dimension . "As a
result of the enhanced distinctiveness of the relevant cue in the
test situation, the new instrumental responses were associated with
the familiar cue more rapidly than with the unfamiliar one (Lawrence,
1949, p. 783)." The design of this experiment was such as to rule

out other explanations of the transfer in terms of a) carry-over of

the same instrunental response, b) learned ability to solve discrim-
inatim problems, 0) external inhibition, or d) acquired reward or
acquired drive (Fidura, 1966).

(he portion of Lawrence ' s mediating process hypothesis was
not confirmed by the data. No significant differences were found
between the negative transfer group and the control group . Lawrence
(19%) makes two assumptims: a) the stimulus which is conditioned
to the differential response has been modified by the mediating
process , and b) the relationship between this modified stimulus input
and the response is governed by the traditional S—R laws . If this
is the case , the negative transfer group should have done poorly be-
cause of the strong association value (the acquired distinctiveness)
of the irrelevant dimension should have interfered with learning.
waever, since the negative transfer group did not differ from the
control there is some indication that those situations influenced by
what Lawrence chooses to call the acquired distinctiveness of cues
may not follow the traditional S—R laws (Fidura, 1966) .

Lawrence (1950) trained one group of rats in a successive
discrimination to respond to a black-white dimension (relevant) and
to ignore the presence or absence of chains hanging in the goal—box
entrance (irrelevant dimension). He trained a second group of rats
(11 the converse of this discrimination. Following the original dis-
crinfinatim he trained all animals on a simultaneous discrimination
with both dimensions relevant and redundant. He then gave two tests,
an opposition test and a reversal test. In the opposition test the

positive attribute of each dimension was opposed so that the

direction of Choice would indicate which dimension was influencing
performance. Finally, all animals re-learned the siJmiltaneous dis-
crimination.with each dimension presented alone; one half of the sub-
jects learned the discriminations with the positive attribute of each
dimension reversed.

The results of the Opposition test show that the animals
respond generally in.terme of the dimension which had been relevant
in their original training. The results of the reversal test show
that the animals not only learned the non—reversed discrimination
faster on the dimension which for them had originally been the rele—
vant one, but also that they reversed faster on that dimension than
on the previously irrelevant dimension. Lawrence again interprets
these results "in terms of the concept of acquired distinctiveness
of cues which postulates that discrimination learning is essentially
a two-stage process, the first stage of which is a change in the
perceptual characteristics of the stimulus (Lawrence, 1950, p. 187)."
As had been mentioned, Reid (1953) found ORE in the results of the
reversal of a simple brightness discrimination. Since that time
others have found simdlar results with rats (Capaldi 8 Stevenson,
1957; Komaki, 1961; Mackintosh, 1962, 1963a, 1965b; North 8 Clayton,
1959; Pubols, 1956; and others). D'Amato and Jagoda (1961) found a
marginally significant ORE using a brightness reversal, but D'Amato
and his co—workers have not been able to replicate that result
(D'Amato 8 Schiff, 1965).

The results of a comparison between overtrained and non—
overtrained rats in a simple position reversal are not as conclusive.

In twenty-two studies cited by Mackintosh (1965) only slightly less

than 1/3 showed any facilitation of reversal following overtraining
(Capaldi, 1963; Ison 8 Birdh, 1961; Pubols, 1956; Theios 8 Blosser,
1965a; and others). Of the other 2/3 of the studies.most showed some
retardation of reversal following overtraining (Clayton, 1963a,
1963b; D'Amato 8 Jagoda, 1962; D'Amato 8 Schiff, 1961+; Galanter 8
Bush, 1959; Hill, 333;” 1962; Komaki, 1962; Mackintosh, 1965c; and
otherS) including four studies in which the retardation was signifi—
cant (Hill a Spear, 1963; Hill, 31;- gr, 1962; Krechevsky a Hozik,
1932; Mackintosh, 1965c). Lovejoy (1966) has summarized these re-
sults by saying that facilitation of reversal following overtraining
on a brightness discrimination (ORE) will usually, but not always be
found, and when the reversal is of a simple position discrimination
ORE will usually not be found.

The facilitation of reversal learning which is found follow-
ing overlearning is contrary to the traditional assumptions of the
S-R theorists that: a) habit strength is directly related to the
number of reinforcements, b) the speed of extinction is inversely re—
lated to the habit strength, and c) reversal learning is simply a
process of extinguishing an old habit and learning a new one in its
place. In attempting to explain the seemingly contradictory results
found in studies of ORE no attempts have been made to modify the
first assumption, a few attempts have been made to change the second
assumption, and even:more to change the third (Lovejoy, 1966; Fidura,
1966).

Attempts to explain the occurrence of ORE by modifying the

assumption that the speed of extinction is inversely related to habit

strength have been based on evidence that the resistance to extinc-
tion is, in fact, not a monotonic functim of the number of rein-
forced trials (Madison, 1961}; North 8 Stimmel, 1960; Siegel 8 Wagner,
1963; Wagner, 1963). This evidence, however, does not provide an
adequate explanaticn of why ORE is found in sore discriminations and
not in others, nor does it explain why within some experiments on ORE
there are same data which support the original assumption. Mackin-
tosh (19 65) states that all investigators who have actually published
data on the resistance to choice response extinction of over—trained
and ncn-overtrained subjects have found that overtraining does in—
crease the resistance to extinction (Kornaki, 1961; Mackintosh, 1962,
1963a, 1963b, 1965b; Reid, 1953). One of the measures which has been
used to determine the resistance to extinction is the length of the
initial error run after reversal (Hill, _e_t_ _a_l_. , 1962; Mackintosh,
1962). Same have argued (3g. , Sperling, 1965) that more rapid ex-
tincticn might follow the initial increased resistance to extinction
of the overtrained subjects. Mackintosh (1963b) showed that to reach
an extinction criterion of 50% accuracy over a block of ten trials

it took animals which had received overtraining longer.

In summary then, it would seem reasonable to divide the re-
versal of a discriminaticn problem into two phases: Phase 1 is the
extinction of the tendency to choose the original positive stimulus
(no: the negative stimulus) and Phase 2 is the acquisition of the
tendency to choose the positive stimulus . Attempts to explain ORE by
saying that overtraining decreases the time, in terms of number of

trials, required for the animal to complete the first phase and,

therefOre, increases the speed of reversal have been shown to be with-
out experimental support. In fact, the data would S€€Hlt0 show the
opposite to be true.

Of the attempts to modify the third assumption, namely that
reversal learning is merely a process of extinguishing an old habit
and learning a.new one in its place, the most general type of explan—
ation has been of the "learning set” type (Fidura, 1966; Lovejoy,
1966). Such explanations have been used by Reid (1953), Spence
(cited by Reid, 1953) and Harlow (1959). Reid's interpretation took
the form of a two-stage hypothesis. He hypothesized that at the same
time as the animal was learning a choice response it was also learn—
ing, at a slower rate, the habit of "response discrimination, i3e3,
learning to respond to a set of stimuli of which the specific stimu—
1us is a member (Reid, 1953, p. 107)." The response to the stimulus
set occurs at a greater distance from the reinforcement than does the
simple approach or avoidance of a particular stimulus value and,
therefore, the acquisition of the response of discrimination is slow—
er due to the greater delay of reinforcement. If after reversal, the
subject contines to make the appropriate response of discrimination
it will be more likely to select the new correct choice after extinc—
tion of the original choice response than will an animal which has
not learned the appropriate cue~discrimination reSponse (Riley, 1968).
Since overtraining allows more trials for the response of discrimina-
tion to become conditioned to the appropriate cues, over—training
should facilitate reversal. Note, however, that this explanation will

not account for the absence of ORE in position reversals.

Spence (cited by Reid , 1953) suggested that perhaps the
overtraining had the effect of equalizing position preferences , and
that the overtrained subjects would be less likely to revert to a
position preference during the early stages of reversal. Pubols
(1956) found ORE in an experiment similar to Reid's (1953), but was
able to reject Spence's hypothesis because the position preferences
for all animals had been equalized.

Riley (19 68) mentions two attempts to explain the negative
cases, that is those cases in which ORE is expected but does not
appear in the data. Paul (1965) suggests that the occurrence of ORE
depends to a large part on the consequences of an incorrect response.
He says that in the instances when ORE has not been found the result
of an incorrect response is an unbaited goal—box. When ORB has been
found, on the other hand, the consequence of an incorrect response
was a locked door in the majority of the cases. Paul (1965) assumes
that in order for ORE to occur the results of an error response must
be very dissimilar from those of a correct response. Why ORE should
be found in these experiments is not clear.

'Iheios and Blosser (1965b) have suggested still another ex-
planation for the negative cases . They have hypothesized that

"resistance to extinction is determined by the strength

of the habit, which is a function of the number of times

the particular actin question has been performed. This

resistance is , lunever, diminished by the disruptive
effects of non—reward. And this disruption in turn is

a fmction of the number of times the subject has been

mardedintlegoalboxandthemmtoftherenard

it has received on each of these occasions. Thus, if

the subject has received a reward many times and the

reward has been substantial , disruption will be great

and the resistance to extinction correspmdingly re-
duced. This analysis, then, really assumes the growth

10

of two habits, one of running down to the goal box

and a second which might be called expectation of

reward (Riley, 1968, p. 1%)."

Theios and Blosser use these two habits to explain ORE in the follow—
ing way: They assune that the instrumental approach habit grows more
quickly than the expectation of reward , but that if the incentive is
fairly substantial both will reach a similar asymptote. ORE is likely
to be found- under these conditions because with the increased numbers
of trials the difference between the habit strength and the reward
expectancy will be less than with a lesser number of trials . However,
studies using a small reward do not show ORE because the reward ex-
pectancy does not reach a high level and the disrupting effect of
errors in reversal is small.

To test their hypothesis Theios and his colleagues (l96u,
1965a, 1965b) have conducted several experiments which seem to support
his position. This hypothesis does not account for all of the results
obtained by Mackintosh and others .

Mackintosh (1965) in his review of the literature on selec—
tive attention in animal discrimination learning offers the following
rather intuitive argument for a mechanism of attention:

"Animals (particularly lower animals) have nervous

systems of limited size and therefore of limited

capacity for processing and storing information.

Thus they are confronted with the problem of selec-

ticn. At scne stage they must discard irrelevant

or redundant informaticn so as not to interfere

with the storage of important information . This

line of argument would seem to provide a general

ratimale for postulating, as Broadbent (1958) does,

the existence of filtering devices in the nervous

system; and if this approach is justified, it is

undoubtedly of first importance to general behavior

theory. To put it at its simplest, if animals do
not respond to all features of their stimulus input,

11

then a.sharp distinction must be drawn between
physical stimuli impinging on an animal in any
given situation and the effective stimulus which
controls the animalis behavior in that situation.
Failure to consider this distinction might lead
to explanations of behavior being offered which
are at best incomplete and at worst totally nus—
conceived... It may be granted, then, both that
there are plausible grounds for postulating a
central mechanism of attention, and that the
question of attention is an important one
(Mackintosh, 1965, p. 1210."

Mackintosh (1965) employs such a mechanisnlin his inter—
pretation of the results of Lawrence's (1950) reversal test, in
which animals learned the reversal of the discrimination on which
they had originally been trained faster than they learned the re-
versal of the originally irrelevant dimension. He explains these
results in terms of the following three assumptions:

"(a) that pretraining with A relevant and §_irrelevant

securely established attention to g, (b) that by the

end of simultaneous discrimination training on the

combined problem, subjects were attending to Q with

a very high probability and to §_with a very low

probability; (c) that, consequently, during reversal

to 5, choice responses would extinguish faster than

attention to the relevant cue while, during reversal

to E, attention to the relevant cue would extinguish

faster (Mackintosh, 1965, p. 136)."

In other words, using the two—stage model described by Lawrence
(1950) and by Mackintosh (1965) the following might be true: If

an animal is trained to approach §_and avoid Q} and is then reversed
on the discrimination problem, the first number of trials following
reversal will be primarily error trials. During this time both the
attention to the relevant dimension and the choice response may

extinguish; it is not, however, necessary to assume that the ex—

tinction will be completed simultaneously. If the attention to the

12

é-é‘ dinensicn extinguishes before the choice response then "rever-
sal learning can only be slow and laborious , since both attention
and choice response IIIJS‘t be built up _d_e_ novo (Mackintosh, 1965,

p. 135)." Hmever, if the choice response to g extinguishes before
the attention to the é—A’ dimension extinguishes, then the reversal
is nnre quickly learned.

"At first sight, it does not seem as if a two-stage

model mist necessarily predict one of these alter-

natives rather than the other; thus there is no need

for it to predict that reversal learning should be

relatively easy. It is clear, however, that one

definite prediction can be made; if attention ex-

tinguishes before choice responses , reversal will

be slower than if choice responses extinguish before

attention , and any procedure that increases the prob-

ability of attending without equally increasing the
response strength, will facilitate reversal (Mackin-

tosh, 1965, p. 1335)."

Fran Lawrence's (1950) experiment it was seen that original
training with one one relevant , 9g. , the black-white cue , and an-
other cue irrelevant not only gave faster discrimination when the
black-white dimension was presented alone, but it also gave faster
reversal on the black-white dinension presented alone . The original
training , then , was enough to prevent extinction of attention to the
relevant stimulus dinensim prior to the extinction of the choice
respmses. This explanation can becone a model for the interpre-
tation of the ORE data.

Lovejoy makes this attentional explanation of ORE slightly
nore specific:

"It is argued that in order to solve a discrimination

problem, animals must first learn to attend to the

relevant dinensicn (e.g. , brightness), and then they
nust learn to make the appropriate response once they

13

have attended to that dimension... Now surely a rat

has a limited information—storage capacity. On a

given trial he might remember'which side he chose,

whether he approached or avoided some special odor,

whether he went to black or white... But he will

probably not remember all these various attributes

of the choice he made. The development which follows

is based on the simple idea that a rat can attend

to only some of the possible dimensions on any given

trial, and that in some sense learning can occur only

on those trials when the subject is attending to the

relevant dimension (Lovejoy, 1966, p. 89)."
Lovejoy (1966) offers the same explanation for the faCilitation of
reversal as that given by Mackintosh (1962, 1965) and Sutherland
(1959), but he also explains why ORE should occur in some discrime
inations and not in others. The presence of OPE is not deternuned
by the comparison of the data obtained from a particular experiment
against some standard set of data, but rather by the comparison of
the reversal scores for the group which received overtraining with
the scores for the group which did not receive overtraining. Love—
joy hypothesizes that ORE should be found when the initial prob-
ability of attending to the relevant dfinension is very low; ORE
should not be found when the initial probability of attending to the
relevant dimension is high. If the relevant dimension used in a dis-
crimination problem is very salient, i323, the initial probability
that the subject will attend to it is very high (Fidura, 1966), then
during the original learning the probability of attending will ap-
proaCh its asymptotic level for all subjects and any overtraining
given some of the subjects will not significantly increase this
probability. Therefore, ORE should not be found, not because both
groups reverse slowly but rather because both groups continue to

attend to the relevant dimension and reverse quickly. If, on the

1H .

other hand, the relevant dimension is not very salient the over—
training given one group can significantly increase the probability
that those animals will continue to attend to the relevant dimension
throughout reversal thereby facilitating reversal. The subjects
whiCh do not receive overtraining are.more likely to stop attending
to the relevant dimension during reversal and will reverse slowly.
Because of the difference in reversal scores ORE will occur.

In studies with rats, discriminations involving a bright—
ness one when overtraining precedes reversal for some of the sub—
jects, ORE will usually be found. [ORE has also been found in
studies with octopuses (Mackintosh 8 Mackintosh, 1963) and in a
study with chicks (Mackintosh, 1965b)]. However, as we have seen,
ORE is usually not found in studies when position is the relevant
cue. This could be expected when the relative saliency of kines—
thetic cues and visual cues to a rat are compared.

Since the appearance of ORE is determined (according to
Lovejoy and Mackintosh) by the initial probability that the animal
will attend to the relevant dimension, it is readily apparent that
relatively mdnute changes in the experimental procedure of in the
apparatus could greatly change this probability and have a corres-
ponding effect on the results of the experiment. Without consider—
ing any other variables this fact alone could probably account for
the discrepancies found in the literature.

Riley (1968) in his review of the various interpretations
which have been offered for ORE lists three objections which might

be.made to the type of explanation offered above: a) while a

15

two-stage nediational theory might account for most of the available
data and might be intuitively attractive, it is more complex than
nonemediational theories; b) the relative determinants of the speed
with.which this mediational process develOps are not made explicit;
and c) what is the relationship between the learning of a reversal
and the learning of a non-reversal shift. In elaboration on the
first criticism, Riley says that a mediational process of selective
attention is intuitively attractive because it "seems to agree with
most observers' phenomenology (Riley, 1968, p. 151)." However,

this theory still requires the assumption that the stimulus which
controls the subjects' behavior is changed in some way during learn—
ing. This assumption is needed in addition to all others required
by non-mediational theories.

The second objection.made by Riley (1968) is an attempt to
compare the overtraining procedure to the repeated discrimdnation
problems used in "learning set” experiments. The question involves
the relative speed with which overtraining can facilitate reversal
as compared to the number of problems needed to establish a learning
set. As has been shown above the learning set theory would predict
ORE in every case and cannot, therefore, account for the data.

The third objection is based on the work of Kendler and
Kendler (1962) who have said that young children and rats (Kelleher,
1956) can shift dimension more easily than they can reverse the
values of a dimension. However, no overtraining was involved in
these studies and the results could be predicted by the Lovejoy—

Mackintosh hypothesis. Marsh (196”) has shown that ORE can be

16

found in studies with young children .

It would seem, then, that the only valid objection is that
which can be made to any two—stage theory. Since the lovejoy»
Mackintosh hypothesis appears to account adequately for a majority
of the data, perhaps the addition of the process of selective atten—

tion might be justified.

CHAPTER II

Experiment I

Lovejoy (1966) has hypothesized that the crucial deterndnant
of ORE is the initial probability that the subject will attend to the
relevant dimension. If this probability is initially very high and
approaches its asympototic level before the subjects meet the cri-
terion for the completion of original learning then overtraining will
show little facilitative effect on reversal performance. All sub—
jects, the overtrained and the criterion (non~overtreined) subjects,
will extinguish the original choice response and learn the new posi—
tive stimulus before their attention to the relevant dimension ex—
tinguishes (Mackintosh, 1965).

On the other hand, if the original probability of attending
to the relevant dinension is low, overtraining will facilitate re—
versal learning. Overtraining will increase the probability that
the subjects will continue to attend to the relevant dimension during
reversal and will facilitate the reversal learning as compared to the
performance of the criterion subjects. The criterion subjects, how-
ever, are more likely to stop attending to the relevant dimension in
favor of a.more salient dimension.

The above hypothesis is based primarily on the results of
studies done with rats. In order to select a dimension which is very

salient for a rat and also to select a less salient dinension, the

17

18

experbmenter is forced to use cues in two different sense modalities,
visual and kinesthetic. Though it is known that rats respond:more
readily to spatial cues than to visual cues, there is not enough in-
formation about the mechanism of selective attention in animals to
assess what interactions might exist between the sense nedality and
the attentional process. For that reason, any test of the Lovejoy—
Mackintosh hypothesis might best be nede using stirmli in a single
sense.modality.

The following test of the hypothesis used two visual stimr

uli in a study with birds. Japanese quail (Coturnix coturnix

 

japonica) are excellent subjects for this type of research for sev-
eral reasons: a) they adapt readily to laboratory conditions and a
pecking response can be rapidly conditioned (Fidura 8 Gray, 1966;
Fidura, 1966); b) they are primarily visually oriented animals and
do possess acute color vision; and c) data are available on the
relative probability that Japanese quail will attend to several
visual stimuli in an experimental situation identical to the one used
in this study (Fidura, 1966, 1969; Fidura 8 Grar, 1966).

Fidura 8 Gray (1966) assessed the acquisition of three cone
Amonly used simultaneous visual discriminations by Japanese quail.
The pecking response was learned prior to any discrimination training
so that the differences found in the number of trials needed to reach
acquisition criterion could be attributed to the relative initial
probability that the subjects would attend to the relevant dimension.
They found that a form discrimination was the most difficult, followed

by pattern, and that a color discrimination was the least difficult

19

to learn. The results are presented in Table l.

Fidura (1966 8 1969) did further studies using a similar
design with complex rather than simple stinwdi. The complex stimuli
were made up of the attributes of the three dimensions mentioned
above presented simultaneously on the two pecking keys; the attri—
butes of the dinensions varied randomly and independently between
the keys making 23 possible stimulus combinations.

For a given subject one dimension was relevant and correla—
ted with reinforcement and the other dimensions were irrelevant and,
therefore, randomly paired with reinforcement. The results were
similar to those found by Fidura 8 Gray (1966); the form discrimina—
tion was the most difficult, followed by pattern and then color.

The results are presented in Table 2.

The following two conclusions might be drawn from these
studies which would be pertinent to this experiment: a) in this
particular experimental chamber for these subjects the most salient
stimulus is color and the least salient is fornu b) the addition of
irrelevant dimensions to color does not significantly affect the
speed of learning, but the addition of irrelevant and more salient
dimensions to fornlgreatly decreases the speed of learning. Since
the initial probability of attending to the color dimension is high,
the acquisition of the discrimination is simply a matter of learning
the correct choice. However, since the probability of attending to
f0rm.is relatively low, the subject must first learn to attend to the
relevant dimension and then to make the correct choice response. The

greater the number of irrelevant dimensions the more difficult this

20

TABLE l.--Means and standard deviations of total and error trials 2
to the first criterion trial for each dimension. (N222)g

 

 

Relevant dimenSion

 

 

Color Pattern Form
Mean 27.5 123.5 360.6
sd 26.9 80.2 180.H
N 22 22 22

*
From Fidura 8 Gray, 1966

TABLE 2.--Means and standard deviations of total trials to criterion
. . . . 3:
on complex discriminations.

 

 

Relevant dinmnsion

 

 

Color Pattern Form
Mean 21.9 Hl2.H l361.u
sd 22.9 164.0 697.5
N 9 9 9

a
From Fidura, 1966

21

becomes and the greater the number of trials needed to reach cri—
terion .

This experiment is designed to test the Lovejoy—Mackintosh
hypothesis that: a) in simple discriminations ORE should not be
found when color is the relevant dinension, but probably will be
found.when the relevant dimension is form; b) in complex discrimin—
ations ORE again should not be found when color is the relevant
dimension, but should be found when form is relevant. There might
be some doubt of the validity of predicting ORE in the reversal of
a simple form discrimination. The hypothesis would suggest that the
subjects that received no overtraining will stop attending to the
relevant dimension during reversal and start attending to some more
salient dimension. However, the procedure used in this experdnent
will perhaps preclude the subjects' attending to any stimulus except
that which is presented on the pecking keys, and will, therefore,
destroy any ORE.

The overtraining procedure which will be used in this ex~
periflent differs from that used in previous studies. Reid (1953)
overtrained two groups of aninals; one group had 50 trials in ad—
dition to original learning and a second group had 150 overtraining
trials. Handler (1968) gave 150 trials past criterion in a black—
white discrimination with rats, and Mackintosh (1969) used 100 over—
training trials in a brightness discrimination in a Grice box. Shepp
and Turisi (1969), however, used 100% and 300% of the original learn—
ing as overtraining on a discrimination with retardates.

Lovejoy (1966) states that overtraining will facilitate

22

reversal if it increases the probability that the subject will con—
tinue to attend to the relevant dimension. The overtraining proce—
dure used in this study was designed specifically to fill that re—
quirement. The subjects learned the relevant dimension presented
alone to meet a specific criterion rather than a set number of trials

or a percentage of the original learning.

Method

Subjects
The §s were 29 male and fenele Japanese quail, 50-60 days

of age at the outset of the experiment. They were maintained at
75—80% of their ad_libitum body weight by daily compensatory feeding
to the 80% level. The §s were randomly selected from the colony
nadntained at the Psychological Laboratory at MiChigan State Univer—

sity.

Apparatus

The apparatus consisted of two identical commercial operant
chambers each with two pecking keys. The chambers were modified by
raising the floor 2 1/8 inches, and by mounting multiple stimulus
projectors behind each key. Reinforcement for a correct response
was presented in a food hOpper located centrally between the two
keys near the floor. The presentation of stimuli, reinforcement, and
the overall functioning of the apparatus was programmed and con-
trolled by a punched—paper tape—reading device which operated through
a system.of relays and timers. The Ss' responses and the total num-

ber of trials in one daily session were recorded on digital counters;

23

presentation of stimuli terminated automatically following the come

pletion of the criterion run of 15 consecutive correct trials.

Procedure

 

The key—peck response was well establisned by requiring the
§s to peck at an amber lighted key vs. a simultaneously present non-
lighted key for 200 food reinforeenents. The amber stimulus alter-
nated randomly between the two keys. Following this pre—training
each §_was randomly assigned to one of four conditions whidh varied
on stimulus complexity and amount of overtraining. The conditions
were as follows:

Amount of Overtraining_

 

 

Overtraining No overtraining
Sinrde N=6 N=6
Complexity
Complex N=6 N=6

Each §_1earned both a color and a form simultaneous discrimination;
the order of the two discrindnation problems was randomly determined
for each S, All Ss completed one discrimination problem, procedur-
ally consisting of a) original learning with the assigned stimulus
complexity and amount of overtraining, and b) reversal of the dis—
crimination at the same level of stimulus complexity, before learning
the second discrimination. The relevant dinensions and their re—

inforced and non-reinforced attributes were as follows:

ZN

 

Original
Dimension Learning Overtraining Reversal
+ - + — + —
Color Red* Green* Red Green Green Red
Form Triangle Circle Triangle Circle Circle Triangle

*Note: The red and green were equated for intensity and previous
researdh (Fidura, 1966) has shown that pretraining with an amber
light does not affect performance on a red vs. green color discrim-
ination.

Reinforced and non—reinforced attributes varied randomly between the
two keys. Complexity was achieved by adding the irrelevant dimension
of pattern (horizontal or vertical white lines) to the relevant dinen—
sion. The pattern varied randomly and independently of the relevart
dimension. Visually the complex stimuli with color relevant appeared
as three white lines, either horizontal or vertical, on a red or
green background; the complex stimuli with fornlrelevant appeared as
three white lines superimposed over a white circle or triangle on a
black background.

In all phases of the experinent the criterion of learning
was 15 consecutive correct trials. Each §_was run 1 hour each day
with the exception of those days on which a criterion was met, then
they were removed from the apparatus at the completion of criterion
and the next phase of the experiment was begun on the following day.

To prevent the Es from perseverating on one key and receiv—
ing 50% partial reinforcement, a type of correction procedure was
used. Following error responses the stimulus sequencing system.did
not advance and the stimulus patterns neintained the sane relative

positions on the next trial. Each trial began with the presentation

25

of the stimuli on the keys. When the S peeked one of the keys the
stimulus lights were turned off and the keys made inoperative for 8
sec.; if the response was correct, reinforcement was available for
the first 5 sec. and the stimulus sequencing system advanced. If
the response was an error the §_had an 8 sec. time—out with no re—
inforcement.

Overtraining consisted of a discrimination of the relevant
dimension presented alone, a procedure identical to the original
learning of the simple discrimination. Reversal learning differed
only in the relative positiveness of the attributes of the relevant

dimension.
Results

Table 3 presents the means and standard deviations of the
total trials to criterion, excluding the 15 criterion trials, for
all Se on the original discriminations. (All trials to criterion
are computed as trials to the first criterion trial in both experi-
ments.) These results are similar to those found by Fidura and
Gray (1966) and by Fidura (1966 8 1969) which are presented in
Tables 1 and 2. Table H presents the means and standard deviations
of the trials to criterion for those Ss which received overtraining
following either complex or simple discriminations in original
learning. These data show an increase in both mean trials and in
variability following simple and complex original discriminations

respectively.

26

TABLE 3.-~Means and standard deviations of total trials to
criterion in original learning.

 

 

Relevant dimension

 

 

Complex v' Simple
Form Color Form Color
MEan 1339.92 93.67 161.92 26.76
8d 828.93 ”3.06 132.76 33.89
N 12 l2 l2 12

TABLE H.——Means and standard deviations of total trials to
criterion in overlearning.

 

 

Relevant dimension

 

 

Form Color
Following Complex Simple Complex Simple
Mean 62.33 27.33 17.00 2.83
8d 58.00 15.67 28.68 6.99

N 6 6 6 6

27

Tables 5 and 6 present the means and standard deviations of
the trials to criterion in reversal learning as a function of dimen—
sion, complexity and amount of overtraining. Tables 7 and 8 present
a summary of the analysis of variance of the data presented in
Tables 5 and 6. No significant differences were feund as a function
of the amount of overtraining on the reversal of either color
(F = .281, df_ 1,20, p_> .20) or form (F = .092, df_l,20, E.) .20).
A significant effect was found as a function of complexity in the
reversal of the form dimension (F = 11.236, df_l,20, p_< .005), but
not in the reversal of the color dimension (F = 1.25, df 1,20,
p_> .20). And no significant interactions were found as a function
of amount of overtraining and complexity for either color (F = .958,

Q: 1,20 p_> .20) or form.(F = .00029, df 1,20, p_> .20).

Discussion

 

The results obtained do not directly support the Lovejoy-
Mackintosh hypothesis of ORE. No facilitation of reversal learning
was found following overtraining in the group which learned the com-
plex discrimination with form relevant. In fact the trend of the
data for this group were in the opposite direction, is?:3 there was
some small suggestion of retardation of reversal following over-
training. The same was true in the simple form reversal, though in
neither case did the level of retardation approach the level of
significance.

The results of the original learning for all subjects sup—

port the findings of Fidura and Gray (1966) and Fidura (1966 8 1969).

28

TABLE 5.-—Means and standard deviations of total trials to
criterion in the reversal of the discrimination
with color relevant.

 

 

 

Simple Complex
Following Oyertraining Criterion Overt raihfiig Criterion
Mean 136.33 126.50 139.17 171.67
sd 38.99 30.19 55.13 75.36
N 6 6 6 6

TABLE 6.-—-Means and standard deviations of total trials to
criterion in the reversal of the discrimination
with form relevant.

 

 

 

Simple Complex
Following Overtrain in g Criterion Overt raining Triterion
Mean 927.33 318.00 1569.67 1971.00
8d 308.57 122.79 1375.70 900.22

N 6 6 6 6

29

TABLE 7.--Smrlnary table of analysis of variance among mean total
trials to criterion as presented in Table 5.

 

 

' 'Y . ‘V

 

Source of variation d: §_s_ m__._s_
Amount of overtraining 1 782.038 782.038
Complexity 1 3980.038 3980.038
(Cells) 3 6929.126

Overtraining X

Complexity 1 2667.05 2667.05
Within-cells 20 55659.999 2782.725
Total 23 62583.625

TABLE 8.-—Summary table of analysis of variance among mean total
trials to criterion as presented in Table 6.

 

 

 

Source of variation df. s§_ §§_
Amount of overtraining 1 69896.00 69896.00
Complexity 1 7902832.66 7902832.66
(Cells) 3 7967899.39
Overtraining X

Complexity 1 170.68 170.68
Within-Cells 20 19066208.66 703310.933

Total 23 22039108.00

30

A similar conclusion to theirs might be drawn:

"...that is, since the motor response was the same for

all dimensions and was well established prior to the

beginning of training, differences in the acquisition

of the ... dimensions reflect differences in the first

or attentional stage of discrimination learning (Fidura,

1966, p. 29—32)."

The results obtained during the overtraining procedure also
indicate that the Se were attending to the color dimension after it
was paired with the pattern dimension in the complex original dis—
crimination with a much greater probability than they attended to
the form dimension presented either alone or in the complex stimulus.
However, these data do show that all Ss were attending to the rele—
vant dimensions with a greater probability following original learn—
ing than they were prior to original learning (see Tables 3 6 9).

The finding of a significant difference in the reversal
scores on the fermxdimension as a function of complexity could be
predicted by the Lovejowaackintosh hypothesis. The addition of a
more salient irrelevant stimulus to a stimulus which is not very
salient greatly decreases the probability that the §_will attend to
the relevant stimulus (see Tables 3 8 6).

Several explanations might be offered for the failure to
obtain the expected results, 139:3 ORE in the form discrimdnations,
but perhaps the two whiCh deserve primary consideration are those
relating to the overtraining procedure and to the criterion estab—
lished for original learning. The first hypothesis is that though
the overtraining procedure was designed to increase the probability

that the subject would attend to the relevant dimension without

significantly increasing the total response strength (Mackintosh,

. 31

1965), the number of trials required for the subject to learn the
overtraining discrimination compared to the number of trials re—
quired to learn original learning was relatively small. This is
particularly true in the form complex group. It might be suggested
then that the overtraining was not enough to produce facilitation
during reversal.

The second hypothesis provides a perhaps more reasonable
explanation of the failure to obtain ORE. The criterion which was
established for learning in this experiment was selected so that the
data could be directly compared.to the results obtained by Fidura
and Gray (1966) and by Fidura (1966 8 1969). It might be suggested,
however, that 15 consecutive correct trials is a difficult criterion
for the subjects to meet. Therefore, some amount of ”overlearning"
could take place in all groups before the subjects meet the criterion
for original learning. The comparatively few trials needed to meet
the overtraining criterion would tend to support this hypothesis.

The result would be that original learning facilitated reversal
learning and no ORE appeared in the data when the groups were com—

pared.

CHAPTER III

Experiment II

One possible explanation of the results of Experiment I is
that for all subjects the learning of the original discrimination
incorporated some amount of "overlearning" and, therefore, facili-
tation of reversal was evident in all subjects' data. Since ORE is
a relative measure between two groups of subjects, the effect would
not appear in the results if all groups reversed quickly (Lovejoy,
1966).

To test this hypothesis and to employ a slightly different
method of overtraining, the following experinent will use a criterion
of 8 consecutive correct trials in original learning. The criterion
for reversal learning will renein at 15 consecutive correct trials
to allow for comparisons between the appropriate groups in Experi—
ments I and 11. None of the subjects in Experinent II will receive
any overtraining.

A similar procedure has been used by Capaldi and Stevenson
(1967). They trained 3 groups of rats on a black-white discrimination
in an elevated T—maze. Group 1 was required to meet Criterion l
(7 out of 8 correct trials) prior to reversal; Group 2 was required
to.meet Criterion 2 (Criterion 1 plus 8 additional consecutive cor—
rect trials) before reversing; and Group 3 was required to have 35

additional correct trials past Criterion 2 prior to reversal.

32

If the subjects in Ebcperiment I did receive overtraining
during original learning then comparisons of the appropriate groups
from Experiment II with those in Experiment I should Show ORE. The
results of this experiment should also help to determine the validity
of the overtraining procedure used in Experiment I.

Method

 

Subjects
The _S_s were 8 male and female Japanese quail of the sane age

and stock as those used in Experiment I and maintained at 76-80% ad

libitum body weight by compensatory feeding.

Procedure

The procedure is identical to that used in Ebcperinent I with
the following two exceptions: a) following pretraining the _S_s were
randanly assigned to two conditions which varied only in stimulus
complexity (N = 14 in each group) and none of the gs received over-
training; b) the criterion for original learning was 8 consecutive
correct trials. All _S_s learned both a color and a form discrimina—
tion at the same level of complexity and the order of which was

randomly determined .
Results

The means and standard deviations for the total trials to
criterion in original learning are presented in Table 9. There is
an obvious reduction in the number of trials needed to reach criter-

icn for the §_s in Eztperiment II as compared to those in Ebcperiment I

3“

(See Tables 3 8 9). Tables 10 and ll present the.means and standard
deviations of the total trials to the reversal criterion of 15 con-
secutive correct trials for color and form respectively. Both the
scores for the non—overtrained _E3_s in Experinent I and the scores for
the groups in EXperiment II are included.

Multiple t_tests were performed as a function of the cri—
terion for original learning. None of the tests were significant and

I

a summary of the results appear in Tables 10 and ll.

Discussion

 

The results obtained in this experiment do not support the
Lovejoy—Mackintosh hypothesis of ORE. No facilitation of reversal
learning was found following a criterion of 15 in original learning
as was predicted. And this particular method of overlearning, though
it differs from that used in Experiment I, apparently does not in-
crease the probability of attending to the relevant dimension sig—
nificantly.

These results are, however, an indication that the learning
criterion of 15 consecutive correct trials used in Experirent I was
not so difficult as to incorporate some amount of overtraining for
all groups. It would seem that the failure to obtain the expected
results is probably due to the relatively small percentage of over-
training received by those groups in Experinent I as compared to the

subjects in Experiment II.

35

TABLE 9.-—Means and standard deviations of total trials to
criterion in original learning.

 

 

Relevant dinension

 

 

Complex Simple
Form Ir‘Color' ‘Torm Color
Mean 999.25 18.75 93.00 10.50
Sd 611.81 23.29 72.89 9.97
N 1+ u u u

TABLE lO.--Means and standard deviations of total trials to
criterion in reversal learning with color relevant.

 

 

 

Sinmfle Complex
Following Criterion 15 Criterion 8 Criterion 15 Criterion 8
MEan 126.50 131.25 171.67 132.00
8d 30.19 8.99 75.36 48.20
N 6 u 6 u
_t_ = .362 _‘E = 1.015

36

TABLE ll.--I“bans and standard deviations of total trials to
criterion in reversal learning with form relevant.

_v w—v

 

 

Simple ' ‘ ' Corrplex
Following Criterion 1‘3 Crit anion 8 Criterion 15 fiCriterion 8
Mean 318. 00 392 .00 1971 . 00 1999 . 75
Sd 122.79 212.68 900.22 993.11
N 6 u 6 u
E. : .209 t_ = .039

CHAPTER IV

Summary of Results

The results of this study do not support the Lovejoy-
Mackintosh hypothesis of ORE. On the other hand, neither do the

data disprove this hypothesis. The failure to find any significant

 

results in the statistical analysis of the comparison of reversal
scores would seem to indicate that this study was not an adequate
test of the hypothesis.

It could be that the hypothesis, as stated by Lovejoy and
Mackintosh is incorrect. However, inasmuch as this study did not
disprove the hypothesis an alternative explanation might be as
follows:

"Eimas (1967, Experinent II) and ErlebaCher (1963)

trained rats to discriminate between black and white

painted goal boxes or goal arms; D'Anato and Schiff

(1965) used different levels of diffuse illumination

as the discriminanda. The initial probability of

attending to the relevant stinmli.ney have been high

in all cases (Mackintosh, 1969, p. 2)."

The hypothesis depends upon the subjects which have not received
overtraining on a difficult discrimination problem attending to other
irrelevant dimensions during the first phase of reversal. As in the
studies cited above, the pretraining method used and the stimulus—
deprived environment of the apparatus could have forced all subjects

to continue to attend to the stimuli presented on the keys during

reversal. If this is true no ORE would be predicted.

37

LIST OF WCES

LIST OF REFERENCES

Broadbent, D. E. Perception and communication. Iondon: Pergamon
Press, 1958.

 

Capaldi, E. J. Overlearning reversal effect in a spatial discrimin-
ation task. Perceptual and Motor Skills, 1963, 16, 335—336.

 

Capaldi, E. J. 8 Stevenson, H. W. Response reversal following differ—
ent amounts of training. Journal of Comparative and Physio—
logical Psycholog, 1957, _5_O_, 1913—198.

 

 

Clayton, K. N. Overlearning and reversal of a spatial discrimina—
tion by rats. Perceptual and Mbtor Skills, 1963, 12)
83—85. (a)

 

Clayton, K. N. Reversal performance by rats following overlearning
with and without irrelevant stimuli. Journal of Experi—
mental Psychology, 1963, 66) 255—259. (b)

 

 

D'Amato, M. R. 8 Jagoda, H. Analysis of the role of overlearning in
discrimination reversal. Journal of Experimental Psychology,
1961, 6_1_, 95-50.

 

INAnato, M. R. 8 Jagoda, H. Overlearning and position reversal.
Journal of Experinental Psychology, 1962, 693 117—122.

 

D'Amato, M. R. 8 Schiff, D. Further studies of overlearning and
position reversal learning. Psychological Reports, 1969,
19, 380-382. '7

 

D'Amato, M. R. 6 Schiff, D. Overlearning and brightness discrimina-
tion reversal. Journal of Experimental Psychology, 1965,
6_9_, 375—381.

 

Eimas, P. D. Overtraining and reversal discriudnation learning in
rats. Psychological Records, 1967, 12) 239-298.

 

Erlebacher, A. Reversal learning in rats as a function of percentage
of reinforcement and degree of learning. Journal of
Experimental Psychology, 1963, 66) 89-90.

 

 

Fidura, F. G. The role of selective attention and the distinctive-
ness of cues in simple and complex discrindnation learning.
Unpublished doctoral dissertation, MiChigan State University,
1966.

38

39

Fidura, F. G. Selective attention and complex discrimination learn-
ing in the Japanese quail (Coturnix coturnix japonica).
Psychonomic Science, 1969, 15, 167—168.

 

Fidura, F. G. 6 Gray, J. A. Visual discrimination of color, pattern,
and form in the Japanese quail (Coturnix coturnix japonica).
Psychonomic Science, 1966, 5, 927—928.

 

Galanter, E. 6 Bush, R. R. Some Tamaze experiments. In R. R. Bush
and W. K. Estes (Eds.), Studies in mthenetical learning
theory. Stanford: University Press, 1959, 265—289.

 

Harlow, H. F. The formation of learning sets. Psychological Review,
1999, 56, 51—65.

 

Harlow, H. F. Learning set and error factor theory, in S. Koch
(Ed.), Psychology: A study of science. Vol. 2. New Yerk:
McGraw—Hill, 1959, 992—537.

 

Hill, W. F. 6 Spear, N. E. A replication of overlearning and rever—
sal in a T—maze. Journal of Experimental Psychology, 1963,
65, 317.

 

Hill, W. F., Spear, N. E. 6 Clayton, K. N. T—maze reversal after
several different overtraining procedures. Journal of
Experinental Psychology, 1962, 63, 533—590.

 

 

Ison, J. R. 6 Birch, D. T—naze reversal following differential and
box placenent. Journal of Experimental Psychology, 1961,
62, 200—202.

 

Kelleher, R. T. Discrindnation learning as a function of reversal
and non—reversal shifts. Journal of Experimental Psye
chology, 1956, 51, 379—389.

 

Kendler, H. H. 6 Kendler, T. S. Vertical and horizontal processes
in problemrsolving. Psychological Review, 1962, 69, 1—16.

 

Komaki, J. Reversal retardation through introducing forced non—
rewarding trials in overtraining by white rats. Annual of
Animal Psychology, 1962, 12, 1-10.

 

Krechevsky, I. 6 Honzik, C. H. Fixation in the rat. University of
California Publications in Psychology, 1932, 6, 13—26.

 

 

Lawrence, D. H. .Acquired distinctiveness of cues: 1. Transfer
between discriminations on the basis of familiarity with
the stimulus. Journal of Experimental Psychology, 1999,
39, 770n789. ‘i'I ‘ V I 7 W 77" ‘7 '

 

90

Lawrence, D. H. Acquired distinctiveness of cues: II. Selective
association in a constant stimulus situation. ‘ Jomnal of
Byerinental Psychology, 1950,“ ‘36, 175—188. ‘ 7 7 V

 

 

Lovejoy, E. Analysis of the overlearning reversal effect . Psycho-
logical Review, 1966, 16, 87-103.

 

Mackintosh, N. J. The effects of overtraining on a reversal and a
nonreversal shift . Journal of Comparat ive and Phygsiological
PSYChOlOEY, 1962, EE) 553'559.

 

Mackintosh, N. J. The effect of irrelevant cues on reversal learning
in the rat. British Journal of Psychology, 1963, 59, 127-
139. (a) _—

Mackintosh, N. J. Extinction of a discrimination habit as a function

of overtraining. Journal of Comparative and Physiological
Psychology, 1963, 66, 892—897.— (D)

 

Mackintosh, N. J. Selective attention in animal discrimination
learning. Psychological Bulletin, 1965, 69, 129-150. (a)

 

Mackintosh, N. J. Overtraining, extinction and reversal in rats and
chicks. Journal of Comparative and Physiological Psycholo-
fl, 1965, 69, 31-36. (b) 7

 

Mackintosh, N. J. Overtraining, transfer to proprioceptive control,
and position reversal. Quarterly Journal of Experimental
Psychology, 1965, _1_7, 26—36. (c)

 

 

Mackintosh, N. J. Further analysis of the overtraining reversal
effect. Journal of Comparative and Physiological Psy—
cholog, 1969, 62, (2, Pt. 2).

 

Mackintosh, N. J. 6 Mackintosh, J. Reversal learning in Octopus
vulgaris Lamarck with and without irrelevant cues.
giarterly Journal of Eagerinental P_sycholog, 1963, 16,
236-292.

 

 

Madison, H. L. Experimental extinction as a function on number of
reinforcements. Psychological Reports, 1969, 11:, 697—650.

 

Mandler, J. M. The effect of overtraining on the use of positive
and negative stimuli in reversal and transfer. Journal
of Comparative and Physiological Psychology, 1968, 66,
110—1175. i _—

 

Marsh, G. Effect of overtraining on reversal and non—reversal
shifts in nursery school children. Child Development,
1969, _3_5_, 1367—1372. 777 7

 

91

McCulloch, T. L. 6 Pratt, J. G. A study of the pre-solution period
in weight discrimination by white rats. JOurnal 'of “Com—
pariatgive and Phys1o1ogic'a1 Psychology, 1939', 16, 271—290.

 

 

North, A. J .1 6 Clayton, K. N. Irrelevant stimuli and degree of
learning in discrimination learning and reversal.
Psychological‘ Reporis, 1959, 6, 905—908.

 

North, A. J. 6 Stimmel, D. T. Extinction of an instrumental response
following a large number of reinforcements. ' Psydiological
Remrts, 1960, _6_, 227—239. i

Paul, C. Effects of overlearning on a single habit reversal in
rats. Psychological Bulletin, 1965, 6_3_, 65—72.

 

Pubols, B. H. , Jr. The facilitation of visual and spatial discrim—
ination reversal by overlearning. Journal of Comparative
Physiological Psychology, 1956, 96, 293—298.

 

 

Reid , I... S . The development of noncontinuity behavior through con-
tinuity learning. Journal of Emerimental Psychology,
1953, 36, 107—112.

 

Riley, D. A. Discrimination learning. ‘Boston: Allyn 6 Bacon, Inc. ,
1968.

 

Siegel, S. 6 Wagner, A. R. Extended acquisition training and re-
sistance to extinction. Journal of Ewerimental Psy—
chology, 1963, 66, 308—310.

 

Shepp, B. E. 6 Turrisi, F. D. Effects of overtraining on the
acquisition of intradimensional and extradimens ional
shifts. Journal of Experimental Psychology, 1969, 61,
96-51.

 

Spence, K. W. An experimental test of the continuity and non-
continuity theories of discrimination learning. Journal
of Experimental Psychology, 1995, _3__5_, 253-266.

 

Sperling, S. E. Reversal learning and resistance to extinction:
A review of the rat literature. Psychological Bulletin,
1965, 6_3_, 281-297.

 

Sutherland, N. S. Stimulus analyzing mechanisms. In Proceedings
of a symposium on the mechanization of thougitw processes.
Vol. 2. london: Tlerj‘lajestyis Stationary Office ,7 1959,
575-609.

 

 

Theios, J. 6 Blosser, D. The overlearning reversal effect and mag—
nitude of reward. Journal of Comparative and Physiological
Psychology, 1965, 69‘, 252—257? (a) v '

 

 

 

92

Theios , J. 6 Blosser, D. An incentive model for the overlearning
reversal effect. "Ps'y‘dionomic SciencE, 1966, .2_, 37—38. (b)

‘hhfifi.w fiw‘w ‘

 

Theios, J. 6 Brelsford, J. Overlearning—extinction effect as an
incentive phenomenon. ' “Journal 'of Experimental Psychology ,
1969, 61, 9631-967. "‘ ‘

 

Wagner, A. R. Overtraining and frustration; PsychOlogiCal Reports,
1963, 13, 717:718. - 7 H

 

 

JUL 21

.-. .4»
C L'_ 3

(.1)

Ml HIGAN STATE UNIVERSITY LIBRARIES

1

293 03056 4425

C

 

3