—'

 

 

ORDER EFFECTS IN LEARNING
A SCALE OF NDGMENT

145
913

That: for the Deg-m of M. A.
MECBIGAN STATE UNIVERSITY

Robert H. Terborg
1966

 

ABSTRACT
ORDER EFFECTS IN LEARNING A SCALE OF JUDGMENT

By Robert H. Terborg

Although some work has been done in the area of
learning a frame of reference or scale of absolute judg-
ment, little has been done pertaining to the effect which
different orders of stimulus presentation have on such
learning. Never and Zener (1928) suggested that when a
subject has no previous knowledge of the stimulus series
he tends to reapond with middle categories to the first
few stimuli presented. They state that once the subject
has knowledge of the end stimuli he uses these stimuli
to judge the other stimuli which are presented. They
suggest that orders which present the end stimuli very
early will lead to faster learning than orders which do
not present the end stimuli very early.

The present study investigated two hypotheses
concerning two different types of orders, type I orders
in which the extreme stimuli were presented very early
in the order and the central stimuli very late in the
order, and type II orders in which the central stimuli

were presented very early and the extreme stimuli were

Robert H. Terborg

presented very late. The first hypothesis was that sub-
jects who received type II orders would outperform
subjects who received type I orders on the first trial
because all subjects would tend to use middle numbers for
their first few responses since they had no previous know-
ledge of the stimuli. The second hypOthesis was that for
the following trials those who had received type I orders
would be more accurate in their judgment than those who
had received type II orders because those who received type
I orders would have immediate knowledge of the stimulus
range and could use the end stimuli to judge the rest of
the stimuli presented in that trial. '

Fifty subjects were randomly divided into two
groups of 25 each. Stimuli consisted of eight slides
with dots on them. Each slide had a different number of
dots on it. Subjects were required to assign numbers
from one to eight to each slide (assigning the number
”one" to the slide which they thought had the least num-
ber hr dots and the number "eight" to the slide which they
thought had the greatest number of dots). Subjects in
Group I were presented with type I orders fer the first
three trials and subjects in Group II were presented with
type II orders on the first three trials. For the fourth
and fifth trials both groups were presented with the same
orders which were equally dissimilar to the type I and

the type II orders.

Robert H. Terborg

Group II made less absolute error than Group I on
trial A supporting Wever and Zener's suggestion that sub-
jects tend to use middle categories as responses to the
first few stimuli presented when they have no prior knowa
ledge of the stimulus series. On trials B and C Group I
made less absolute error than Group II, also the amount
of information transmitted was greater for Group I. On
trials D and E there was no difference between Groups I
and II in absolute error or in amount of information trans-
mitted. It was concluded that orders which present the
extreme stimuli very early and the central stimuli very
late lead to greater accuracy in judgment than orders
which present the central stimuli very early and the ex-
treme stimuli very late. However the amount of learning
transferred to unbiased orders is approximately the same
from both types of orders. This confirms Never and Zener's
(1928) suggestion that orders which present the end stimuli
at the beginning of the order lead to quicker learning of

the scale of judgment.

 

ORDER EFFECTS IN LEARNING A SCALE
OF JUDGNENT

Ev

Robert H. Terborg

A THESIS

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

MASTER OF ARTS

Department of Psychology
1966

To 1‘wa Wonderful Parents

ii

ACKNOWLEDGELKTS

It is with much appreciation and gratitude that I
acknowledge those who have helped make this research a
success. The greatest thanks are due my committee chair-
man, Dr. Donald H. Johnson, whose guidance and patience
were gratefully felt throughout the project. I also wish
to thank Dr. Terrence M. Allen and Dr. Paul Bakan for
their encouragement and for consenting to serve on the

thesis committee.

iii

TABLE OF CONTENTS
Chapter Page

I. INTRODUCTION . . . . . . . . . . . . . . . .
II e IETHOD e e e e e e e e e e e e e e e e e e 0
Subjects . . . . . . . . . . . . . . .

Katerials and Apparatus . . . . . . . .
Procedure . . . . . . . . . . . . . . .

msqxa \J +4

III 0 RESULTS 0 O O O O O O O O O O O O O O O O O 12

Error Analysis . . . . . . . . . . . .
Correlation Analysis . . . . . . . . .
Information Analysis . . . . . . . . .
Analysis of Extremity of Response . .

16
l8

l9
IVe :DI:)C:IJTESS:[OI.T e e e e e e e e e e e e e e e e e 21

V. SUIH'MRY O O O O O O O O O O O O O O O O O O 28

REFERBIJCES O O O O O O O O O O O O O O O O O O O O O O 3]—

iv

Table
1.

Analysis of variance of the absolute error
Of SUb‘jeCtS O O O O O O 0 O O O O O O O O O O

Kean absolute error per subject per trial . .

Mean absolute error per subject after trials
were combined . . . . . . . . . . . . . . . .

Correlation between stimulus and response
for each group on trials A-E . . . . . . . .

Information transmitted per trial (in bits) .

1A
15

16

18
19

LIST OF FIGURES

Figure Page

1. Scatter plot of the responses to the 8
stimuli for all 50 83 on trial E. . . . . . . 13

2. Mean absolute error per subject per trial . . 17

vi

LIST OF APPENDICES

Appendix Page
A. ORDERS OF STIMULUS PRESENTATION . . . . . . . A-l

vii

INTRODUCTION

Research on organization of a frame of reference
or a scale of absolute judgment has in general, followed
the psychophysical tradition and emphasized stimulus varia-
bles, but recently a few experimenters have been concerned
with learning variables. Eriksen and Hake (1955b) have
shown that learning of a frame of reference or scale of
absolute judgment is easier for multidimensional stimuli
than for unidimensional stimuli. Eriksen (1958) states
that knowledge of results helps very little in a subject's
learning a scale of judgment. He suggests "that what the
subject learns on this task is not to recognize individual
stimuli as such, but instead to establish a frame of re-
ference for the series of stimuli as a whole."

Parducci (1956) has shown that judgment scales may
be incidentally learned, although such learning is less
effective and subject to greater interference than inten-
tional learning. Murphy (1966) has shown that knowledge
of results facilitates the learning of Escales of judgment
when the stimuli used are various durations of time.

Gibson (1953) has stated that in experiments on
absolute judgment a subject's knowledge of the range of a

set of stimuli can affect his judgments and stabilize them

in predictable ways. Even though the subject does not
receive knowledge of results, his experience with the
stimuli seem to provide him with adequate knowledge of
the range. Johnson (1945) and Never and Zener (1928)
imply that an individual can make judgments fairly well
after he has had an opportunity to become familiar with
the range of the stimulus series.

Wedell (1934) in an experiment concerning the
absolute judgment of pitch has shown that subjects use
the end stimuli of the series as reference points for
judging the other stimuli in the series. His subjects
were also able to recognize the end stimuli when they
occurred. Eriksen and Hake (1957) - in an experiment
involving the absolute judgment of hues - showed that
subjects tend to pick one or two hues from the stimulus
series and then proceed to use these hues as references
to judge the other stimuli in the series. These stimuli
are usually the end stimuli on the continuum of the stim-
ulus series.

Garner (1953) suggested that the prominence of
such end effects might be partially due to a response
attenuation effect that artifactually led to unidirection-
al errors at the ends of the stimulus series, i.e., in
the middle of the stimulus series, the subject can make
an error in either of two directions when he assigns a

response to the stimulus while for the end stimuli, the

subject can make an error in only one direction. Garner
(1953) and Eriksen and Hake (1955a) have shown that this
artifact only partially accounts for such anchor effects

at the end stimuli. Eriksen and Hake (1957) present an
alternative explanation which they call the subjective-
standard hypothesis. This hypothesis states "that S when
confronted with a series of stimuli in an absolute-judgment
task, selects certain stimuli that he then uses as stan-
dards for judging the remaining stimuli. When a stimulus
is presented, S attempts to recall one of these standard
stimuli and uses it as a reference in judging the presented
stimulus." The stimuli selected are usually (but not al-
ways) the end stimuli.

Johnson and King (196A) investigated the conditions
under which an end stimulus would be selected as the stan-
dard to which the response scale is anchored. They found
that an end stimulus serves as an anchor when and only when
it has some salient property which identifies it "as an
end beyond which there are no other stimuli to be judged.”

Little, if any, work has been done pertaining to
the effect which different orders of presentation have on
the learning of a frame of reference in absolute judgment.
If a general rule could be discovered which would enable
one to construct orders which would facilitate the learning
of a frame of reference, rather than those which might in-

hibit such learning, it would be possible to present the

members of the stimulus series, in an order which would
lead to faster or greater learning than other orders.

It was the purpose of this experiment to attempt to find
a general type of order of presentation of the stimuli,
such that the learning of the frame of reference would
be facilitated.

Since individuals seem to establish a frame of
reference for a stimulus series rather than learning to
recognize each individual stimulus, and since this is
done by selecting stimuli which have some salient property
and using them as standards for judging the remaining stim-
uli, it was hypothesized that orders which present the
salient stimuli very early and the other stimuli later
in the order would produce faster learning than orders
which present the central stimuli very early and the sa-
lient stimuli very late in the order. In the present
study the salient stimuli were the end stimuli and thus
if the end stimuli vvere presented early in the order, the
subject should be able to learn the range of the stimuli
quicker than he would be able to if the salient end stim-
uli were presented late in the order, because then the
subject would not have immediate knowledge of the stimulus
range. This is similar to the suggestion made by Wever
and Zener (1928).

In a pilot study it was found that subjects who

receive the central stimuli early in the order (Group II)

performed better on trial A then those who received the
extreme stimuli early in the order (Group I). It was
felt that this was due to the construction of the orders
which Groups I and II were presented. Since the subjects
had no prior knowledge of the stimuli, the first trial
was mainly a measure of performance rather than learning.
It was felt that subjects would tend to use central hump
bers to judge the first few stimuli presented on trial A
because this is more conservative than using extreme
numbers. If the subjects proceeded in such a manner then
Group II would perform better than Group I (because the
orders for Group II present the central stimuli very early in
the order). Never and Zener (1928) state that when a per-
son has no prior knowledge of the stimuli to be presented
his first judgment tends to be the middle of the possible
response categories.

The investigator's first hypothesis was that Group
II would perform better than Group I on the first trial.
The second hypothesis was that Group I would learn the
scale of judgment faster than Group II because they re-
ceived the extreme stimuli very early in the order on the
first three trials. This aspect was looked at in two ways:
1) by comparing Group I with Group II on trials two and
three, 2) by comparing Group I with Group II on the un-
biased orders presented in trials four and five.

The pilot study also suggested that after four or
five trials further learning could only be expected to oc-
cur in situations of controlled attention, and motivation
because subjects tend to become bored and fatigued on this
task. It also suggested that a subject's expectation of
the end of the experiment hinders his performance on the

last trial presented.

l-ETHOD

Subjects: Subjects consisted of 50 undergraduate
students enrolled in an introductory psychology course at
Michigan State University. Subjects received experimental
credit in this course for participating in the experiment.
They were randomly assigned to one of two equal groups
(n=25).

Material and Apparatus: The materials used con-

 

sisted of eight, 2 X 2 slides with black dots on a white
background. Each slide had a different number of dots on
it. Slides one through eight had 22, 26, 30, 35, Al, 48,

56 and 65 dots on them respectively. The number of dots

on each slide increased logarithmically so that approximate
linearity of response from the subjects could be assumed.
Reese (1953) gives evidence which indicates that a logarith-
mic increase would yield approximately linear responses.

A pilot study was conducted by the investigator and approxi-
mate linearity of response was achieved.

A tachistisc0pic projector was used to present the
slides on the light colored wall of a very dimly illumi—
nated room. The projected image on the wall was 5 feet,

A inches square. The projected dots had a diameter of 3
inches. Subjects sat approximately lh% feet from the center

of the projected image.

The end stimuli (the slides with 22 and 65 dots)
were shown by a pilot study to be salient to the extent
that subjects used them to judge the other stimuli.

Procedure: The subjects were randomly divided
into two equal groups: I and II. Those in Group I were
presented with orders which had the extreme stimuli at
the beginning and the central stimuli at the end of the
order. Those in Group II were presented with orders which
were the exact reverse (i.e., inverted) of those presented
to Group I (thus for Group II the central stimuli appeared
in the early part of each order and the extreme stimuli
appeared in the last part of the order). Extreme stimuli
were slides l, 2, 7, 8. Central stimuli were slides 3, h,
5, 6.

Subjects were presented with a total of gigg
trials so that the onset of boredom and fatigue could be
prolonged until after the fifth trial. Only the data from
the first five trials were analyzed. Trials 1-5 are de-
snibed below. Trials 6-9 were simply presentations of
orders chosen at random.

A trial consisted of a separate presentation of
each of the eight slides. The intertrial interval was AS
seconds and the interstimulus interval Was approximately
7 seconds. The stimulus duration was 1 second. For each
of the two groups there was a different order for trials

l-3, i.e. for Group I there were three different orders

but they all had the extreme stimuli appearing very early
in the order and the central stimuli appearing very late
in the order. The orders for the first two trials with-
in each group were complements of each other. Therefore
the difficulty of each would be expected to be the same.
This was done so that the hypothesized differences between
Groups I and II could not be attributed to a difference in
order difficulty. Group II was presented orders which were
exactly the same (on trials 1-3) as for Group I except
these orders were inverted. On trials four and five both
groups were presented with stimuli in the same order.
These orders were ones which had been constructed in such
a way that they were equallycfissimilar to the orders which
had been presented to Groups I and II on the first three
trials.

The method of constructing the orders for trials
four and five was such: the stimuli were divided into
two groups — extreme (I, 2, 7, 8) and central (3, A, 5, 6).
These two groups were subdivided into: a) extreme extremes
(BE) (1, 8), b) central extremes (CE) (2, 7), c) extreme
centrals (EC) (3, 6) and d) central centrals (CC) (h, 5).
Fbr each half of the orders there was a stimulus from each
of the four groups. Furthermore the orders were counter-
balanced in such a way that the first stimulus presented
was in a class complementary to the class of the eighth

stimulus presented. The class of the second stimulus was

lO

complementary to the class of the seventh stimulus. The
class of the third stimulus was complementary to the class
of the sixth stimulus presented. The class of the fourth
stimulus presented was complementary to the class of the
fifth stimulus presented. Classes CC and EB were consid-
ered complements of each other, as were classes CE and EC.
Thus the performance of the two groups could be compared
on orders which were equally dissimilar to those presented
to either group on trials 1-3.

In order that the subjects would judge the slides
by the density of the dots rather than by other character-
istics (e.g. "that one had three dots in the first 'row'")
the slides were not shown in the same position twice.
Rather they were rotated 900 (clockwise) after each show-
ing for trials l-h. Each slide was projected backwards
on the fifth trial.

Subjects were run in groups of five at a time. The
following instructions were given to the subjects:

You will be shown a number of slides one at a
time. Each slide will have dots on it. There are
eight different slides (i.e. each slide has a dif—
ferent number of dots on it). Some will have many
dots, others will have only a few dots. You are
to assign a number from 1-8 to each slide. Assign
the number "one" to the slide which you think has
the least number of dots on it, the number "two"
to the slide you think has the second lowest num-
ber of dots on it. The number "eight" should be
assigned to the slide which you think has the
greatest number of dots, the number "seven" to the
slide you think has the second highest number of

dots on it. Middle numbers such as "four" and
"five" should be assigned to the slides which you

11

feel have a medium number of dots in compari-
son to the other slides.

Obviously you will have to guess on the
first few slides since you have no prior know-
ledge of the slides. Even if you must guess,
assign a number to every slide. Do not change
your answer after the next slide hasfi been pre-
sented. If, for example, you have used the
number "5" and are presented with a slide which
you think is the one which really should have
been judged as a "5" write the number "5" again
but do not change your previous judgment.

Before I present each slide I will say "ready"
so that you can focus on the wall and will not be
caught unawares by the slide. Please do not talk
or look at your neighbor's data sheet during the
entire experiment or rest periods.

After each trial subjects were told: "You have
completed trial A (etc.), if you judged each slide cor-

rectly you have used each number from 1-8."

RESULTS

When the slides of dots were constructed it was
assumed that increasing the number of dots logarithmic-
ally would yield approximate linearity of response. A
pilot study by the investigator and the work of Reese
(1953) also suggested this. A scatter plot was made of
the responses to the eight stimuli for all 50 subjects
on trial E. Figure 1 shows the scatter plot. By look-
ing at the scatter plot one can see that approximate
linearity of response was obtained.

Error Analysis: The data for each subject was

 

treated in the following manner. The absolute value of
each error was determined, e.g. if the S responded with

a five to stimulus seven then the absolute value of error
was [5-7| =2. Then these were added to get the total
error for each subject on each trial.

The analysis of variance was computed to see if
an interaction of orders (groups) by trials was present.
If both hypotheses are correct such an interaction should
be present. The results are shown in Table l. The in-
teraction of orders (groups) by trials was significant

beyond the .001 level (F=32.2h5 df.=5, and 191).

12

RESPOKSBS

13

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

[
21431.
5 9 2616
h- 26 ll 10
3 2161022
5 ll 15 7 A
358261022
2 19 18 10 h
l 26 16 l l
l 2 3 A 5 6 7 8
STIMULI
Figure l. Scatter plot of the responses to

the 8 stimuli for all 50 5'3 on
trial E.

 

14

 

 

 

 

 

 

 

 

 

Table 1. Analysis of Variance of the absolute Error of
Subjects

Source of Variation S.S. df MS F
Orders (Groups) 28.900 1 28.900 3.715
85 within orders 373.376 #8 7.779
Trials 798.976 h 199.7hh h1.312*
Orders x trials 779.520 5 155.90h 32.2h5*
Trials x 33
within orders 923.50h 191 h.835

Total 290h.276 249

.001 level

* significant at the

15

Table 2 lists the mean errors per subject for both

groups on the first five trials.

Table 2. Mean absolute Error per Subject per Trial

 

Trials
A B C D E
Group I 13.hh 3.88 h.08 5.72 5.h0
Group II 5.92 5.6a 6.32 5.80 5.44

 

In order to analyze trials B and C, and trials D and
E together, these trials were combined by adding the absolute
error of each S together. The purpose in combining these
trials was to reduce the number of comparisons between the
two groups. In reducing the number of comparisons the prob-
ability of obtaining significant results by chance is reduced.
The results are shown in Table 3.

Group II had a significantly less total absolute
error than Group I on trial A. Group I had a significantly
less total of absolute error than Group II for trials B and
C combined. There was no significant difference in absolute

error between Groups I and II on trials D and E combined.

Table 3. Kean absolute Error per Subject after Trials
were combined

 

Trials
A B+C D+E
Group I 13.4h 7.96 11.12
Group II 5.92 11.96 11.24
t 9.82 -3.81 -O.l2
p* .001 .001 n.s.

 

 

 

 

 

* 2—tailed t—tests were performed

Figure 2 shows graphically the difference between
the groups on the first five trials.

Correlation Analysis: Correlation coefficients
were computed between stimulus and response for both
groups on each of the first five trials. The results are
shown in table A. Because the correlations are represent-
ative of group performance rather than individual perfor-
mance, no tests of significance were performed. The
differences are the same as in error analysis. Group II
had a higher correlation between stimulus and response on
trial A than Group I. On trials B and C Group I had a
higher correlation between stimulus and response than
Group II. On trials D and E the correlations were approxi-

mately the same for both groups.

17

Figure 2. Mean absolute Error per Subject per Trial

1h

12

10

Error

A

x;__x Group I
o___o Group II

 

 

Trials

18

Table A. Correlation between Stimulus and Response for
each Group on Trials A-E

 

Trials
A B C D E
Group I .559 .934 .931 .914 .91h
Group II .761 .925 .887 .903 .915

 

Information Analysis: The Shannon-Wiener measure

 

of information (Attneave, 1959) was used to compute the
amount of information transmitted from stimulus to response,
on trials- B and C, and trials D and E combined for each
subject in each group. The amount of information trans-
mitted was not computed on trial A because in using in-
formation theory one assumes that the subjects have learned
the scale of judgment fairly well. Since trial A is mainly
a measure of performance and not of learning, to use infor-
mation theory here would violate the above assumption and
would probably lead to spuriously high estimates of
information transmitted. Several studies report that in
learning scales of absolute judgment there is a rapid
movement toward a stabilized scale of judgment (Tresselt
and Volkman, 19h3; Rambo and Johnson, 1964); and Rinturn
and Reese (1951) found that reduction in error of judgment
occurred very suddenly, as early as the latter part of the
first trial. Thus it would seem that using information

theory as early as trial B does not violate the assumption

19

that the subjects have learned the scale of judgment
fairly well. Also the high correlation between stimulus
and response for each group as a whole for trials B-E
seem to indicate that subjects had learned the scale of
judgment sufficiently well so that information-theory
measures could be applied to these trials. The average
amount of information transmitted per trial for trials B

and C, and D and E combined is shown in Table 5. It

TABLE 5

Information transmitted per trial (in bits)

 

Trials
B and C D and E
Group I 2.444 2.300
Group II 2.209 2.304
p* .001 n.s.

 

* two-tailed tests were performed

should be noted that the maximum amount of information
transmission possible is three bits.

The amount of information transmitted was signif-
icantly higher for Group I than for Group II on trials B
and C. No significant difference in amount of information
transmitted was found between Groups I and II on trials D
and E.

Analysis of Extremity of Responses: The responses
to the first four stimuli presented in trial A were ana-

lyzed as to whether they were central responses (3, 4, 5, 6)

20 '

or extreme responses (1, 2, 7, 8). For the first four
stimuli presented Group I responded with 76, 84, 60,

and 44 per cent central numbers, respectively. Group II
responded with 76, 96, 80, and 84 per cent central num-
bers respectively. Therefore both groups used central
responses more frequently then extreme responses in judg-

ing the first few stimuli presented on trial A.

DISCUSSION

Before discussing the results a remark or two
should be made about the effectiveness of the three types
of analyses used. The error analysis is the most power-
ful measure used. The correlational analysis is less
powerful than the error analysis because it was done for
each group as a whole rather than for each individual
subject. It was felt that a correlational analysis for
each subject would yield the same results as the error
analysis, therefore this analysis was performed for each
group rather than for subjects within groups. The infor-
mation analysis is not as strong as the error analysis
because it does not take into account the magnitude of
each error. One reason for using the information analy-
sis was because much of the literature on the learning
of scales of judgment presents its results in terms of
information transmitted. In respect to the purpose of
this study the results from the three types of analysis
are in agreement.

Group I had a significantly greater amount of
absolute error per subject than Group II on Trial A.

In looking at the correlation coefficients for the re-

lationship between stimulus and response it was found

21

22

that for trial A Group I had a correlation of .559 and
Group II had a correlation of .761. Both of the above
measures confirm the investigator's first hypothesis,
namely that Group II would outperform Group I on trial

A. The suggested reason for this is that subjects who
have no prior knowledge of the scale of judgment tend

to assign middle categories of response to the first few
stimuli presented. The analysis of extremity of response
indicated that subjects in both groups tended to assign
middle response categories to the first few stimuli pre-
sented. Thus, since for trial A Group II received an
order in which the middle stimuli were presented very
early in the order, they would have a greater probability
of choosing the correct response than would Group I who
were presented with an order in which the extreme stimuli
were presented very early in the order and the middle
stimuli very late in the order. Many subjects attempted
to use all eight response classes on each trial. If a
subject tended to use middle responses in the beginning
of the trial and attempted to use all eight responses, he
would have to use the extreme responses very late in the
trial. Thus this effect could last for the entire trial.
The investigator concluded that the superior performance
of Group II over Group I on trial A was due to the con-
struction of the different orders used for the two groups

on the first trial. This strengthens Wever and Zener's

23

(1928) suggestion that subjects who have no prior know-
ledge of the stimulus series tend to use the middle
response categories for their first few judgments.

An analysis of the amount of information trans-
mitted was not attempted for trial A because the use
of information theory assumes that the subjects have
learned the scale of judgment fairly well. In light of
the low correlations between stimulus and response on
trial A, for both groups, it was felt that this assump-
tion could not be met.

For trials B and C combined it was found that
Group I had significantly less absolute error per subject
than Group II. The correlation between stimulus and
response was .934 for trial B and .931 for trial 0, for
Group I, while it was .925 for trial B and .887 for trial
C, for Group II. The average amount of information trans-
mitted per subject was 2.444 bits per trial for Group I on
trials B and C, while it was only 2.209 bits per trial for
Group II. This difference was significant at the .001
level. (Note: it is entirely possible that if more stim-
uli were included in the stimulus series that the amount
of information transmitted would increase. Miller (1956)
has shown that the amount of information increases as the
number of stimuli increase and thalit levels off asymptot-
ically. In the present study it could be that more than

eight stimuli are necesany'for the asymptote to be reached.)

24

Thus those subjects who were presented with
orders where the extreme stimuli appeared very early
and where the central stimuli appeared very late, learned
the scale of absolute judgment faster than those subjects
who received orders where the central stimuli appeared
very early and the extreme stimuli appeared very late.
The orders which presented the extreme stimuli first re-
sulted in less absolute error, greater correlation
between stimulus and response, and in a greater amount
of information transmitted per trial than those orders
which presented the central stimuli very early in the trial.

0n trials D and E both groups were presented with
the same order of stimuli on trial D and the reverse of
that order on trial E. These orders were constructed in
such a way as to be equally dissimilar to the types of
orders which either group had received on the first three
trials. A t-test showed no significant difference between
Groups I and II in absolute error for trials D and E com-
bined. The correlations between stimulus and response for
trials D and E, were approximately the same for both
groups. There was no significant difference between groups
as to the amount of information transmitted over trials D
and E. The average amount of information transmitted for
Group I was 2.300 bits per trial and for group II was
2.304 bits per trial. Thus both groups performed approxi-

mately the same on trials D and E. This finding was

25

contradictory to expectation as it was expected that Group
I would trutperform Group II on trials D and E.

The reason for this last finding is unclear and
there are several possible explanations for it. The first
explanation is that subjects tend to become bored and fa-
tigued quite easily on such a task. A pilot study by the
investigator showed that when subjects became bored and
fatigued they tended to make more errors and to make
errors of greater magnitude. If the subjects became bored
or fatigued as early as trials D and E this might cause
poor performance in the judgment of the stimuli and the
subjects in both groups might be expected to perform at
the same level.

A second explanation is that in learning scales
of absolute judgment, subjects tend to become quite stab-
ilized in their judgments in just a few trials. After
this their performance is asymptotic. Since the subjects
in Group I learned the scale of judgment quicker than
those in Group II it is expected that they would reach
the peak of their performance earlier. If this peak came
at the end of trial C their performance would remain
stable while that of Group II would be approaching the
asymptote. One might point out however that the perfor-
mance of Group I on trials D and E decreased from that of
trials B and C and thus the above explanation would not

hold, but there is another factor which must be taken into

26

account...it is possible that the unbiased orders are
more difficult than the orders which Group I received on
trials A—C. If the unbiased orders were of greater dif-
ficulty then this could cause the decrease in their per-
formance, while an expected decrease in Group II's
performance might be offset by additional learning.

A third possible explanation is that the subjects
in Group I might have been learning the scale of absolute
judgment in the context of orders which presented the ex-
treme stimuli early and the central stimuli late in the
order, while the subjects in Group II were only learning
the scale of absolute judgment in the context of orders
which presented the central early and the extreme stimuli
late in the order.) Thus while orders which have the ex-
treme stimuli early and the central stimuli late produce
greater learning than orders which have the central stim-
uli early and the extreme stimuli late, when a subject
switches from either of these orders to an unbiased order
the amount of transfer to that particular unbiased order
may be the same for both groups. If so, this would ex-
plain the failure to find a difference between Groups I
and II for trials D and E.

It is concluded that when individuals are given
a task in absolute judgment, orders which present the
extreme (and also salient) stimuli early in the order and

the central stimuli late in the order will lead to greater

27

learning of the frame of reference than orders which pre-
sent the central stimuli early in the order and the extreme
stimuli late in the order, although the amount of transfer
from both orders to performance on the unbiased orders is
approximately the same. The above result confirms Wever
and Zener's (1928) expectation that orders which present
the end stimuli result in greater learning.

It is suggested that orders which present the ex-
treme stimuli very early result in greater learning because
the subject is provided with immediate knowledge of the
stimulus range. Once the subject has been presented with
the end stimuli, he uses these stimuli to judge the other
stimuli which are presented. Thus subjects who are given
such orders may be giving themselves "knowledge of results"
on each trial while subjects who are presented with the
central stimuli early in the order and the extreme stimuli
late in the order may be unable to give themselves "know-

ledge of results" throughout the trial.

SUMMARY

Although some work has been done in the area
of learning a frame of reference or scale of absolute
judgment, little has been done pertaining to the effect
which different orders of stimulus presentation have on
such learning. Never and Zener (1928) suggested that
when a subject has no previous knowledge of the stimulus
series he tends to respond with middle categories‘to the
first few stimuli presented. They state that once the
subject has knowledge of the end stimuli he uses these
stimuli to judge the other stimuli which are presented.
They suggest that orders which present the end stimuli
very early will lead to faster learning than orders which
do not present the end stimuli very early.

The present study investigated two hypotheses
concerning two different types of orders, type I orders
in which the extreme stimuli were presented very early
in the order and the central stimuli very late in the
order, and type II orders in which the central stimuli
were presented very early and the extreme stimuli were
presented very late. The first hypothesis was that sub-

jects who received type II orders would outperform subjects

28

29

who received type I orders on the first trial because all
subjects would tend to use middle numbers for their first
few responses since they had no previous knowledge of the
stimuli. The second hypothesis was that for the followe
ing trials those who had received type I orders would be
more accurate in their judgment than those who had received
type II orders because those who received type I orders
would have immediate knowledge of the stimulus range and
could use the end stimuli to judge the rest of the stimuli
presented in that trial.

Fifty subjects were randomly divided into two
groups of 25 each. Stimuli consisted of eight slides
'with dots on them. Each slide had a different number of
dots on it. Subjects were required to assign.numbers from
one to eight to each slide (assigning the number "one” to
the slide which they thought had the least number of dots
and the number "eight" to the slide which they thought had
the greatest number of dots). Subjects in Group I were
presented with type I orders for the first three trials
and subjects in Group II were presented with type II or-
ders on the first three trials. For the fourth and fifth
trials both groups-were presented with the same orders
which were equally dissimilar to the type I and the type
11 orders.

Group II made less absolute error than Group I

30

on trial A supporting Never and Zener's suggestion that
subjects tend to use middle categories as responses to
the first few sthmuli presented when they have no prior
knowledge of the stimulus series. On trials B and C
Group I made less absolute error than Group II, also the
amount of information transmitted was greater fer Group
I. On trials D and E there was no difference between
Groups I and II in absolute error or in amount of infor-
mation transmitted. It was concluded that orders which
present the extreme stimuli very early and the central
stimuli very late lead to greater accuracy in judgment
than orders which present the central stimuli very early
and the extreme stimuli very late. However the amount
of learning transferred to unbiased orders is approximately
the same from both types of orders. This confirms Wever
and Zener's (1928) suggestion that orders which present
the end stimuli at the beginning of the order lead to
quicker learning of the scale of judgment.

REFERENCES

Attneave, F. Applications of Information Theory to
Psychology. Bolt, New York, 1959.

Eriksen, C. W. Effects of practice with or without
correction on discrimination learning. Amer.
J. Psychol. 1958, 21, 350-358.

Eriksen, C. W. and Hake, H. W. Absolute judgments as
a function of stimulus range and number of
stimulus and response categories. J. Exp.
Psychol. 1955a, 32, 323-332.

Eriksen, C. W. and Hake, H. W. Multidimensional stim-
ulus differences and accuracy of discrimination.
J. Exp. Psychol. 1955b, 59, 153-160.

Eriksen, C. N. and Hake, H. W. Anchor effects in absolute
judgments. J. Exp. Psychol. 1957, 53) 132-138.

Garner, N. R. An informational analysis of absolute
judgments of loudness. J. Exp. Psychol. 1953,
Aé. 373-380.

Gibson, E. J. Improvement in perceptual judgments as a
function of controlled practice or training.
Psychol. Bull., 19532.29: 401-431.

Johnson, D. M. A systematic treatment of judgment. Psychol.
Bull., 1945, 4g, 193-224.

Johnson, D. M. and King, C. R. Systematic study of end
anchoring and central tendancy of judgment. J.
Exp. Psychol., 1964, 61, 501-506.

Killer, G. A. The magical number seven, plus or minus two.
Psychol. Rev., 1956, 63) 81—97.

Kinturn, A. L. and Reese, T. W. The effect of differential
reinforcement on the discrimination of visual num-
ber. J. Psychol., 1951, 31, 201-231.

Hurphy, L. Absolute gudgments of duration. J. Exp.
Psychol., 19 6, 21, 260-263.

31

32

Parducci, A. Incidental learning of stimulus frequencies
in the establishment of judgment scales. J. Exp.
Psychol., 1956, 52, llZ-ll8.

Rambo, U. and Johnson, E. Practice-effects and the esti-
mation of adaptation level. Amer. J. Psychol.

1964, _7_z, 106-110.

Reese, E. P. Psycholphysical research summary report
1946-1952. USN Spec. Dev. Ctr. Tech. Reot. No.
SDC-lBl-l-5, Psychophysical Research Unit,

Kt. Holyoke College.

Tresselt, M. E. and Volkman, J. The production of uniform
opinion by non-social stimulation. J. abnorm.

soc. Psychol., 1942, 21, 23L-2h3.

Wedell, C. H. The nature of the absolute judgment of
pitch. J. Exp. Psychol., l93h, l2, 485-503.

Wever, E. G. and Zener, K. E. The method of absolute judg-
ment and psychophysics. Psychol. Rev., 1928, 25J
héé—Ag3o

APPENDIX A

ORDERS OF STIMULUS PRESENTATION

GROUP

Trials

151-2

’EOUP II

1

G

Trials

do

do

do

'1 H'GAN STATE L‘WERSITY LIB?“

l ir'llll'l' II I” Hi I l‘

3003175