13.
‘

'm a

q
r
1

«in

'N

‘1'.

r m. 2......"
t . .
P... 3‘

‘
4t?!‘
9‘

as)
.‘E‘i.

3.3"“;
'4:

r
, u
‘4! . a

:A
3:.
0-5.

2

8

21'

SERSAL FQSI‘E‘QQN $52335? 5413 3:
T3

 

Hg,.:_:__::___:_:_i:_:_:_:_:g

an

599mg

’ 1

‘9

i

'314‘55

g:
d

5w
Ga

5"}.

NW 25

31

WW a L

:1 714' 5"

LIBRARY-

Michigan State

' .
Umvcrsity

PLACER WNXtomovothhdnckouMyumd.
TO AVOID FINEBMunonorbdmddo duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

MSU ION! Affirmative ActfiorVEquol Opportunity Indium
Was-m

ABSTRACT

SERIAL POSITION EFFECT AS A FUNCTION
OF INTER-TRIAL INTERVAL

by Louis G. Lippman

A discrimination theory for serial position effect
was developed. Basically, this theory stated that
primacy-recency is primarily dependent upon the
inter-trial interval (ITI) for providing a discriminable
order cue. Experimental conditions, (1) where the ITI
was set equal to the inter-stimulus interval, with and
without (2) dummy items at the onset of learning, and
(3) where the ITI was systematically varied through all
positions in the list were compared to a control
condition where the ITI was presented in a constant
location as in traditional serial learning.

All experimental conditions significantly reduced
primacy-recency effects and addition of dummy items
further curtailed primacy effects. When the ITI was
systematically relocated, learning was significantly
slower than when the ITI maintained a constant position,
thus indicating that rate of learning is more dependent
upon consistent order cues than on massing or spacing.

Finally, measures of "actual first“ and "objective first"

 

Louis G. Lippman

-2-

indicated that gs utilize and, when necessary,
establish discriminable cues for primacy. Reducing
the applicability, availability, or saliency of these
cues slows learning and reduces prrmacy-recency
effects. Consequently, the discrimination theory

was supported.

Wed/W MW

mteflfifl/éfé’)

SERIAL POSITION EFFECT AS A FUNCTION
OF INTER-TRIAL INTERVAL

By
Louis G. Lippman

A THESIS

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

MASTER OF ARTS
Department of Psychology
1963

ACKNOWLEDGMENT

The author wishes to eXpress his gratitude and
appreciation to Dr. M. Ray Denny, chairman of his
committee, for his guidance in this research and for
his assistance in the development of this manuscript.

He also wishes to thank Drs. Charles Henley and
Stanley 0. Ratner for their assistance, and Mr.

Joseph R. Levine for his advice during the formulation

of this research.

11

INTRODUCTION

PROCEDURE .
Subjects
Apparatus
Material

0

TABLE

Instructions . .

OF CONTENTS

Design and Measurement . . .

RESULTS . .
DISCUSSION .
SUMMARY . .
REFERENCES .
APPENDIX . .

iii

Page

oommms-tr-s-H

NMMH
[DI-’00)

TABLE
I.

II.

III.

IV.

LIST OF TABLES

Page

Summary of.Analysis of Variance for .

Learning Curves . . . . . . . . . . . . 14

Summary of Analysis of Variance
Comparing Groups' Performance on Items
1 through 4 of Serial Learning. . . . . 16

Summary of Analysis of Variance
Comparing Groups' Performance on Items
5 through 10 of Serial Learning . . . . 16

Summary of Analysis of Variance
Comparing Groups' Performance on Items
11 and 12 of Serial Learning . . . . . 16

Summary of Multiple Comparisons

between Treatment Means of the

Analyses Summarized in Tables I, II,

and IV, Applying Duncan's Multiple
RangeTeSt..............17

iv

LIST OF FIGURES

FIGURE
1. gearniggEgurves for Groups C, E1,
2 ' an O O O O O O O O O O O O O
2. Serial-position curves for Group
C andnGroup E2, plotted from "actual
first 0 O O O O C C O O O C O O C .
3. Serial-position curves for Group E1,
plotted from "actual first" and
'objective first". . . . . . . . . .
4. Serialhposition curves for Group E2,
plotted from "actual first" and
objective first". . . . . . . . . .
5. Serial-position curves for Group E3,

plotted from "actual first" and
"objective first". . . . . . . . . .

Page

10

ll

12

13

INTRODUCTION

Glanzer and Peters (1962), investigating influences
of temporal variables upon the shape of the serial
position curve, found a marked flattening of the serial
position curve when the inter-trial interval (ITI) was
eliminated. This suggests that the serial position
effect is primarily dependent upon the ITI's serving
as a highly discriminable cue for primacy (onset of
the list). Eysenck (1959) Performed a similar but less
extensive experiment and concluded that all theories
of serial position effect, except von Restorff type
(or discriminability) explanations, were endangered by
his findings.

A discriminability theory for serial position
effect can be formulated and stated as follows: The
ITI makes the beginning, or the first item of the list,
discriminable by temporally Spacing the last from the
first, by visual spacing or asterisks, and through
constant location (order cue). Through repetition and
constant location of the ITI the order cue builds up
and extends beyond the first item into the list
(Ebenholtz, 1963). Not only is the initial item
perceptually segregated from the remainder of the list

because of the ITI (primacy), but doubtless the second

-1-

-2.-

and third items are similarly made more discriminable,
though to a lesser extent. Further along in the list,
however, there is no clear or definite differentiation.
Thus with a long list the serial position curve tends
to flatten near the center, with the extent of the
flattening varying directly with the length of the
list (Murdock, 1962). In other words, order cue qua
primacy functions only for the first few items; the
items_in the middle are indiscriminable with respect
to order. With an ITI present the end of the list is
also somewhat distinctive. There are temporal cues
present as one nears the end of the list that indicate
to g that he is well along in the list. Therefore an
order one qua recency is also Operative. Consequently,
primacy-recency effects are the main order effects in
the list, with the ITI serving as the main means for .
the establishment of this compound temporal cue. Serial
position effect is thus viewed as a van Restorff effect
(Osgood, 1953, p. 504) with ITI providing the distinctive
stimulus (order).

The following are some predictions from this theory:
(a) Increasing the differential between rate of presen-
tation of the verbal items and the ITI (increasing the
difference between the ITI and the inter-stimulus
interval, 181) increases the discriminability of the

-3-

ITI and enhances the bowing of the serial position
curve. Bowing was increased by Hovland (1938b) when
the ISI was reduced, relative to ITI; and by Glanzer
and Peters (1962) when the ITI was increased relative
to the ISI. (b) Setting the ISI equal to the ITI
(essentially, eliminating the ITI) or constantly
relocating the ITI markedly reduces serial position,
effects because the order cue is severely attenuated.
(c) Increasing the discriminability of the first
verbal item in any way accelerates learning, as
Jensen (1963) has found. Conversely, elimination of
factors which make the first item discrhminable
retards learning; thus if the first item is not at
all distinctive and ITI equals ISI, then serial
position effects will be entirely eliminated.

The purpose of the present study was to test
(b) above and the second part of (c), as predictions

from the discrimination theory.

PROCEDURE

SubJects

§s were 80 students, secured on a volunteer basis,
who were enrolled in introductory psychology at Michigan
State University. gs who failed to understand or follow
instructions or who had previously participated in verbal
learning experiments were eliminated from the sample.
Those who had participated formerly in an auditory
paired-associate learning experiment, however, were
retained.
Apparatus

Apparatus for this experiment consisted of an MTA
100 Scholar (teaching machine), rollers for adaption
to repeated-loop materiall, and a Lafayette Repeat Cycle
Timer, model number one, for an external pacer. The
window on the teaching machine was reduced to 17/16 by
9/16 inches and was adjusted so that the verbal items
appeared approximately at the center of the window.
This type of apparatus was used in preference to a
standard memory drum because it allows trial-to-trial

change in presentation and eliminated influences from

 

I_Repeated-loop material was used in conditions C
and only; conditions E2 and E3 were presented

by continuous tape.

-4-

consistent extraneous cues, such as imperfections in
the tape.
Material

Verbal material employed in this experiment was
a specimen list of twelve CVC trigrams, following
typical rules for construction of nonsense syllable
lists, and having an average association value of
22.7 percent, according to Glaze: MEP, GAW, KOJ,
CIB, ZUR, TEY, WOQ, XIG, NAH, JEC, QUT, YOF (Hilgard,
1958, p. 540). The syllables were presented at a
two-second rate (ISI); the ten-second ITI consisted
of four two-second blank spaces followed by a two-
second presentation of asterisks.
Instructions

Hovland's (1938a) instructions, with minor
modifications, were used (Appendix). §s were given
a copy of the instructions which they followed while
the eXperimenter read the instructions aloud. Comments
clarifying §fs understanding of his task were made
when required.
pesign and Measurement

There were one control group and three experimental
groups:

(1) Group C where the syllables were presented
in traditional fashion with an ITI following each

complete trial.

(2) Group E1 where the trigrams were presented
without any ITI.

(3) Group E2 which was identical to E1 except
that four trigrams (dummy items which were presented
before the actual list began and never reappeared)
were added. These dummy items (VAF, SIJ, FEE, ZIW)
were of zero association value according to Glaze
(Hilgard, 1958, p. 544).

(4) Group E3 where the ITI was systematically
varied through all positions in the list. Specifically:
Trial one: l,2,...11,l2, (ITI),1; Trial two: 2,3,...
12,1, (ITI),2; Trial three: 3,4,...l,2, (ITI),3; etc..
Although all responses were recorded, §s' anticipations
to the repeated item following the ITI were eliminated
from consideration in order for comparability between
groups.

Responses were recorded for 24 trials, trial one
being Ss' first exposure to the material. All §s were
started in the task by the experimenter's removing a
slip of paper from the window simultaneously with the
appearance of the first item.

At the conclusion of the 24 trials, the experimenter
asked gs in the E conditions whether any item in the

list seemed to be first in any way, i.e., whether §

-7-

perceived any item as "first"; the experimenter noted

Ss' choices of "subjective first”.

RESULTS

The results are graphically presented in terms
of learning curves (Figure l) and serial position curves
(Figures 2,3,4, and 5). Two techniques for plotting
serial position curves from the E groups' data were
applied: (a) From.the item actually presented first
(having first position on the list) and (b) from'gs
"objective first", defined as the item correctly
anticipated most frequently by an §/. In case of ties
the tied item correctly anticipated earliest in learning
was employed. The measure of "subjective first”,
described in the procedure, was discarded since gs
evaded the question, failed to understand what was
being asked, or tried to recall which item had, in fact,
been presented first.

A visual examination of the serial position curves
indicates that group C produced a typical serial position
curve with primacy and recency effects. E1, when plotted
from "actual first", showed slight, if any, primacy
effects and marked reduction in recency; however, the
plot from "objective first" retained both primacy and
recency. E2, plotted from "actual first" showed no
primacy-recency; the plot from "objective first" displayed
very slight primacy-recency. Finally, E3, plotted from

-8-

MEAN N9 CORRECT ANTICIPATIONS

Figure 1.

GROUP C\ 'I’\ z”\\ /’
GROUP Eu ./‘-\ ,’ W“ a

-‘ v" GROUP E3
. / \
.. GROUP :2

.1.1.1.1.1.1.141.i
4 6 8 IO l2 l4 l6 I8 20 22 24
TRIALN9

Learning curves for Groups C, El' E2, and E}.

-10-

 

 

 

.1
S '3 -
a:
F h
35
m H '-
(I)
Z L
Q
+- 9 - GROUP c
3.5
2 I-
.—
3; 7 ' GROUP £2
5 _ (FROM “ACTUAL FIRST”)
h]
K 5 //’\\\
g I" r" J \
U ,’ \\ I’ \\
\ I I
g 3 b I, \\ I, \\ IA\
/ \ I \\ \x
Z ’ \ I VI \
< *- I, \y’ \\
U \
2 | t ’ \\
l l l . I I I
O |23456789I0ll|2

SERIAL POSITION

Figure 2. Serial-position curves for Group C and Group E2,
plotted from "actual first".

MEAN N9 CORRECT ANTICIPATIONS PER TRIAL

G?

63

Q)

‘4

C”

(o

i

-11-

 

 

 

I.
t I
I
FROM “OBJECTIVE FIRST”
- “/
I- I‘ FROM “ACTUAL FIRST”
I
‘ I
I
_ I
p
b
I I l j J l 1 1 l I I I
I23456789IO|Il2
, SERIAL POSITION

Figure 3. Serial-position curves for Group E , plotted

from "actual first" and "objectiIe first".

v12-

I
I
I
I
9)- I‘
I
\KFROM “OBJECTIVE FIRST"

I
\ FROM “ACTUAL FIRST"
I
I
I
I

CORRECT ANTICIPATIONS PER TRIAL

   
 

MEAN N9
"as

 

o L 1 l l 1 L 1 1 l J

I 2? 3 44. .5 6 7' 8 9 IC) II I2
SERIAL POSITION

Figure 4. Serial-position curves for Group E2, plotted
from "actual first" and "Objective first".

II -
OM “OBJECTIVE FIRST”

FR
FROM “ACTUAL. FIRST ”

 

 

MEAN N9 CORRECT ANTICIPATIONS PER TRIAL
V

I
5 I’ I ’\
I \
+- I,’ \
\\
3 - “‘~.
\\\~ ll---
.- ~‘~‘, -- -~’
| p
L n 1 1 1 L 1 4 L 1 1 n
O|23456789|O|l|2

SERIAL POSITION

Figure 5. Serial-position curves for Group E3, plotted
from "actual first" and "objective first".

-l4-

"actual first", showed a spread of primacy and absence
Of recency; the plot from "Objective first" showed a
slight primacy and virtually no recency.

An analysis of variance was performed on the data
plotted in Figure 1 (Table I), and all possible
comparisons between treatment means were made (Table V,a).
The F is significant at the .005 level, and the multiple
comparisons indicate that group C learned significantly
faster than each of the E groups. There are no
significant differences between E groups but certain
trends appear. The learning curves show that groups
E2 and E1, in that order, tended to learn faster than
E3. Group E3 appeared to learn slowly at the start but
later in learning surpassed groups E and E .

l 2

Table I. Summary of Analysis of Variance for Learning

 

 

 

Curves
Source d.f. Mean Square 3
Between groups 3 3424.19 5.57*
Within groups 92 614.45

 

*p < .005

-15-

The serial position curves were arbitrarily
partitioned into three sections and the groups were
compared for each of these measures in order to
determine the effects of the experimental manipulations
(E groups) upon primacy-recency effects. It was
assumed that the first four items would adequately
represent any primacy effects, the last two items
would describe recency effects, and items 5 through
10 would be relatively independent of primacy-recency
(Murdock, 1962). The results of each analysis of
variance using these measures are summarized in Tables
II, III, and IV; the results of multiple comparisons
between treatment means for these analyses are
summarized in Table V, b and c.

The'g for primacy effects is significant at the
.0005 level; the multiple comparisons indicate that
all E conditions clearly reduced primacy effects.
0f great interest is the fact that the comparison
between E1 and E2 on primacy showed that addition of
dummy items significantly curtailed primacy effects
(p (.05). The F for recency effects is significant
at the .0005 level; the multiple comparisons show
that all experimental manipulations significantly
reduced recency effects. Finally, thelg for the
"middle" items is not significant, in agreement with
Murdock's (1962) findings.

-15-

Table II. Summary of Analysis of Variance Comparing
Groups Performance on Items 1 through 4
of Serial Learning

 

 

 

 

Source d.f. Mean Square 3
Between groups 3 2185.48 8.90*
Within groups 76 245.67

*p {.0005

Table III. Summary of Analysis of Variance Comparing
Groups Performance on Items 5 through 10
of Serial Learning

 

 

 

 

Source d.f. Mean Square 33
Between groups 3 221.28
Within groups 76 232.30

*n.s.

Table IV. Summary of Analysis Of Variance Comparing
Groups Performance on Items 11 and 12 of
Serial Learning

 

 

 

Source d.f. Mean Square F
Between groups 3 308.88 13.04*
Within groups 76 23.69

 

*p (.005

-17-

 

 

.m.c OH.~ .m.e om.m .m.e Na.a a .m» m
.m.e mm.o .m.: mm.m .n.a mm.~ mm .m» as
.m.: mm.a mo. ma.~a .m.c mo.m m .m> an
ace. $6 80. 8.3 mo. :3: mm .2 0
H00. mm.m Hoe. oa.m~ Hoe. as.am .Nm .m> 0
H00. oa.a mo. mm.ma moo. mw.H~ Hm .m> o
a oososommfin_cwos a oOCOHOHMHQ new: A oosouomman sees

 

NH use Ha mamSH Heasom
AsH caneav

AIR maopH Haautm
AHH mansev

mo>nso wsfisameq

AH massav

voaegsoo manage

 

 

o

Aoqauoma .am .ooma .messeemv smog amass oflaasass m.:moasa messages .sH new .HH .H
moahwa ca possessssm momhadsd esp mo enema Psospmena soospon.msomfismmaoo camwpfisa Mo huessdm

m

4

.> sands

DISCUSSION

The results Of this experiment clearly support
the discrimination theory of serial position effect.
When the cue value of the ITI is reduced, rate of
learning drops and primacy-recency effects are curtailed;
of considerable interest is the fact that E} learned no
faster than the other E conditions although it was
Spaced. Primacy-recency effects are directly dependent
upon the ITI for providing the order cue. When the ITI
equals the ISI and when dummy items are introduced so
that even on the first trial there is no "first" item,
then primacy-recency effects are totally eliminated.
In other words, seeing the first item as first on Just
one trial makes the first item more discriminable than
the other words on the list.' In condition E2, Ss'
only possible cue of this sort was their perception
Of an item's repetition; but since the first item so
recognized was directly dependent upon individual
differences and not upon a consistent, repeating cue
in the list, primacy-recency effects were eliminated.
The results from the plots from "objective first"
generally support this assumption that g uses

idiosyncratic primacy cues.

-18-

-19-

Finally, the results for group E seem to indicate

3
that SS in serial learning situations will take a
potentially confusing cue and convert it into a cue

to facilitate learning. Specifically, in E3, some

§s attended to the ITI as a cue to indicate "the next
item to be learned", i.e., once an s learned the pattern
of ITI relocation there was an extra cue for this §,

To eliminate this possibility, in any future research
the location of the ITI could be a random affair.

SUMMARY

A discrimination theory for serial position effect
was developed. Basically, this theory stated that
primacy-recency is primarily dependent upon the inter-
trial interval (ITI) for providing a discriminable
order cue. Experimental conditions, (1) where the ITI
was set equal to the inter-stimulus interval, with and
without (2) dummy items at the onset of learning, and
(3) where the ITI was systematically varied through
all positions in the list were compared to a control
condition where the ITI was presented in a constant
location as in traditional serial learning.

All experimental conditions significantly reduced
primacy-recency effects and addition of dummy items
further curtailed primacy effects. When the ITI was
systematically relocated, learning was significantly
slower than when the ITI maintained a constant position,
thus indicating that rate of learning is more dependent
upon consistent order cues than on massing or spacing.
Finally, measures of "actual first" and "objective
first" indicated that gs utilize and, when necessary,
establish discriminable cues for primacy. Reducing
the applicability, availability, or saliency of these
cues slows learning and reduces primacy-recency effects.
Consequently, the discrimination theory was supported.

-20-

REFERENCES

Ebenholtz, S. M. Position mediated transfer between
serial learning and a spatial discrimination task.

g, 25p. Psychol. 1963, §§, 603-608,

Edwards, A. L. Ex erimental design ip psychological
research. Revised ed.), New York: Holt, Rinehart

and Winston, 1960.

Eysenck, H. J. Serial position effects in nonsense
syllable learning as a function of interlist rest
pauses. Brit. g, Psychol. 1959. 59, 360-362.

Glanzer, M. and Peters, 8. C. Re-examination 0f the
serial position effect. g. exp. Psychol. 1962,
64, 258-266.

Hilgard, E. R. Methods and procedures in the study Of
learning. In S. S. Stevens (Ed.), Handbook pf
experimental s cholo . New York: John Wiley &
Sons, Inc., 195 , Pp. 517-566.

Hovland, C. I. Experimental studies in rote-learning
theory. I. Reminiscence following learning by
massed and by distributed practice. g, 252. Psychol.
1938a, pp, 201-224.

Hovland, C. I. EXperimental studies in rote-learning
theory. II. Reminiscence with varying speeds of
syllable presentation. g, gap, Psychol. 1938b,
2—2.! 338-353 0

Jensen, A. R. Transfer between paired-associate and
serial learning. g. verbal learn. verbaI Behav.

Murdock, B. B., Jr. The serial position effect of free
recall. g. exp. Psychol. 1962, 64, 482-488.

Osgood, C. E. Method and theopy pp experimental
psychology. New York: Oxford University Press,
953.

 

-21-

APPENDIX

INSTRUCTIONS

This is an experiment in learning a list of
nonsense syllables, and not a psychological test. We
are interested in certain complex relationships in the
learning process which are common to all people, and
are not concerned with your personal reactions.

Shortly after the apparatus starts you will see
a three-letter syllable in the window. You are to
spell this syllable and those that follow it. As you
learn, you are to try to anticipate the syllables; in
other words, as you see one syllable you are to spell
the syllable that will follow it BEFORE it appears.

If you think you know what a syllable will be, but are
not sure, guess, because it will not hurt your score
any more than to say nothing. And if you get it

right it will count as a success. Always try to Spell
the syllables as distinctly as possible. You will
continue anticipating the syllables until I tell you
to stop.

Do not ask questions about the purpose of the
study until the experiment is over. Do you have any

questions regarding your task?

-23-

RU‘ji‘ Up»:- ONLY