EFFECTS OF pELAY OF INFORMATION FEEDBACK ; ,j, _ .,
AND LENGTH OF POSTFEEDBACK INTERVAL 9N LINEAR ' f
PROGRAMED LEARNING -- ' ' " '

1965 --------- » NEG i} E: I75ff5é- E'7: 51

 

 

mama

LIBRARY

Michigan State
University

 

This is to certify that the

thesis entitled

EFFECTS OF DELAY OF INFORMATION FEEDBACK AND LENGTH OF
POSTFEEDBACK INTERVAL ON LINEAR PROGRAMMED LEARNING
I5

L .

-,1

presented 1513

Frederick J. Boersma

has been accepted towards fulfillment
of the requirements for

Ph.D. degree in Education

Major professor

Date December 14, 1964

 

0-169

 

 

 

 

 

 

 

 

 

 

 

 

 

ABSTRACT

EFFECTS OF DELAY OF INFORMATION FEEDBACK
AND LENGTH OF POSTFEEDBACK INTERVAL
ON LINEAR PROGRAMED LEARNING

by Frederick J. Boersma

The controversial findings on delay learning were
attributed to the failure of experimenters to control for the
confounding effects of the intertrial interval with delay of
information feedback (DIF), and post-information feedback
delay (PIF). The study attempted to determine the indepen-
dent effects of these delay variables in a complex learning
task in which subjects (§s) learned a series of symbolic
logic rules by programed instruction. Fifty-six §s served
individually in a factorial design which combined two de-
lays of IF (0 and 8 sec.), two postalF intervals (0 and 8
sec.) and a sex factor.

The data did not support the prediction that human
learning would be facilitated by increasing the post-IF
interval, nor did it yield a significant delay or IF effect.

But it did show that a simple explanation in terms or the

3

independent effects of delay of IF or postuIF is inadequate.

In addition, the analyses of answer latencies revealed that

' “ ts
SS Spent more time examining correct answers when treatmen

. .‘ I 0 - S
were associated with delayed rather than immediate IF. A

 

 

 

Frederick J. Boersma

predicted, all delay effects for response latencies were
insignificant. A competing response interpretation of delay,
similar to that developed by Spence and Denny, was formulated

with respect to error scores, answer latencies and the de—

lay variables.

 

 

 

 

 

EFFECTS OF DELAY OF INFORMATION FEEDBACK
AND LENGTH OF POSTFEEDBACK INTERVAL
ON LINEAR PROGRAMED LEARNING

By

 

Frederick J. Boersma

 

 

A THESIS

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

DOCTOR OF PHILOSOPHY
College of Education

1965

 

 

 

ACKNOWLEDGEMENTS

To the members of my Doctoral committee, I owe a
special word of thanks. To Professor Bernard R. Corman, for
his valuable guidance in the development of my Doctoral pro-
gram, and assistance in the direction of this dissertation,
I am especially indebted. I also wish to express my appre-
ciation to Dr. Walter R. Stellwagen for his assistance in
the complexities of the experimental design. A further word
of thanks is due to Dr. M.R. Denny and Dr. Jean LePere for
their constructive criticisms of this study, and to
Cornelius VanderVeen for his aid in experimentation.

Finally, to my wife, Linda, for her consistant sup-

port throughout my Doctoral program, I am grateful.

ii

 

 

TABLE OF CONTENTS

ACKNOWLEDGEMENTS O O O O O O I O O O O O O O O 0

LIST OF TABLES O O 0 O O O O O O O O O O O O O 0

LIST OF FIGURES . . . . . . . . . . . . . . . .

THEORETICAL ORIENTATION AND REVIEW OF LITERATURE

METHODOLOGICAL CONSIDERATIONS . . . . . . . . .
Subjects . . . . . . . . . . . . . . . .
Procedure and Apparatus . . . . . . . .
Design . . . . . . . . . . . . . .

Design Note . . . . .
Treatments . . . . .
Dependent Variables .
Reliability . . . . . .
Hypotheses . . . . . . . . . . . . . . .

O 0 O 9
O
C
0

RESULTS . . . . . . . . o o o o o e o o o o o 0

Analyses of Error Scores . . . . . . . .
Analyses of Latency Scores . . . . .
Exposure Time Data . . . . . . . . . .
Intertrial Interval Data . . . . . . .
Questionnaire Data . . . . . . . . . . .

DISCUSSION AND CONCLUSIONS . . . . . . . . . . .

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . .

APPENDICES , . , , , , . . . . . . . . . .

iii

Page
ii

iv

53

58

 

 

 

Table

10

11

LIST OF TABLES

Analysis of Variance on Program Errors .

Analysis of Variance on Criterion Test
Errors 0 O O D O O U 0 O O I O O O O O

Summarized Mean and Variance Data for
Program Error Scores . . . . . . . . .

Summarized Mean and Variance Data for
Criterion Test Error Scores . . . . .

Analysis of Variance on Average Response
Latency Scores . . . . . . . . . . . .

Analysis of Variance on Average Answer
Latency Scores . . . . . . . . . . . .

Summarized Mean and Variance Data for
Average Response Latency Scores . . .

Summarized Mean and Variance Data for
Average Answer Latency Scores . . . .

Summarized Mean Exposure Time Data for
Program Material . . . . . . . . . . .

Summarized Data on Intertrial Interval
Exposure Time . . . . . . . . . . . .

Percentage Breakdown of Questionnaire
Responses as a Function of Treatments

iv

Page
23

24

26

27

30

31

32

33

36

38

39

 

 

 

 

 

Figure

LIST OF FIGURES

Temporal Intervals Common to the Present
S tUdy . Q C O O O O O O O O O O I O O

A Geometric Interpretation of the Delay
of IF by Post-IF Interaction for
Program and Criterion Test Error
Scores . . . . . . . . . . . . . . .

Page

18

25

 

 

 

 

 

 

Appendix
A

HED’UWUOW

«O’UOZKHWLI

LIST

OF

APPENDICES

I I O 0 O Q
I I I l I I
I I I I I
I I I I I
I I I I
I I I I I
I I I I I
I I I

I I I I
I I I I I
I I

I I
I I I

I I

I I I

I
I I I I I

Page

 

 

 

 

THEORETICAL ORIENTATION AND REVIEW OF LITERATURE

Delay of reinforcement as an independent variable in
the experimental study of learning dates at least to Watson's
1917 experiment on delay of reward in a digging response
(Renner, 1964). Watson found that a 30 second delay had no

1 effect on acquisition and questioned the then widely held
supposition that receiving a reward "stamped in" the most
recent of a learner's responses. Watson's experiment marked
the beginning of a continuing controversy about how empiri-
cally established relationships between the immediacy of
reinforcement and acquisition are to be interpreted.

Contemporary theorists have mainly favored and exu
tended either Hull's (1943) "law of effect" interpretation
or Guthrie's (1952) "contiguity" explanation, although a few
theorists have advocated two—factor theories in an attempt
to bridge the gap between these conflicting positions
(Skinner, 1938; Mowrer, 1960; Spence 1956, 1960). In part,
theoretical distinctions between "contiguity" and "effect"
interpretations of reward have been narrowed by the realiza—
tion that differences often represent linguistic preference.
For instance, even though Hull spoke of "drive—reduction"
and Guthrie, "stimulus change" their respective students

(e.g., Spence, 1956; and Sheffield, 1954) arrived at strikingly

 

 

 

 

 

 

 

similar interpretations of the role of reinforcement. Both
Spence and Sheffield considered acquisition to be the result
of the contiguous occurence of stimulus and response and
held to Tolman's (1955) distinction between learning and
performance.

In any case, the experimental evidence has not con—
vincingly established the supremacy of any one monistic or
two—factor theory, though there is little doubt that rein-
forcement does play an important role in controlling both
animal and human performance. Research with animals has
established that delay of reinforcement retards acquisition,
that delay combines cumulatively with other variables within
the immediate situation, and that delay leads to an increase
in resistance to extinction when S5 can mediate the delay
interval.

The explanation of these established relationships
is in dispute. Spence (1956), argued that the constant den
lay of reinforcement situation, in which §s remain without
any necessity or opportunity for making the response to be
learned, is a much more complex phenomenon than previously
realized. A theoretical examination of classical and instruw
mental conditioning with respect to such delay gives con-
siderable support to his position.

In classical and instrumental conditioning, for both
animal and human gs, delay of reinforcement appears to be

either a form of partial reinforcement in which the proportion

 

 

 

 

 

 

of reinforcements is inversely related to length of the de-
lay interval, or a form of experimental extinction in which
the immediate situation remains unchanged and reinforcement
fails to occur. This hypothetical interpretation of delay
implies that the same frustration-aroused competing response
tendencies which occur during extinction and lead to a decre—
ment in conditioned response strength, will also occur dur-
ing delay of reinforcement.

Spence (1956, p. 154) stated that "occuring as these
responses do to essentially the same components as does the
to be learned instrumental response they likewise become
conditioned to them," and that once such competing responses
are established, that they "will have the effect of increas—
ing the time it will take the appropriate response (to
occur)."

Rieber (1961), and a series of studies reported by
Spence (1956), provided support for the supposition that
maximum performance is dependent upon minimizing competing
behavior (Carlton, 1954; Harker, 1950; and Shilling, 1951).
Ramond's 1954 study, also reported by Spence (1956), hypoth—
esized that competing behavior accumulates when S5 are not
able to minimize conflicting behavior. A similar hypothesis
has recently been proposed by Renner (1963).

Spence (1960), surmised, contrary to Hull's work»
fatigue concept of inhibition, that failure of occurrence of

a reinforcer following a response results in a type of

 

 

 

 

 

 

 

 

frustration inhibition. He stated that the strength of this
frustration varies directly with the strength of the non-
reinforced response tendency, and that this effect occurs
since §s have learned to expect or anticipate reward which
no longer occurs. This explanation led Spence to refer to
his theory as a type of S—R expentancy theory.

Spence's interference-frustration notions of inhibi-
tion are similar to those proposed by Denny and Adelman
(1955). Denny and Adelman hypothesized that frustration
(non-reward) constitutes an eliciting state of affairs where
cues associated with non-reward acquire the property to
elicit avoidance responses, and that these avoidance responses
compete with the to—be—learned responses. Similarly, Amsel
(1958) stated that introduction of delay into a previously
reinforced situation results in an emotional reaponse of
frustration which promotes conditioned avoidances and a
decrement in performance. Additional support for this
frustration hypothesis has been reported by Adelman and
Maatsch (1955). These Es found that rats allowed to escape
from an empty goal box during extinction showed less loss of
original habit strength than did those who were retained in
the box. Adelman and Maatsch's data indicated that escape
from frustrating conditions of extinction is conducive to
the elicitation of fewer competing responses, and conse—

quently, greater resistance to extinction.

 

 

 

A somewhat more detailed analysis of the processes
of acquisition and extinction in terms of competing responses
has been published by Pubols (1958). Pubols hypothesized
that resistance to extinction varies inversely with number
of competing responses which have been extinguished during
acquisition, and that readaptation to the original situation
is a necessary condition for extinction. His position sug=
gested that the greater the opportunity for competing res—
ponses to occur, the slower the rate of acquisition and the
greater the resistance to extinction.

Although there are many theoretical interpretations
of why constant delay appears to operate similarly to partial
reinforcement during extinction, in experiments with animals
this effect is probably best attributed to the acquisition
of situational and/or reSponse produced cues formed in
original learning which become reactivated during extinction
and make it difficult for S3 to discriminate between condi—
tions of acquisition and extinction.

Researcher's have found it necessary to introduce
additional variables and/or constructs to explain the effects
of delay of reinforcement in human, as distinct from animal
learning. Much of the research in human learning has dealt
with the influence of delay of knowledge of results (KR) in
simple psychomotor tasks. Greenspoon and Foreman (1956)
found that increasing length of delay interval before in—

forming §s about their performance reduced rate of acquisition

 

 

in learning to draw a three inch line. A replication of the
Greenspoon and Foreman study, using tighter controls, by
Bilodeau and Ryan (1960), reported that delay of KR does not
retard psychomotor performance, but a similar study by Ryan
and Bilodeau (1962) proved inconclusive. The majority of re—
ports, however, have indicated that delay does not retard at-
tainment of a performance standard (Bilodeau and Bilodeau,

1958; Denny et al., 1960; McGuigan, 1959; Noble and Alcock,

 

1958; and Saltzman, Kanfer and Greenspoon, 1955). The delay
of KR in these studies typically has been up to 30 seconds,
although Bilodeau and Bilodeau (1958) used a seven day delay
interval.

Even more controversial findings have been reported
when concept formation and verbal learning tasks were empha-

Sized. For example, though Bourne (1957) did find that delay

 

of information feedback (IF) retarded concept attainment in
his 1957 study, his replication using tighter controls, failed
to produce a similar effect (Bourne and Bunderson, 1963).
In verbal learning, Saltzman (1951) found that delay inhib-
ited acquisition, whereas Brackbill and Kappy (1962) found
that it had no effect on learning. Generally speaking, re—
vieWS of the literature on delay in more complex forms of
human learning have reported that performance decreases as
delay increases (Ammons, 1956; Renner, 1964; and Wolfe, 1951).
More consistent relationships have been reported in
the programed learning literature where the majority of the
studies support Skinner's (1954) supposition that immediate

reinforcement is a necessary condition for effective learning.

 

 

 

 

 

For example, Angell (1949) using a Presseyatype punchboard,
and Meyer (1960) using a linear program, found that students
receiving immediate KR had significantly higher final exam—
ination scores than did those who were given results at the
next class meeting. Coulson and Silberman (1960), and Lums-
daine (1960) have also attributed the effectiveness of auto—
instructional material to the immediacy of KR as a reinforcer.
But, other studies have reported that delay of KR has no
effect on learning (Evans, 1960; Feldhusen, Ramharter and
Birt, 1962; Hough and Revsin, 1963; McDonald and Allen, 1962;
and Moore and Smith, 1961). And while a number of other
researchers have reported significant differences in favor

of immediate KR, they were mainly concerned with testing
different methods of providing feedback (e.g., Krumboltz and
Bonawitz, 1962; and Kanner and Sulzer, 1961).

Thus, though the study of experimental delay goes
back nearly 50 years, and though this research has demon—
strated the retardation effects of delay in lower organisms,
it has failed to generate such consistent findings with
humans.

In an attempt to resolve some of these contradictions,
Bilodeau and Bilodeau (1958) ran a series of five psychomotor
experiments to see what effect, if any, the intertrial in—
terval (1T1) had on performance. Their rationale was that
Es had failed to consider the confounding effects of 1T1 with

delay or KR in the majority of the psychomotor studies. Four

 

 

 

 

of these experiments indicated that the lTI was a critical
variable in this type of learning task, while the fifth ex-
periment suggested that post-KR delay may facilitate per-
formance. These studies showed that delay of KR, up to a
week did not effect performance in drawing a three inch line,
but that performance varied inversely with length of the

ITI. Denny, et al. (1960) attempted to control for possible

 

carry—over effects associated with the familiar three inch
line used by Bilodeau and Bilodeau by having §s draw 60
"glubs" (also a three inch line). They found, as did Bilo—
deau and Bilodeau, that the period of time between responses
was the crucial variable, i.e., the 1T1 (better learning
with shorter intervals), and that delay of KR did not effect
performance. Moreover, they surmised that "...previous ex-
perience in conjunction with knowledge of the response to
be learned will diminish the importance of the intertrial
interval as a variable" for very simple motor tasks (Denny,

et al., 1960, p. 327), Data reported by Becker, Mussina

 

and Persons (1963) support this hypothesis in that they
found the only significant variable operating in the drawing
of a familiar three inch line was that of KR itself. Al-
though their findings supported Bilodeau and Bilodeau's (1958)
position that delay of KR is not crucial in simple psychomotor
tasks, their data did not yield a similar ITI effect.

Another recent attempt to explain the differential

effects of IF and lTl was made by Bourne and Bunderson (1963).

 

 

 

 

 

To control for possible confounding effects of the lTl with
delay of IF and post-IF delay, they developed an experimental
design which permitted them to vary these variables ortho—
gonally, but maintain an average ITI between treatments.
Bourne and Bunderson used this design to replicate Bourne's
1957 concept formation experiment. In this replication they
found that delay of IF did not retard concept attainment,
that increases in the post-IF interval lead to significant
decreases in error scores, and that the post-IF effect became
more prevelant as concept complexity increased.

Bourne and Bunderson (1963) surmised that the dis—
crepancy between Bourne's 1957 and their 1963 study, and the
majority of the research reported on human delay learning,
is due to the fact that Es have failed to control for pos—
sible confounding effects of ITI with the post—feedback
interval, and to consider such methodological factors as
task complexity.

Thus, it appears that the lTI ceases to function as
a crucial variable for very simple tasks when §s have had
previous experience with these tasks (Becker, Mussina and
Persons, 1963), and for complex tasks (five dimension con—
Cepts) when §s have not had previous experience with such
tasks (Bourne and Bunderson, 1963). Results from studies
employing wide ranges of task complexity which build in

experience, as in programed instruction, are needed in order

 

 

 

to determine optimal acquisition intervals for such tasks,
and to demonstrate the effects of complexity on performance.

Brackbill and Kappy (1962) hypothesized that §s used
mediational cues acquired during acquisition to bridge the
time gap between response and KR in their serial learning
experiment. They suggested that the differential delay
effects reported in human learning literature may be due to
the fact that experimental tasks were such that §s could
mediate during the delay interval in some cases, but not in
others. In an attempt to provide support for this media—
tional hypothesis, Brackbill (1964) replicated the 1962 study
by initiating an interpolated task during the delay interval.
She found that in this instance delay retarded performance
and concluded that §s were unable to mediate the delay due
to interfering (competing) behavior elicited by the inter—
polated task. It should be noted, however, that even though
Brackbill and Kappy (1962), and Brackbill (1964), did con-
trol for the 1T1, they did not consider the effects of the
post-delay on performance, as did Bourne and Bunderson (1963),
and that consequently, their data is confounded and mis—
leading.

But, Brackbill and Kappy's mediation hypothesis is
in agreement with Lorge and Thorndike“s (1935) earlier ques—
tioning of the premise that delay of IF or KR retards all
forms of performance. Lorge and Thorndike suggested that

there are delay situations (unfilled) in which §s can

 

 

 

11

reinstate (symbolically) some form of response after-effect,
and reach criterion just as quickly as if they had received
immediate reinforcement. Similar mediational and/or response
produced delay interpretations have been developed by other
psychologists (e.g., Bilodeau and Bilodeau, 1958; Brown,
1949; and Erickson and Lipsett, 1960).

These contradictory findings within the delayed
learning literature might also be attributed to innate dif-
ferences between animal and human learning with respect to
the reinforcement variable. Such a differential explanation
of reinforcement would hold that it is appropriate to speak
of "reward" in animal learning, but that the constructs of
"knowledge of results" and "information feedback" are more
exact in human learning. This distinction, however, appears
to be hypothetical rather than demonstrable, since the moti—
vational characteristics of the §s used in the experiment
usually determine whether or not reward (incentive) is pro—
vided, and not differences in the reinforcement variable
per se. For example, if intrinsic motivation is insufficient,
as with animals and very young children, reward is usually
provided; whereas, if it is adequate, as with normal adults
and older children, only knowledge of results (KR) or in—
formation feedback (IF) is given. Thus, it appears that
reinforcement, KR and IF are operating similarly in both
animal and human learning. Furthermore, these terms have

been used interchangeably by a number of prominent

 

 

 

 

12

psychologists (e.g., Bourne and Bunderson, 1963; Brackbill
and Kappy, 1962; and Skinner, 1954).

Certainly unwarranted in human learning is the com-
plete acceptance, without qualification, of delay of rein-
forcement generalizations which state that a few seconds
delay between response and reinforcement may mean the dif-
ference between maximal learning and no learning whatever.
There are, as we have shown, situations in which gs appear
to mediate the delay interval and minimize its decremental
effect. The frequent conflicting findings reported for the
effects of delay in human learning may be due to the fact
that $5 have generally failed to control for confounding
effects of ITI with delay of IF and post-IF. Even in the

Bourne and Bunderson (1963) study on concept formation,

 

which did control for confounding effects of delay with 1T1,
the data were not representative of the more complex forms
of school learning.

The experiment to be reported in this paper has
attempted to extend existing knowledge on human delay learn—
ing by examining the independent effects of delay of IF,
and post—IF delay, on performance of a programed learning
task in symbolic logic. This study differed from the pre—
viously reported learning research on delay, in that the
differential complexity of material within the program neces-
sitated the use of §, rather than the E controlled, feedback

intervals. This initiated a new variable into Bourne and

 

 

13

Bunderson's (1963) design on which the present study was
based. But, this change in experimental procedure was im-
perative if §s were to receive adequate IF. The signifi-
cance of this study lies in its close resemblance to actual
school learning, and in its use of a variable, not pre-
viously examined as a function of delay, i.e., the S con-
trolled feedback interval.

It was predicted that the post—IF interval would be
the crucial variable in this complex learning task, and that
the effect of delayed IF would be negligible. It was further
surmised that the § controlled feedback interval would not
correlate with error scores, and that the data would resemble
that obtained by Bourne and Bunderson (1963), since for all
practical purposes, the E had controlled for confounding

effects of ITI with the delay variables.

 

 

METHODOLOGICAL CONSIDERATIONS

Subjects

Fifty-six volunteer students, 28 men and 28 women,

enrolled in the College of Education at Michigan State Uni-
versity were used as §s in this experiment. Their median
age was 20 years and their cumulative mean grade point
average was 2.66. These §s were randomly assigned to the
four experimental conditions (14 per experimental treatment)
to be described below, with sexes equally represented in

each treatment.

Procedure and Apparatus

The Linear program in symbolic logic developed by
Evans, Glaser and Homme (1962) was modified for use in this
experiment. The final form of the modified program con=
sisted of seven rules to be learned in 78 frames (See Appen-
dix A). Items within the program progressed from simple to
complex. Each frame was programed according to the Ruleg
System developed by Evans, Homme and Glaser (1962). Com-
posed responses were used in order to avoid cues associated
With constructed responses. The symbolic logic material was
chosen since it permitted the E to control for previous ex-

perience, was fairly complex and yet held §s‘ interest.

14

 

 

 

 

The apparatus consisted of a MTA Scholar teaching
machine, a Hunter timer and an Esterline-Angus recorder.
The teaching machine and timer were wired so that the E
could control the delay intervals. The Esterline-Angus re-
corded latency to response and amount of time the § spent
examining the correct answer. The program was advanced by
die internal mechanism of the apparatus when S5 pressed a
button.

Subjects were allowed to control the amount of time
they spent on each frame. The §s constructed a response
following which either immediately or after an eight second
delay they were informed of the correct answer. The amount
of time spent examining the correct answer was also con—
trolled by SS. After this second self—determined interval
they were exposed again, either immediately or after an
eight second delay, to the next frame. In all treatments,
the stimulus material and the S's constructed response were
withdrawn from view during the delay intervals. The lengths
of the respective delays, i.e., the delays of IF and post=IF,
were predetermined as a function of the treatment specifica-
tions. Detailed initial oral instructions were given before
BXposure to any of the programed materials to each § regard-
ing the experimental procedure (See Appendix B). No further

instructions were given to the §s during the period of ex=

perimentation.

 

 

 

 

 

 

Each § received a five minute break midway through
the program and before administration of a criterion test
after completion of the program. This test consisted of 10
true-false questions, 10 recall questions and 8 deductive—
proof questions (See Appendix C); an open-ended general in—
formation questionnaire was also included in the test battery
(See Appendix D). Average total program running time per S
was approximately two hours, and varied from one hour and

twenty-five minutes to two hours and thirty—five minutes.

Design

A 2 X 2 X 2 factorial design was used with tWO de-
lays of IF (0 and 8 sec.), two lengths of post—IF interval
(0 and 8 sec.) and a sex factor. This experimental design
permitted the E to vary the effects of delay of IF (DIF) and
post-IF (PIF) orthogonally, to maintain an average inter-
trial interval of 8 seconds between these two factors, and
to control for possible massing effects between the various

treatment conditions.

Manama

Due to the differential complexity of material with-
in the symbolic logic program, it was necessary to allow §s
to control exposure time to the correct answer, i.e., the
length of the IF interval. This initiated a new variable
into Bourne and Bunderson's experimental design on which the

present study was based. However, this change was imperative

 

 

 

 

 

17

if the §s were to receive adequate information feedback.

It was assumed that this change would not alter the relation-
ship between length of IF intervals and number of errors
associated with the respective treatment conditions. All
zero order correlations indicated that longer feedback in-

tervals were not accompanied by lower error scores.

Treatments

 

The four treatment conditions were identified as
follows: 0-0 indicated complete self-pacing by §s, i.e., no
delay of IF or post-IF delay; 0—8 indicated no delay of IF
but an eight second post-IF interval; 8-0 indicated an eight
second delay of IF but no post-IF delay; and, 8-8 signified
an eight second delay interval before and after exposure to

correct answer, i.e., before and after IF.

Dependent Variables

The dependent variables used in this study were two
error scores, a response latency score and an answer latency
score. Number of errors indicated to what degree §s responded
correctly to program and criterion test items, and to the
eleven deductive—proof and easiest items within the program.
The correlation between program and criterion error scores
was .875. Response latency referred to the average amount
Of time spent by §s responding to programed items, and answer
latency to the average amount of time spent by §s examining

the correct answers to these items.

 

 

 

 

 

 

18

 

Figure 1 represents the temporal intervals. The
intertrial interval was defined as that period of time be-
tween response and the presentation of the next stimulus.
The delay of IF interval (DIF) referred to the amount of
time §s waited after responding before receiving IF, while

the post-IF interval (PIF) referred to the amount of time

 

 

§s waited after receiving IF before being exposed to the
next frame. The IF interval was synonomous with the answer

latency variable.

FIGURE 1

 

Temporal Intervals Common to the Present Study

 

DIF IF PIF
/F_A‘_\ /”—”k_—‘\
s, ——————— m. -—-—+s———4$2
\ Y J
ITI

Re liabili_t_y

 

Scorer reliability (testuretest) over a seven day
Period was .963 for program errors and .981 for criterion
test error scores. Recorder (interjudge) reliability was
.994 for response latencies and .999 for answer latencies.

The reliability of the criterion test was .843 (KR 21).

Hypotheses
General Hypothesis:

Recent research in the area of human learning

suggests that the post-IF interval may be a critical

IIIIIIlll::::_________________________¥ i,

 

 

l9

variable in mediated tasks and that delay of IF may not.

 

The data indicate that human learning may be facilitated by
increasing the post-IF interval and that the temporal rela—
tionship between delay variables is negligible. This study
will attempt to replicate these variables in a complex learn-

ing task.

 

 

Statistical Hypotheses:
A. Analysis of summated error scores on pro-
gram and criterion test
1. The delay of IF interval will not con-
tribute a significant source of variance

to the analysis, i.e., an increase in

 

the delay of IF interval will not be
accompanied by an increase in mean num-
ber of errors. Rejection of this hy—
pothesis will give support to the tra-
ditional interpretation that delay
retards acquisition.

2. The post—IF interval will yield a
significant source of variance such
that an increase in the post-IF inter-
val will be accompanied by a decrease

in mean number of errors.

 

 

The sex factor will not contribute a
significant source of variance to the
analysis, i.e., it is hypothesized that
females will perform as well as males
on this task.

The DIF and PIF interaction will be
insignificant. Rejection of this hy-
pothesis will indicate that delay of
IF does affect post-IF delay, and that
this effect is not constant over all
levels of delay of IF in this form of
learning.

The DIF by sex, and the PIF by sex in—
teractions will not yield a significant
source of variance to the analysis.
Rejection of these hypotheses will in-
dicate that the delay intervals have
differential effects on sex.

The three~way interaction between DIF,
PIF and sex will be insignificant in-
dicating that these variables are

independent.

Analysis of mean response latency scores

There will be no significant statisti-

cal findings associated with any of the

main or interaction effects for the analysis

 

 

 

 

21

of response latency scores. This will

indicate that amount of time spent to

solution will not differ (correlate) with

the various treatment conditions or sex.
C. Analysis of mean answer latency scores

1. The delay of IF interval will not con—

 

 

tribute a significant source of variance
to the analysis, i.e., an increase in
the delay of IF interval will not be
accompanied by an increase in answer
latency scores. If this hypothesis is
rejected, it will indicate the §s spent
a significantly greater amount of time

examining a correct answer when IF is

 

delayed than when it occurs immediately.
2. It is hypothesized that all other main
effects and interactions will be in-
significant. If this supposition is
supported, it will indicate that PIF'
and sex do not affect answer latency
scores, and that DIF, PIF and sex do

not correlate with each other.

 

RESULTS

Analyses of Error Scores

The analysis of variance summarized in Table 1 was
performed on total program error scores. All main and in-
teraction effects were insignificant except the delay of IF
by post—IF interaction (1><.05 ). A similar analysis, sum-
marized in Table 2, was performed on criterion error scores.
In this analysis, the delay of IF by post-IF interaction
failed to reach the .05 level of significance; however, it
did exceed the .10 tabled value of the F distribution.
Figure 2 represents geometrically the delay of IF by post—IF
interactions for total program and criterion error scores.

Summarized mean and variance data for total program
error scores is presented in Table 3. Individual tests be-
tween treatment variances revealed at the .01 level that
treatment 8—0 was significantly more variable than treat-
ments 0-0 and 8-8. All other tests between program error
variances were negligible. Table 4 denotes mean and variance
data for criterion error scores. All individual tests be-
tween program error variances were insignificant.

The analysis of variance for error scores on the
eleven deductive-proof items within the program is presented

in Appendix F. Here again, as in Table 1, only the delay of

22

 

 

 

 

 

23

TABLE 1

Analysis of Variance on Program Errors

 

 

 

Source DF F
Sex 1 .603
Delay of IF (DIF) 1 1.212
Postfeedback interval (PIF) 1 .004
Sex X DIF 1 .691
Sex X PIF l .041
DIF X PIF 1 5.640*
Sex X DIF X PIF 1 1.301
Residual MS 48 (37.50)

 

;><.05

 

 

 

 

 

 

24

TABLE 2

Analysis of Variance on Criterion Test Errors

 

 

 

Source DF F

Sex 1 .002
Delay of IF (DIF) 1 2.161
Postfeedback interval (PIF) 1 .018
Sex X DIF 1 .001
Sex X PIF 1 .000
DIF X PIF 1 2.821
Sex X DIF X PIF 1 2.922
Residual MS 48 (22.59)

 

 

 

MEAN NUMBER OF ERRORS

25

FIGURE 2
A Geometric Interpretation of The Delay
Of IF by Post-IF Interaction For

Program and Criterion Test Error Scores

 

——-——- PROGRAM ERRORS
I4 _ ———-CRITERION ERRORS
O-PIF
I2 -
'0 ‘ 8—PIF
/°O-PIF
8 _.
6 _ '8-PIF
4 _
I J

 

 

 

 

O-DIF 8‘9”

 

26

TABLE 3

i
!

Summarized Mean and Variance Data

For Program Error Scores

 

 

 

 

 

 

Source N Mean Variance
Male 28 11.15 42.78
Female 28 9.87 32.23
0—0 14 7.62 23.74
0-8 14 11.62 35.82
8=0 14 13.30 67.51
8—8 14 9.52 22.96
DIF—0 28 9.62 29.78
DIF-8 28 11.41 45.24
PIF—0 28 10.46 45.63
PIF=8 28 10.57 29.39

 

 

 

 

 

i TABLE 4

Summarized Mean and Variance Data
1 For Criterion Test Error Scores

 

 

 

 

 

 

Source N Mean Variance
Male 28 6.52 21.70
Female 28 6.47 23.49
O=O 14 4.59 26.29
0-8 14 6.55 28.55
8-0 14 8.58 19.19
8—8 14 6.27 16.35
DIF-0 28 5.57 27.42
DIF-8 28 7.43 17.77
PIF—O 28 6.59 22.74

 

PIF-8 28 6.41 22.45

 

 

 

28

IF by post-IF interaction was significant ( p<:.05 ).

The analysis of variance, summarized in Appendix G,
was performed on error scores for the eleven easiest program
items. These items were operationally defined as those
items on which average response latencies were equal to or
less than 15 seconds. All main and interaction effects were
insignificant for this analysis. These data indicate that
the delay of IF by post-IF interaction is not present for
program errors when length of program and/or difficulty is
removed from the analysis.

Appendices H and I summarize respectively mean and
variance data for the eleven deductive—proof and easy items.
The largest error scores in both of these appendices, as in
the analysis of program and criterion test material, were
associated with treatment 8-0. A rank order between treat—
ments similar to that observed in the analysis of program
and criterion error scores was obtained for deductive-proof,

but not for easy items.

Analyses of Latency Scores

The analysis of variance, summarized in Table 5,
was performed on average total program response latency
scores. A significant sex effect was found at the .05 level
indicating that females spent less time responding to each
frame on the average than did males. All other main and

interaction effects were insignificant.

 

 

 

A similar analysis on average total program answer
latency scores is presented in Table 6. This breakdown
showed that female answer latency scores were on the average
shorter than those of males (I><305 ). And, that §s
spent a significantly greater amount of time examining cor-
rect answers when their treatments were associated with
delayed feedback conditions ( p<.01 ).

Tables 7 and 8 summarize respectively mean and
variance data for average program response and answer
latency scores. No significant variance differences were
found for response latencies. But, the breakdown of answer
latencies revealed that treatments associated with delayed
IF were significantly more variable than those associated
with immediate IF ( p(.05 ).

Appendices J and K represent respectively the
analyses of variance performed on average response and
answer latency scores for the eleven deductive-proof items.
As in the analysis of program response latencies, females
were found to have significantly shorter response latency
scores than males. However, the analysis of deductive-proof
answer latencies failed to reveal, as did the analysis of
program answer latencies, significant sex and delay of IF
effects.

Appendix L gives mean and variance data for average

deductive-proof response latency scores. Male response

 

 

 

   

 

 

 

30

TABLE 5

Analysis of Variance on Average

Response Latency Scores

 

 

 

 

Source DF F
Sex 1 4.085*
Delay of IF (DIF) 1 .398
Postfeedback Interval (PIF) l .255
Sex X DIF l .105
Sex X PIF l .413
DIF X PIF 1 .543
Sex X DIF X PIF 1 .263
Residual MS 48 (134.67)

 

 

 

 

*p <.05

 

 

 

 

31

TABLE 6

Analysis of Variance on Average

Answer Latency Scores

 

 

 

 

Source DF F
Sex 1 4.908*
Delay of IF (DIF) 1 12.676**
Postfeedback Interval (PIF) 1 .007
Sex X DIF 1 .268
Sex X PIF 1 1.266
DIF X PIF 1 1.911
Sex X DIF X PIF 1 .701
Residual MS 48 (2.80)
~kp<.05

*7‘:
p<<.Ol

 

 

 

 

 

 

 

32

TABLE 7

Summarized Mean and Variance Data for

Average Response Latency Scores

 

 

 

 

 

 

 

Source N ' Mean Variance
Male 28 48.83 185.19
Female 28 42.56 99.55
0-0 14 46.64 73.16
0—8 14 42.79 198.28
8-0 14 46.37 111.19
8-8 14 47.03 86.91
DIF—O 28 44.77 165.72
DIF—8 28 46.70 99.05
PIF-0 28 46.51 92.18

PIF-8 28 44.91 142.60

 

 

 

 

 

 

33

TABLE 8

Summarized Mean and Variance Data for
Average Answer Latency Scores

 

 

 

 

 

 

‘ Source N .Mean Variance
Male 28 4.73 3.62
Female 28 3.74 1.97
0-0 14 3.11 2.18
0—8 14 3.77 .98
8=0 14 5.32 4.51
8~8 14 4.74 3.52
DIF-0 28 3.44 1.58
DIF-8 28 5.03 4.02
PIF-0 28 4.22 3.55
PIF—8 28 4.26 2.25

 

 

 

 

 

 

 

 

 

34

 

latencies, although failing to reach the .05 level of sig-
nificance, were approximately four times more variable than
those of females. A similar summarization for deductive-
proof answer latency scores is presented in Appendix M.
Individual analyses of answer latency variances revealed that
treatments associated with delay(s) were less variable than
the non-delay treatment.

Appendix N represents the analysis of variance per-
formed on average response latency scores for the eleven
easiest items. No statistically significant effects were
observed; but, females spent less time responding to ques-
tions than did males. The analysis of variance summarized
in Appendix 0 abstracts answer latency data for these easy
items. Significant sex and delay of IF effects similar to
those obtained for total program items were found in this
analysis. Shortest answer latencies were associated with
females, and with treatments having immediate IF.

Appendices P and Q summarize mean and variance data
respectively for average response and answer latency scores
for the eleven easiest items. All individual variance com—
parisons between response latencies were negligible. How-
ever, tests between answer latency variances revealed at
the .01 level that female performance was significantly less
variable than that of males, and that treatments associated
with delayed—feedback were significantly more variable than

those associated with immediate-feedback.

 

 

 

 

 

 

 

 

 

 

35

Exposure Time Data

Table 9 denotes summarized average exposure time
data for total program latency scores. No statistically
significant differences were found between the four treat-
ments for response latencies, however, the analysis of
,answer latency scores revealed that a significantly greater
amount of time was spent by SS examining correct answers
when treatments were associated with delayed-feedback.
Total average exposure time in seconds for the four treat-

ments was 49.75 for 0-0, 54.56 for 0-8, 59.64 for 8-0 and

 

67.77 for 8=8. As would be expected, the shortest exposure
time was associated with treatment 0-0 and the longest with

treatment 8-8.

Intertrial Interval Data

Theoretical and actual intertrial interval exposure
time (ITI) for individual treatments is presented in
Table 10. Theoretical ITI exposure time for treatment 0—0
was 0 seconds; for 0—8, 8 seconds; for 8-0, 8 seconds; and
for 8-8, 16 seconds. However, the actual average ITI for
treatment 0-0 was 3.11 seconds; for 0-8, 11.77 seconds;
for 8-0, 13.32 seconds; and for 8—8, 20.74 seconds. The
average theoretical ITI over all treatments was 8 seconds,
whereas the actual average ITI over all treatments due to
the confounding of ITI wita answer latency was 12.24

seconds.

 

 

 

 

 

36
TABLE 9

Summarized Mean Exposure Time

Data for Program Material*

 

 

Treatments 0-0 0-8 8-0 8—8

 

Total Exposure Time 49.75 54.56 59.64 67.77
Response Latency 46.64 42.79 46.32 47.03
Answer Latency 3.11 3.77 5.32 4.74

 

 

In seconds

 

 

 

 

 

 

Questiqnpaire Data
The analysis of questionnaire data is presented in
Table 11. §s responses were categorized independently by
two judges into behavioral classifications of fatigue
(physical), frustration (competing responses) and mediation
(task oriented information processing). Examples of repre-
sentative classifications of responses are presented in
Appendix E. There was 95% agreement between the classifica-

tions of these judges.

 

 

 

 

 

38

TABLE 10

 

Summarized Data on Intertrial
Interval Exposure Time*

 

 

Treatments 0—0. 0—8 8-0 8-8

 

Theoretical ITI Exposure
Time in Seconds 0 8 8 16

Actual Mean Exposure Time
as a Function of Confound-
ing ITI with the Answer
Latency Variable 3.11 11.77 13.32 20.74

 

Theoretical Average ITI Equals 8 Seconds

Actual Average ITI was 12.24 Seconds

 

*
In seconds

 

 

 

 

 

 

 

39

TABLE 11

Percentage Breakdown of Questionnaire Responses

As a Function of Treatments

 

 

 

 

Classification Treatments

of Responses 0-0 0—8 8-0 8-8
Fatigue 43 (6) * O (O) 7 (l) 7 (1)
Frustration O (0) 57 (8) 57 (8) 57 (8)
Mediation 71 (10) 86 (12) 100 (14) 79 (11)

 

7':
Each cell is based upon an n of 14. Numbers within

parantheses indicate number of §s upon which percentage
score is based.

 

 

 

 

 

DISCUSSION AND CONCLUSIONS

The data led to rejection of the general hypothesis
which stated that human learning of complex material would
be facilitated by increases in the post-IF interval irres—
pective of variation in the delay of IF interval, and that
the interaction between these delay variables would be neg-
ligible. The results show that a simple explanation in
terms of the independent effects of delay of IF or post-If
is inadequate.

The following decisions were drawn with respect to
the statistical error hypotheses. First, the hypothesis
that delay of IF in itself would not contribute a signifi—
cant source of variance was accepted since treatments asso-
ciated with delayed-IF were not accompanied by significantly
larger error scores. This result was inconsistent with much
of the literature on delay of reinforcement, but is in
agreement with the conclusions of Bourne and Bunderson (1963)
and Brackbill and Kappy (1962).

Secondly, the hypothesis that the post—IF variable
would contribute a significant source of variance was also
rejected in that treatments associated with post-IF delay

were not accompanied by significantly lower error scores.

40

 

 

 

 

41

 

1 Thus, the present data fail to support Bourne and Bunderson's
1 (1963) supposition that the post-IF interval is a critical
variable in human learning. Where Bourne and Bunderson
found that increases in the post-IF interval were accompanied
by decreases in error scores, the effects of post-IF delay
in this experiment were negligible.
The unique result of the present study was the in=
teraction between the delay of IF and post-IF variables.
This interaction was found for both complete program and
deductive—proof error scores. A similar interaction,
although not as pronounced, was also present for criterion

error scores .

 

For all analyses, except that for easy items, where
mean error scores were approximately the same, the fewest
number of errors was associated with the self-pacing treat—
ment, 0-0; second fewest with the treatment with the longest
intertrial interval, i.e., treatment 8~8; third fewest with
the treatment 0-8; i.e., the treatment with immediate feed~
back and an 8 second post-IF delay, and the greatest number
of errors with the treatment with delayed feedback and no
post-IF delay, i.e., treatment 8~O. The largest variance
values for total program items were found with treatments
having the greatest number of errors, and the smallest

variances with treatments having'the fewest number of errors.

 

 

 

 

42

Statistically significant differences in error scores for
sex or the other interaction effects were not found.

The data obtained in this study are at variance with
that reported by Bourne and Bunderson (1963), and by Brack—
bill and Kappy (1962), in that while the former found a
significant post-IF effect, this study did not, and while
the latter found no delay of IF effect in agreement with
Bourne and Bunderson, this experiment found that there were
conditions under which immediate IF did lead to a decrease
in error scores.

Both in Bourne and Bunderson's, and the present
study, delay of IF and post—IF delay were varied orthogonally.
The essential difference between these two studies was that
Bourne and Bunderson's tasks were peaked at a constant level
of difficulty such that IF could be presented in a fraction
of a second (a light flashed on the apparatus) and, as a
result, the E completely controlled the ITI. Whereas, in
the present study task complexity varied throughout the
program and necessitated the use of E, rather than E con-=
trolled IF, and consequently, complete E control of the ITI
was impossible.

Thus, Bourne and Bunderson's treatments are rela~
tively more massed than are those used in the present study.
The effect of massing should be most prevalent in their self-
paced treatment, Gel, and consequently, this treatment

should be associated with.maximum error scores. Massing

 

 

 

 

 

43

.

1 should not occur, however, under self—pacing, such as that

1 used in the present study, where the E controls length of
the IF interval. Moreover, the rank order between compar-
able delay treatment means in Bourne and Bunderson's and the
present study should be similar since massing should dissi—
pate when delay or IF and/or post—IF delay is introduced
into the experimental situation.

This interpretation was supported by a comparison
of comparable treatments between the two studies.1 Where
Bourne and Bunderson found their 0-1 treatment to be asso-
ciated with the largest number of errors, in the present
study the comparable 0-0 treatment produced the fewest num—
ber of errors. However, an examination of treatments where
delay intervals were similar, revealed identical rank orders
for all treatment means except those of easy items. The
greatest number of errors was associated with treatment 0-8
(0-9), the second greatest with treatment 8—8 (8u9), and the
third greatest with treatment 8—0 (8—1).
Moreover, even though massing per se did not appear

to be a crucial factor affecting performance in the present
Study, 43% of the §5 in the selfupaced treatment did com-
plain of physical fatigue due to length of program, whereas

7% or less made similar complaints in each of the delay

 

 

1Bourne and Bunderson‘s five dimensional concept
treatments were used for this comparison. These treatments
were chosen since the complexity of these tasks more closely
resembled that of the tasks used in the present study, and
since Bourne and Bunderson found the post—IF effect to be
more pronounced for complex concepts.

 

IIIIIIZL_______________________________44 444444

 

 

 

44

 

treatments. If fatigue due to massing is operating during
complete self-pacing, as it appears to be, its influence
should.be less than that associated with self-pacing in which
the E provides IF in a fraction of a second.

The controversial findings between Brackbill and
Kappy's (1962) serial learning study, and the present one,
can be attributed, in part, to the fact that their Es always.
received IF in a fraction of a second (one second to be
exact), whereas in this study, Es did not. Another dif-
ference was Brackbill and Kappy's failure to consider the
independent effects of delay of IF and post-IF delay. For
example, even though they maintained an ITI of 20 seconds
between their learning tasks, their immediate-feedback group
.was accompanied by a 19 second post-IF interval and their
delayed-feedback group (10 second delay) by a 9 second post-
IF interval. Thus, their immediate and delayed feedback
treatments were actually more representative, respectively,
of the 0-8 and 8-8 treatments used in the present study. A
comparison of their treatments, with the 0—8 and 8-8 treat-
ments of the present study, revealed that the largest number
of errors was associated with treatment O~8 (0-19), and
fewest with treatment 8-8 (10~9).

If Brackbill and Kappy had varied their delay vari-
ables orthogonally as did Bourne and Bunderson, and had run
0-1 and 8—1 treatments, it is probable that they would have

obtained an ordering among their treatment means similar to

 

 

 

 

 

 

 

 

45

 

Bourne and Bunderson's. If such data were compared with

: that obtained in the present study, it should differ only
with respect to treatment 0-0. And, this difference would
be attributable to the fact that the E controlled the feed-

back interval in the present study, whereas the E would

have controlled it in Brackbill and Kappy's study.

 

Analyses of mean response latency scores for total
program and deductive—proof items, revealed that females
spend significantly less time, on the average, answering
questions (responding to), than did males, and showed less

variability in performance. Similar findings, although not

 

as pronounced, were found for easy items. All other null
hypotheses were accepted as tenable.

The analyses of mean answer latency scores revealed
significant sex and delay of IF effects for total program

and easy items, but not for deductive-proof items. More-

 

over, the data indicated that females spent less time examin-
ing correct answers than did males, and that a significantly
greater amount of time was spent by Es examining correct

answers when their treatments were associated with delayed,

 

rather than immediate feedback conditions. It appears,

that when IF was delayed, that the number of interfering
(Competing) responses which the E had to process and/or
eliminate while examing the correct answer was increased,

and that consequently, it was necessary for him to spend more
time examining correct answers under delayed, than immediate

feedback conditions.

 

—¥— _

 

 

 

 

 

46

This interference interpretation of delay predicts
that as delay increases competing responses will: (1) occur
with increasing frequency, (2) lead to an increase in nume
ber of errors, and (3) increase the amount of time neces~
sary for the appropriate response to be acquired. The in-
significant delay of IF effect associated with deductive-
proof items appears to be due to the fact that these items
were too complex for the E to process (remember) over delay,
and consequently, when IF did occur, the E could not remem—
ber his original response 9359 he just moved on to the next
frame.

Additional support for this interference interpreta—
tion of delay was obtained in the analyses of main effect
variances where the data revealed, for both total program
and easy items, that performance was significantly more
variable under delayed than under immediate feedback con-
ditions. This finding was reversed, however, for the
analyses of deductive-proof variances where treatments asso—
ciated with delay(s) were found to be less variable than
the non-delay treatment. It appears that Es were not over-
whelmed by the differential complexity of tasks within the
total program per se, but that they were when the effects
of deductive—proof items were considered alone. Thus, the
data for total program items indicates that Es remembered
their responses over delay when tasks were not to complex,

and that Es increased the length of the feedback interval

 

 

 

 

 

 

 

 

47

 

when necessary, to determine where they had made a particular
error. However, when tasks were extremely complex, as they
were for deductive-proof items, it appears that there was

too much information for Es to remember over delay, and that
consequently, when IF did occur, they just moved on to the
next task without trying to figure out where or why they

had made a particular error.

Further support for this interference interpretation
was obtained from the analysis of the general information
questionnaire. The questionnaire revealed that 57% of the
§s in each of the delay treatments complained of frustration
while working through the program. It appears that competing
responses were initiated during delay, that these responses
were overtly characterized in terms of frustration, and
that Es were able to use the IF interval to process and/or
eliminate many of these competing responses. Moreover, it
appears that if delay of IF had been extended to the point
where Es could no longer cope with these competing responses,
that a significant delay of IF effect would have emerged.

Similar competing response interpretations of delay
have been formulated by Spence (1960) and Denny and Adelman
(1955) for animal learning, and by Denny for human learning.2
These Es surmised, as does the present paper, that delay

elicits competing responses which are overtly characterized

 

 

2Personal communication with M.R. Denny. Michigan
State University, East Lansing, Michigan, Fall, 196 .

 

 

 

 

 

 

 

 

 

48

by frustration, that these responses become conditioned to
the same stimuli as does the to~be-learned response, and
that acquisition time increases as a function of increments
in the delay variable. Consequently, such theories predict
maximal learning to be associated with immediate feedback
conditions.

This competing response interpretation of delay ex-
plains the individual treatment effects obtained in the
present study. As would be expected, the self-paced treat-
ment, 0—0, where delay was absent, was characterized by the
fewest number of errors, and by shorter answer latency scores.
Further, the questionnaire revealed that Es in this treatment
were not bothered by frustration (competing responses) while
working through the program. But, it did show that 43% of
the Es in the self~paced treatment complained of physical
fatigue due to length of program, whereas 7% or less made
similar complaints in the treatments associated with de-
laY(S).

It appears that competing behavior was minimized
under self-pacing, and that massing, rather than length of
program was responsible for the fatigue effect in treatment
0-0, since for all practical purposes the effect disappeared
when delay was introduced. Although the questionnaire also
revealed that 71% of the Es gave responses indicating that
they were actively involved (mediating) while working through

the program, the data is somewhat misleading when compared

 

 

 

 

 

 

 

 

g

49

with that obtained for delayed treatments since forced delays
were not common to this treatment. However, the data of the
self-paced treatment did show, when compared with that ob-
tained for delayed treatments, that when delays are not in-
discriminately forced upon Es, that Es vary delay as a func-
tion of need, and that mediation still plays an important
role in acquisition.

In the treatment associated with the second fewest
number of errors, i.e., treatment 8-8, Es spent a signifi-
cantly greater amount of time examining correct answers than

they did under the self—paced treatment. The questionnaire

 

revealed that 79% of these Es were actively processing in-
formation during delay, and that, whereas Es did not show
signs of frustration under self—pacing, that 57% of the Es
in this treatment did. These findings indicated that Es
needed and used the IF interval to eliminate competing res-
ponses which had developed during delay of IF, and that the
post-IF interval allowed Es to further process feedback in—
formation by minimizing competing behavior which would have
occurred with the immediate presentation of the next frame.
A secondary facet contributing to the effectiveness of
treatment 8—8 was its association with the longest ITI, and
consequently, with longest total program exposure time.

The analysis of questionnaire data for the treatment
associated with the third fewest number of errors, i.e.,

treatment 0—8, revealed that 86% of these Es were processing

 

 

 

 

 

    
   
      
      
   
   
   
   
   
   
   
    

50

information during the post-IF interval, and that 57% of
these Es complained of frustration while working through the
program. It appears that even though competing responses
were minimized by immediate IF, that Es were not able to
eliminate competing responses which occurred during post-IF
delay, since they were not able to return to the original
question or the correct answer, and that the presentation
of additional frames led to an increment in the number of
competing responses already present, and consequently, to
an increase in error scores.

The data associated with the treatment with the
greatest number of errors, i.e., treatment 8-0, indicated
that Es spent a significantly greater amount of time exam-
ining correct answers under this treatment than they did
under treatments associated with immediate feedback. The
questionnaire showed that all (100%) Es were actively in-
volved (mediating) during acquisition, and that 57% of these
Es complained of frustration. These findings indicated as
did those for treatment 8-8, that Es needed, and used, the
IF interval to eliminate competing responses elicited dur-
ing the delay of IF interval.

However, the data for treatment 8—0 also indicated
that the immediate presentation of the next frame, after IF,
was conducive to the development of additional competing
responses, since, this treatment was associated with maximum

error scores and differed from treatment 8-8 only in that it

L

 

 

 

 

 

 

51

lacked the 8 second post-IF interval. It appears that com-
peting responses associated with the immediate presentation
of the next frame combine with those already developed, and
that therefore, error scores reach a maximum under this

treatment. Although both Ramond (1954) and Renner (1963)

 

have found support in animal research for such an additive
inhibitory effect due to failure on behalf of Es to minimize
competing responses, this problem has yet to be extensively
investigated with humans.

In summary, the data led to the rejection of the

 

general hypothesis which stated that human learning of com»
plex material would be facilitated by increases in the post~
IF interval irrespective of variations in delay of IF, and
that the interaction between these delay variables would be
negligible. The results show that a simple explanation in
terms of the independent effects of delay of IF or post—IF
is inadequate. Fewest number of errors was associated with
the self—paced treatment 0-0, second fewest with treatment

8-8, third fewest with treatment 0-8, and the greatest num-

 

ber of errors with treatment 8—0.

As predicted, all delay effects for response latencies
were insignificant. However, the analyses of answer laten—
Cies revealed that Es spent a greater amount of time examin-
ing correct answers when treatments were associated with de-

layed-IF. A competing response interpretation of delay was

 

 

 

 

o

 

formulated with respect to error scores, answer latencies
and the delay variables.

Moreover, it appears that self-pacing with immediate
reinforcement increases the probability of a desired form of
behavior, that delay of IF and/or post—IF delay initiates
frustrative behavior which interferes with acquisition, and
that, post—IF delay is not a crucial variable in complex
human learning (programed) unless it is accompanied by de-

layed-IF.

 

 

 

 

 

 

 

 

 

 

 

 

 

BIBLIOGRAPHY

  

Adelman, H.M., and Maatsch, J.L. Resistance to Extinction
As a Function of the Type of Response Elicited by
Frustration. g. Exp. Psychol., 1955, 29, 61-65.

Ammons, R.B. Effects of Knowledge of Performance: A Survey
and Tentative Theoretical Formulation. g. Gen.
Psychol., 1956, 54, 279-299.

Amsel, A. Drive Properties to the Anticipation of Frustra-
tion. Psyczol. Bull., 1958, 52, 102-119.

Angell, G.W. The Effect of Immediate Knowledge of Quiz
Results on Final Examination Scores in Freshman
Chemistry. J. Educ. Res., 1949, 5;, 391-394.

Becker, P.W., Mussina, C.M., and Persons, R.W. Intertrial
Interval, Delay of Knowledge of Results, and Motor
Performance. Percept.Mot.Skills,1963, ll, 559a563..

Bilodeau, E.A., and Bilodeau, I. McD. Variation of Temporal
Intervals Among Critical Events in Five Studies of
Knowledge of Results. g. Exp. Psychol., 1958, 5;,
603—612.

Bilodeau, E.A., and Ryan, F.J. A Test for Interaction of
Delay of Knowledge of Results and Two Types of
Interpolated Activity. Q. Exp. Psychol., 1960, 22,
414=4l9.

Bourne, L.E., Jr. Effects of Delay of Information Feedback
and Task Complexity on the Identification of Con-
cepts. g. Exp. Psychol., 1957, 54, 201-207.

Bourne, L.E. and Bunderson, C.V. Effects of Delay of
Informative Feedback and Length of Postfeedback
Interval on Concept Identification. ;. Exp. Psychol.,
1963, pg, 1—5.

Brackbill, Y. The Impairment of Learning Under Immediate
Reinforcement. Unpublished paper, University of
Colorado Medical Center, 1964.

 

 

 

 

 

 

54

Brackbill, Y., and Kappy, M.S. Delay of Reinforcement and
Retention. J. Comp. Physiol. Psychol., 1962, 55,
14-18 3

Brown, J.S. A Proposed Program of Research on Psychological
Feedback (Knowledge of Results) In the Performance of
Psychomotor Tasks. In Research Planning Conference
pp Perceptual gpg Motor Skills. San Antonio, Texas:
Hum. Resour. Res. Cent., Lackland A.F.B., 1949.
(Conf. Rep. 49-2), 81-87.

 

Carlton, P.L. Response Strength as a Function of Delay of
Reward and Physical Confinement. Unpublished
Master's Thesis, University of Iowa, 1954.

Coulson, J.E., and Silberman, H. Results of an Initial
Experiment in Automated Teaching. Q. E9. Psychol.,
1960, 51, 135—143.

Denny, M.R., and Adelman, H.M. Elicitation Theory: 1. An
Analysis of Two Typical Learning Situations.
Psychpl. Rev., 1955, 62, 290—296.

Denny, M.R., et al. Supplementary Report: Delay of Knowledge
of Results, Knowledge of Task, and Intertrial In-
terval. J. Exp. Psychol., 1960, pg, 327.

 

Erickson, M,, and Lipsitt, L.P. Effects of Delayed Reward
On Simultaneous and Successive Discrimination
Learning in Children. J. Comp. Physiol. Psychol.,
1960, 5;, 256—260.

Evans, J.L. An Investigation of "Teaching Machine" Variables
Using Learning Programs in Symbolic Logic. Unpub—
lished Doctoral Dissertation, University of Pitts-
burgh, 1960.

Evans, J.L., Glaser, R., and Homme, L.E. An Investigation
of "Teaching Machine" Variables Using Learning Pro—
grams in Symbolic Logic. i. pg. §g§., 1962, 55,
433=452.

Evans, J.L., Homme, L.E., ard Glaser, R. The Ruleg System
For the Construction of Programmed Verbal Learning
Sequences. Q. Ed. Res., 1962, 55, 515—520.

Feldhusen, J.F., Ramharter, H., and Birt, A.T. The Teacher
vs. Programed Learning. Egg. 1. Ed., 1962, 25, 8~10.

 

 

 

 

 

 

 

55

 

Greenspoon, J., and Foreman, S. Effect of Delay of Know-
ledge of Results on Learning a Motor Task. g. Exp.
Psychol., 1956, 21, 226-228.

Guthrie, E.R. The Psychology of Learnin (Rev. Ed.), New
York: Harper and Bros., 19

 

Harker, G.S. An Experimental Investigation of the Effects
of Changes in the Delay of Reinforcement Upon Level
of Performance of an Instrumental Response. Unpub- !

lished Doctoral Dissertation, University of Iowa, , 1
1950.

Hough, J.B., and Revsin, B. Programed Instruction at the
College Level: A Study of Several Factors Influenc-
ing Learning. Phi Delta Kappan., 1963, 44, 286-291.

Hull, C.L. Principles pf Behavior. New York: Appleton-
Century-Crofts, 1943.

 

Kanner, J.H., and Sulzer, R.L. Overt and Covert Rehearsal
of Fifty Per Cent vs. Hundred Per Cent of the
Material in Film Learning. In A.A. Lumsdaine, (Ed.)
Student Response in Programed Instruction: A S
posium on Experimental Studies of Cue and Response
Factors* in Gropp and Individual Learning From In»
structional Media. Sponsored by Headquarters, Air
Research and Development Command, U.S. Air Force.
Washington, D.C.: National Academy of Sciences,
National Research Council, 1961, 427-441.

Kromboltz, J.D., and Bonawitz, B. The Effect of Receiving

the Confirming Response in Context in Programed
Material. g. Ed. Res., 1962, 25, 472-475.

 

Lorge, 1., and Thorndike, E.L. The Influence of Delay in

Affect-Effect ofra Connection. g. Exp. Psychol.,
1935,_1§, 187~194.

 

Lumsdaine, A.A. The Deve10pment of Teaching Machines and
Programmed Self-Instruction. New Teaching Aids for
the American Classroom. Stanford University: The
Institute for Communication Research, 1960, 136-173.

McDonald, F.J., and Allen, D. An Investigation of Presenta-

tion Response, and Correction Factors in Programed
Instruction. g. Ed. Res., 1962,_§§, 502-507.

 

 

 

 

56

McGuigan, E.J. The Effect of Precision, Delay, and Schedule
of Knowledge of Results on Performance. ;. Exp.
Psychol., 1959, éfi, 79-84.

Meyer, S.R. Report on the Initial Test of a Junior High
S 001 Vocabulary Program, 1960. In A A. Lumsdaine
and R. Glaser (Eds.), Teaching Machines and Pro-
grammed Learning. Washington, D. C. : National Educa-
tion Association of the United States, 1960, 229— 246.

Moore, J.W., and Smith, W.I. Knowledge of Results in Self“
Teaching Spelling. Psychol. Rep., 1961, 2, 717-726.

Mowrer, O.H. Learning Theory and Behavior. New York:
Wiley, 1960.

Noble, C. E., and Alcock, W. T. Human Delayed-Reward Learning
with Different Lengths of Task. g. E p. Psychol.,
1958, 56, 407— 412.

Pubols, B.H. Delay of Reinforcement, Response Preservation,
and Discrimination Reversal. g. Exp. Psychol.,
1958, pg, 32=40.

Ramond, C.K. Performance in Instrumental Learning as a Joint
Function of Delay of Reinforcement and Time of De—
privation. g. Exp. Psychol., 1954, 41, 248-250.

Renner, K.E. Delay of Reinforcement: A Historical Review.
Psychol, Bull., 1964, 61, 341-361.

Renner, K.E. Influence of Deprivation and Availability of
Goal Box Cues on the Temporal Gradient of Reinforce-
ment. g. Comp. Physiol. Psychol., 1963, 56, 101-104,

Rieber, M. The Effect of CS Presence During Delay and Re-
ward on the Speed of an Instrumental Response. p.
Exp. Psychol., 1961, p;, 290—294.

RYan, F. J., and Bilodeau, E. A. Countertraining of a Simple
Skill with Immediate and 1-Week Delays of Informa-
tion Feedback. _ Ex xp. Psychol., 1962, pg, 19- 22.

Saltzman, I.J. Delay of Reward and Human Verbal Learning.
:1. EXQ. PS!ChOl., 1951, EL, 437*439-

Saltzman, I.J. Kanfer, F.H., and Greenspoon, J. Delay of
Reward and Human Motor Learning. Psychol. Rep.,
1955, 1, 139—142.

 

 

 

 

 

 

 

 

57

Sheffield, F.D., et al. Drive Reduction versus Consummatory
Behavior as Determinants of Reinforcement. g. Comp.
Physipl. Psychol., 1954, 41, 349-354.

Shilling, M. An Experimental Investigation of the Effect
of a Decrease in the Delay of Reinforcement Upon
Instrumental Response Performance. Unpublished
Master's Thesis, University of Iowa, 1951.

Skinner, B.F. The Behavior pf Organisms. New York:
Appleton-Century-Crofts, 1938.

Skinner, B.F. The Science of Learning and the Art of
Teaching. Harvard Ed. Rev., 1954, 24, 86-97.

Spence, K.W. Behavior Theory and Conditioning. New Haven:
Yale University Press, 1956.

Spence, K.W. Behavior Theory and Learnipg. Englewood
Cliffs, New Jersey: Prentice-Hall, 1960.

Tolman, E.C. Principles of Performance. Psychol. Rev.,
1955, 62, 315-325. .

Wolfle, D. Training. In 8.8. Stevens (Ed.), Handbook pf
Experimental Psychology. New York: Wiley, 1951.

 

 

 

 

 

 

 

 

 

 

 

__.,_._-_ .4

APPENDICES

 

 

 

 

 

 

 

APPENDIX A

Rules Covered in the Programed Material

 

Wedge Adding (WA)
1. a
2. a.V’b (any letter) WA 1

Wedge Identity (WI)

1. a.V/a
2. a WI 1
Spear Separation (SS)
1. a—elb 1. 0
2. a p; 2. o—ajp
3. b SS 1,2 3. p SS 2,1

Tent Jointing (TJ)
1. z

2. a
3. z.flxa TJ 1,2 pg a/A.z TJ 2,1

 

Tent Disconnecting (TD)

 

1. a/A.b

2. a TD 1
D uble S ear (DS)

0 1. a24>b 1. c—a d ‘

2. b-+ c p; 2. b—e c

& mécDSLZ & bédDSLl

tt Exchan e (LE)
Le 1?r a/A b g 1. a\/ b

2. bAa LE1 95‘- 2. bVa LE1

58

 

 

 

APPENDIX B

 

Instructions Given to Subjects

    

 

You are about to learn some material by the method
of programed instruction. Your task is to answer each ques-
tion to the best of your ability from the material which

you have learned. Please write down your answer on the

Paper provided. When you have answered the question press

the button and wait for the correct answer to appear. This

Will enable you to see whether or not you have made a cor-

rect response. When you have finished checking the correct-

ness of your answer press the button again and wait for the

next question to appear.

If you have any questions please ask them now. You

Will receive a short break mid-way through the program. Now

press the button and begin.

59

M'

 

 

 

 

APPENDIX C

-- -‘—— __.__

Examples of Criterion Test Items

The following examples show the instruction to SS

and three sample items from each of the subtests.

 

True-False Test

In the following examples circle "T" if the LAST

step is a correct example of the rule indicated. If it is

 

an incorrect example, circle "F".

T F 1. g—911

2. g
3. g SS 1,2 (Spear Separation)

T F 1. a—a'b 1

2 x
3. b«—9c
4. a-dyc DS 1,3 (Double Spear)

T F 1. b—é b

2. a\/ a
3. b WI 2 (Wedge Identity)

60

 

 

 

 

 

 

Solve the following problems by using the rule

61

Recall Test

given at the LAST step of each example.

a‘V/e

LE 1

TJ 2,1

DS 2,1

(Letter Exchange)

(Tent Joining)

(Double Spear)

 

 

 

 

 

f

- __— ' —~ . h-‘ --

62
Deductive-Proof Test

In the following problems, use any rules that you
want in any order that you want and as often as you.want to
get to the Winning Step or "W". Do not spend too much time
on any one problem. You may go back and work on an income
plete problem later. Complete as much of each problem as

you can. Do NOT forget rule initials and step numbers.

1. x—ézr
2. x / W: r‘v’k

1. ‘a/\ b /' W: b

1. In_911

2. n-—)O
3. x/W: KARI—'90

 

 

 

 

 

 

APPENDIX D
General Information Questionnaire

The following questions are those used in the gen-

eral information questionnaire.

Instructions to S:

Please answer the following questions to the
best of your ability. Do NOT be afraid to state

your true impressions of the experiment.

1. What did you find yourself thinking about and/

or doing after making a correct response while

 

waiting for the next frame to appear?

2. What did you find yourself thinking about and/

or doing after making an incorrect response

while waiting for the next frame to appear?

3. Write down any comments which you might have

on this experiment.

CH
O.)

 

 

 

 

 

 

 

_ APPENDIX E

Representative Classifications of Responses

for Questionnaire Data

 

Fatique:

Frustration:

Mediation:

"It seemed to get tedious towards the end."

"I kept thinking, 'When will it ever be
finished?'"

"I felt somewhat discouraged and became aware
of my tiredness towards the end."

"Possibly the program was about ten minutes
too long."

"I wished the machine would hurry up--it was
frustrating when you had to wait."

"It was somewhat confusing as the rules
multiplied."

 

"I found myself wishing I had a reference
to all the rules in relation to each other

at one time for study."

"I often thought of unrelated things while
waiting for the next frame to occur."

"I kept going over in my mind the steps I
had used to arrive at the answer."

"I tried to think where I went wrong during
the delay so that next time I would not

make the same error."

"I kept trying to remember the method I used
for future problems."

"I found myself rehearsing when I had to
wait."

64

 

 

 

APPENDIX F

Analysis of Variance on Error Scores for
the Deductive-Proof Items

 

 

 

Source ' DF , F

Sex 1 .018

Delay of IF (DIF) 1 1.310

Postfeedback Interval (PIF) 1 .000

Sex X DIF 1 .080

Sex X PIF 1 .012

DIF X PIF 1 4.372*
Sex X DIF X PIF l .031

Residual MS 48 (5.02)

7kp < .05

65

 

 

 

 

 

 

 

 

APPENDIX G

Analysis of Variance on Error Scores for
the Eleven Easiest Items

 

 

 

 

 

Source DF F

Sex 1 .877
Delay of IF (DIF) 1 2.013
Postfeedback Interval (PIF) 1 .211
Sex X DIF 1 1.628
Sex X PIF 1 2.093
DIF X PIF 1 .318
Sex X DIF X PIF 1 1.494
Residual MS 48 (.62)

66

 

 

 

 

- .--——. .fi‘

Summarized Mean and Variance Data for Error

APPENDIX H

Scores on the Deductive-Proof Items

 

 

 

 

 

Source N Mean Variance
Male 28 3.24 4.50
Female 28 3.32 5.54
0-0 14 2.31 4.67
O=8 14 3.56 4.60
8-0 14 4.25 5.90
8-8 14 2.99 4.90
DIF-0 28 2.94 4.63
DIF—8 28 3.63 5.40
PIF-0 28 3.28 5.28
PIF-8 28 3.28 4.75

 

67

 

 

 

 

 

 

APPENDIX I

Summarized Mean and Variance Data for Error
Scores on the Eleven Easiest Items

 

 

 

 

 

 

Source N Mean Variance
Male 28 .78 .88
Female 28 .59 .35
0-0 14 .43 .25
0-8 14 .65 .49
8—0 14 .85 1.14
8-8 14 .82 .91
DIF-0 28 .54 .37
DIF-8 28 .83 1.03
PIF-0 28 .64 .70
PIF-8 28 .73 .70

 

68

 

 

 

 

 

 

 

APPENDIX J

Analysis of Variance on Average Response Latency
Scores for the Deductive-Proof Items

 

 

 

 

Source DF F
Sex 1 5,0779:
Delay of IF (DIF) 1 .228
Postfeedback Interval (PIF) 1 .029
Sex X DIF 1 .612
Sex X PIF 1 2.631
DIF X PIF 1 1.130
Sex X DIF X PIF l .002
Residual MS 48 (689.07)
7kp< .05

69

 

 

 

 

 

 

 

 

 

 

APPENDIX K

Analysis of Variance on Average Answer Latency

Scores for the Deductive-Proof Items

 

 

 

Source DF F

Sex 1 .655
Delay of IF (DIF) 1 2.007
Postfeedback Interval (PIF) 1 .048
Sex X DIF 1 .341
Sex X PIF 1 3.880
DIF X PIF 1 .415

1 .407
48 (22.87)

Sex X DIF X PIF

 

Residual MS

70

 

 

 

 

 

 

APPENDIX L

Summarized Mean and Variance Data for.Average Response

Latency Scores on the Deductive-Proof Items

 

 

 

 

 

 

 

Source N Mean Variance
Male 28 90.95 1160.70
Female 28 75.14 217.43
0-0 14 84.51 322.26
0-8 14 78.23 827.39
8-0 14 80.40 544.43
8-8 14 89.04 1062.19
DIF-0 28 81.37 574.82
DIF-8 28 84.72 803.31
PIF-O 28 82.45 433.35
PIF-8 28 83.64 944.79
71

 

 

 

 

 

APPENDIX M

Summarized Mean and Variance Data for Average Answer
Latency Scores on the Deductive-Proof Items

 

 

 

 

 

 

Source N .Mean Variance
Male 28 9.58 17.55
Female 28 8.54 28.15
0-0 14 7.89 50.43
0-8 14 8.43 10.16
8-0 14 10.52 15.50
8-8 14 9.42 15.42
DIF-O 28 8.16 30.29
DIF-8 28 9.97 15.46
PIF-O 28 9.20 32.96
PIF-8 28 8.92 12.79

 

72

 

 

 

APPENDIX N

Analysis of Variance on Average Response Latency
Scores for the Eleven Easiest Items

 

 

 

 

Source DF F
Sex 1 3.970
Delay of IF (DIF) 1 1.824
Postfeedback Interval (PIF) l .167
Sex X DIF 1 1.567
Sex X PIF 1 .720
DIF X PIF 1 .832
l .080

Sex X DIF X PIF

Residual MS 48 (7-74)

73

 

 

 

 

 

A PPEND IX 0

Analysis of Variance on Average Answer Latency
Scores for the Eleven Easiest Items

 

 

 

Source DF F
Sex 1 4.369*
Delay of IF (DIF) 1 9.440**
Postfeedback Interval (PIF) 1 .084
Sex X DIF l .685
Sex X PIF 1 .159
DIF X PIF 1 1.696
Sex x DIF x PIF 1 .944
48 (1.20)

Residual MS

 

*p<.05

*1—
p<.Ol

 

74

 

 

 

 

 

APPENDIX P

Summarized Mean and Variance Data for Average Response
Latency Scores on the Eleven Easiest Items

 

 

 

 

 

Source N Mean Variance

Male 28 10.88 6.48

Female 28 9.40 9.00

0—0 14 10.13 7.72

0-8 14 9.15 8.49

8-0 14 10.46 5.79

8-8 14 10.83 8.96

DIF-O 28 9.64 8.11

DIF-8 28 10.64 7.37

PIF-O 28 10.29 6.75

PIF—8 28 9.99 8.72
1
1
1
1
1
1

75

 

 

 

 

fi‘

 

APPENDIX Q

Summarized Mean and Variance Data for Average Answer
Latency Scores on the Eleven Easiest Items 1

 

 

 

 

 

 

Source N Mean Variance
Male 28 1.90 2.13
Female 28 1.28 .26
0~0 14 1.01 .12
0—8 14 1.27 .26
8-0 14 2.29 2.57
8-8 14 1.79 1.84
DIF-0 28 1.14 .19
DIF-8 28 2.04 2.20
PIF-0 28 1.65 1.35
PIF-8 28 1.53 1.05

 

76

 

 

 

 

 

 

 

 

 

 

 

 

 
 

 

 

 

 

 

 

 

   

 

 

 

 

 

 

 

 

3 1293 03196 0085

111111111