W

HIWIIHIH

IHI

HIMIHH

1
I
l

 

(a. gv\h.§¢‘me 2““. STIR?" {ti ENE) QNQHQ “Rikﬁ

‘T‘Eaesix $73? $423 taxes a-% M“ A

"9‘ a ﬁjd-zgts-‘fl’
*“5:5:H.éﬁik Suﬁ?“ :I'Nixtfhxi‘wi’s‘g

 

 
 

‘ LIBRARY
" Michigan State

University

   

 

 

 

 

 

 

 

ABSTRACT

A SYSTEMATIC STUDY OF
END ANCHORING

by Calvin Reginald King, Jr.

An experiment was designed to investigate and clarify
the concept of anchoring. Anchoring was defined as the
special way in which certain response categories appear
tied to certain stimuli.

Seven stimulus sets were selected from a systemati-
cally varied stimulus series. The complete stimulus series
was composed of thirty—seven photographic slides, each
containing thirty-six blue and green dots such that the
entire series varied systematically from an all-blue slide
to an all-green slide. This permitted the experimenter
to control the color mass of the stimuli selected for
presentation.

An eleven—point scale was used to express judgments.
Seventy subjects were recruited from introductory psychology
courses, and were randomly assigned to groups of five subjects.
Half of the subjects judged greenness, and half judged blue-

ness .

Calvin Reginald King, Jr.

The main argument of the experiment was that the
salience of a stimulus, rather than its placement in the
series. determined anchoring. A salient stimulus was
defined as a stimulus that is easily noticeable and readily
associated with a response category.

Other hypotheses were offered relating the effective-
ness of anchoring to the salience of the end stimuli, and
the influence of anchoring upon judgments of adjacent stimuli
to the salience of the end stimulus and to the distance of the
judged stimuli from the end stimulus.

The data were analyzed in terms of errors, an error
being defined as any incorrect association of a response
category to a stimulus. An overall analysis of Variance of
errors indicated that the data gathered under the two color
conditions could be combined into a single body of data.

An overview of the data was then presented graphically by
calculating the median and the interquartile range of judg-
ments for each stimulus in a set. This analysis indicated
that anchoring effects occurred as predicted.

Subsequently, an analysis of variance of errors for
each set was run to provide a more refined analysis of the
magnitude and extent of anchoring effects. Significant

F-ratios were obtained on all sets with the exception of the

Calvin Reginald King, Jr.

mid—range set where no anchoring was predicted to occur. The
F-ratios obtained decreased in significance as the salience of
the end stimulus decreased, supporting the hypotheses.

It was suggested that current interpretations of
anchoring would have to be revised to incorporate the concept
of salience. The most acceptable interpretation is the
subjective-standard hypothesis which embraces salience as the
key concept for explaining the selection of a stimulus as the
subjective standard. The data of this experiment indicated
that anchoring is not primarily a function of the placement of
a stimulus in a series. Rather, for a stimulus to act as an
anchor, it must be easily identified and associated with a

particular response category.

Approved ﬂw 77/; 1W

A SYSTEMATIC STUDY OF

END ANCHORING

BY

Calvin Reginald King, Jr.

A THESIS

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

MASTER OF ARTS

Department of Psychology

1963

_\ )

To Jeanne

ii

ACKNOWLEDGMENT

The author expresses his deepest appreciation to
Dr. Donald M. Johnson, chairman of his committee, for his
counsel and support, so vital to the completion of this
thesis.

The writer also extends his thanks to Dr. Paul
Bakan and to Dr. Robert E. McMichael for their helpful

advice and criticism.

iii

TABLE OF CONTENTS

INTRODUCTION

Stimulus Arrangements and Judgment Scales
Dependent Variables

Classification of Anchoring Effects
Problems and Criticisms

THE PROBLEM . . . . . . . . . . . . .
Hypotheses
METHOD . . . . . . . . . . . . . . . . . . . . .

Stimulus Series
Judgment Scale

Apparatus

Subjects

Procedure
RESULTS . . . . . . . . . . . . . . . . . . . . .
DISCUSSION . . . . . . . . . . . . . . . . . . .
SUMMARY . . . . . . . . . . . . . . . . . . . . .
BIBLIOGRAPHY . . . . . . . . . . . . . . . . . .

iv

Page

[—I

\.C>U'|-l>l--J

11
11
13
13
14
15
15
16
18
27
33

36

Table

LIST OF TABLES

Page
Overall analysis of variance of errors . . . . 19
Analysis of variance of errors for each set . . 22
Mean errors (2) for each slide and critical
difference values (CDV) for each set . . . 23

LIST OF FIGURES
Figure Page

1. The effects of end anchoring on the dispersion
of judgments of each slide in each set . . 20

Vi

LIST OF APPENDICES
Appendix Page

A. Judgment Recording Sheet . . . . . . . . . . 38

vii

INTRODUCTION

Stimulus Arrangements and Judgment Scales

In experiments involving the formation of a judgment
scale relative to a specified stimulus series, it has been
observed that the judgment scale is related to the stimuli
judged in a generalized way: it corresponds to the judged
stimuli in position and width. The judgment scale is said
to be stimulus-anchored (Rogers, 1941), which is merely
another way of describing the conformity of the scale t)
the stimulus series.

To be more specific, however, anchoring is considered
to be the descriptive term for the special way in which
certain response categories appear tied to certain stimuli
(Johnson, 1955). As such, anchoring would seem to be
intimately related to the experimental stimulus arrangement
presented to the subject. The following three modal
stimulus conditions have been described by Sherif and
Hovland (1961).

1. Formation of a scale on the basis of a well—graded

stimulus series having an explicit standard within it.

This arrangement is exemplified by experiments using the
method of constant stimuli, in which for each stimulus
presented for judgment, a comparison or standard is
simultaneously presented, usually with a value near the
middle stimulus of the series. The judgment scale formed
using this method is more stable and better fitted to
stimulus values near the value of the standard stimulus.
According to Sherif and Hovland, this means that judgments
are more accurate and less variable near the middle of the
series.

2. Formation of a scale on the basis of a well—graded
stimulus series without an explicit standard. This is
synonymous with the method of single stimuli, also known
as the method of absolute judgment, formally introduced into
psychophysics by Wever and Zener in 1928. This technique
consists of the single presentation of members of the
stimulus series with the requirement that each stimulus be
judged in absolute terms as it is presented. Through the
repeated presentation of the stimulUs series, the subject
(S) gains knowledge of that series so that he is able to
express an absolute judgment. The distribution of his
judgments is called the "absolute series" or "absolute

scale." When this method is used, it is the end stimuli

of the series which control oscillations in the judgment
scale, according to Volkmann (1951),while the middle stimuli
have no functional significance. As Needham (1935) stated
it, it is as though S "first 'learns' to recognize the
boundaries within which he is judging, and to assign to

these limiting stimuli the relatively more correct judgments"
(p. 282).

3. Formation of a scale without a graded stimulus series.
This arrangement is perhaps most pertinent to social situations
where social stimuli are complex and cannot easily be ordered
along a dimension. Such a judgment scale would be highly
influenced by internalized standards operating within s,
such as attitudes, and by the judgments of other persons
if estimates are given in the presence of other people.

Of the above three modal stimulus conditions, the
one most commonly thought of in reference to anchoring is
the method of single stimuli. In the first place, it
demonstrates the fact that a judgment scale can be formed
relative to a series of stimuli without the explicit
introduction of a standard. Secondly, it can show the
specific linkage between a response category and a stimulus
in the absence of a standard. Furthermore, as a psycho-

physical method, it is more flexible than the method of

constant stimuli: it permits greater manipulation of the
stimulus series and the response scale. And it may be
expected to yield more precise data than an ungraded series

of stimuli.

Dependent Variables

 

Operationally, anchoring is generally described in
terms of the effects of the independent variable on one or
more of the following dependent variables: (1) frequencies
of responses; (2) latencies of response; (3) variability of
response; (4) confidence attending responses: (5) frequency
of correct responses or the converse, frequency of errors.

Several writers (including Volkmann, 1951; Volkmann
and Engen, 1961) have noted that the measurement of these
dependent variables reveals the following effects, which
have been taken as indications of anchoring. (1) A set
of correlated changes in response—frequency known as a shift
of a judgment scale. (2) A selectively decreased latency
of responses. (3) A selectively decreased variability of
responses. (4) A selectively increased confidence attending
the judgment responses. (5) A selectively increased
frequency of correct responses, or conversely, a selectively

decreased frequency of errors.

Classification of Anchoring Effects

Experiments in anchoring specify the type of mani-
pulation of the stimuli used to obtain the desired effect.
Hence, it is feasible to classify anchoring in terms of these
experimental operations (see Johnson, 1955, and Volkmann
and Engen, 1961).

Anchoring effects may be correlated with specific

stimulus values. Here we may speak of sgpplied anchoring

 

and end-anchoring. In supplied anchoring, the anchoring
stimulus is added to the range of stimuli to be judged.
Its placement may be either within or outside the range
of the stimulus series. Perhaps the best illustration of

supplied anchoring comes from studies of assimilation and

 

contrast effects, a term introduced by Rogers (1941). In

a study using weights by Sherif, Taub, and Hovland (1958), it
was hypothesized that (l) the introduction of anchors at or
near the end points of the series will cause displacement

of the distribution of judgments in the direction of the
anchor (assimilation) and (2) the introduction of anchors

at increasing distances from the upper and lower ends of the
stimulus series will cause displacement of the distribution

of judgments away from the anchor and the judgment scale

will be constricted to fewer categories. The results of
the experiment supported the hypotheses.

End anchoring, perhaps the best known type of

 

anchoring, occurs as a function of both the range of stimuli
used (Volkmann, 1951) and the ease of identification of the
end points of the stimulus series (Johnson, 1955). Volkmann
(1951) reports on an experiment in which triangular stimuli
varied only in area. The smallest triangle was to be judged
as "l," and the largest triangle as "7." The results indi-
cated a higher frequency of errors in identifying the middle
stimuli. Hence it is the end stimuli which are responsible
for controlling the oscillations of the judgment scale.

Natural anchoring effects occur when the intrinsic

 

character of a stimulus ties down a judgment category,
independent of the specific stimulus series or the examiner's
(E) instructions. A good example of a natural anchor in
judging the inclination of lines is the vertical and the
horizontal. Onley and Volkmann (1958) in a study of perpen-
dicularity, had Ss adjust two straight lines to be mutually
perpendicular in the absence of visual cues of the horizontal
or vertical. They report that the subjective horizontal

and vertical axes play a conspicuous role when S is adjusting

two lines to be perpendicular. As the reference line

increased in angular displacement, variability in perpendi-
cularity increased.

An infrequent effect is derived anchoring. An anchor

 

is derived from a discriminable aspect of the stimulus series
and may represent a subjective act of the S. In the Onley
and Volkmann study cited above, 83 also used a secondary set
of reference axes corresponding to subjective coordinates
tilted at 450 to the normal or primary axes. Volkmann and
Engen report a study conducted by E. P. Reese (and others)
in which ananchoring effect was found at the 450 angle
as well as at the horizontal and the vertical. They believe
that the S is subjectively bisecting the angle between 00
and 900.

Anchoring may also occur via instruction without its
being present in the stimulus series. Volkmann (1936)
had his 85 first judge inclinations of lines without any
anchor. He then told them to call the horizontal a "1"
even though it was not presented. Anchoring then occurred.

This is an example of the operation of an internal anchor.

 

With less methodological precision, the term "internal

anchor has also been used in studies of social judgment
to describe the operation of an individual's attitudes and

opinions. In this context, the attitude or opinion held by

an individual in reference to a particular issue is said to
anchor his judgments relative to that issue. This use of

the term "internal anchor" is similar, if not analogous,

to the concept of an internal frame of reference. Generally,
in studies using this meaning of internal anchoring, the S's
stand on an issue is determined and his placement of perti—
nent items on a scale is observed. For example, in a study
by Secord, Bevan and Katz (1956), Ss were shown photographs

of unknown Negroes and Caucasians, and were asked to judge

the photographs on the basis of ten physiognomic traits
associated with "Negroidness" and fifteen personality at—
tributes commonly associated with stereotypes of Negroes.

It was found that 83' stands toward Negroes anchored their
judgments. The more anti—Negro the judge, the more blantantly
he attributed all physiognomic and stereotyped characteristics
to a Negro, regardless of how Caucasian-like he appeared.

A novel phenomenon is that of response—anchoring.

 

In an experiment conducted by Eriksen and Hake (1957), a
circular hue continuum was used which avoided end-anchoring
effects. Ss were required to name each hue as it was pre—
sented with a given number from 1 to 20. They found that 85
selected numbers at the ends of the response continuum as

anchoring agents, regardless of the stimuli to which they

were attached. This suggests that responses on the ends
of an ordered set of responses can produce anchoring, just
as end-stimuli produce anchoring.

Another rare type of anchoring occurs when there is
a high degree of compatibility between a stimulus and a
response. Pitts and Deininger (1954) found anchoring when
a high degree of similarity exists between a stimulus and
a response, such as attaching the response nonsense syllable
YEL to the color yellow in judging hues. This seems to be
an artifact of the experiment rather than a true case of

anchoring.

Problems and Criticisms

 

Few studies offer any explicit reasons for the observed
anchoring effects. While some interpretive hypotheses have
been offered, and these will be presented in a later section,
most fail to specify what qualities a stimulus possesses
that cause judgments made in relation to it to be more
accurate and less variable than others. In this thesis,
the author will speak of the salience of a stimulus to
mean that it is more noticeable than others. A stimulus
that is salient is easily associated with a response.

That the salience of a stimulus can be defined leads

10

to another criticism of experiments: many fail to construct
a stimulus series that varies systematically and quantifiably
among its elements. Hence, the possible influence of an
unsystematically varied stimulus series upon the results
cannot be ruled out. Furthermore, the series selected for
study rarely contains all possible stimuli of the physical
property being judged; that is, the stimulus series is some-
what arbitrarily defined, as opposed to being bounded as a
natural consequence of its construction.

It follows that if the salience of the stimuli can
be quantified and systematically varied, then anchoring may
be studied under different conditions of stimulus salience,
and the influence of a salient anchor upon other stimuli
in the series can be observed. Ideally then, the stimulus
series should vary in only one parameter: hence, observed
effects of the stimulus condition upon judgments can be
attributed to that experimentally controlled variation. A
problem will now be posed that attempts to attribute anchor—
ing to the salience of the stimuli using a stimulus series
that is quantifiable and systematically varied. The argument
will be introduced that the end of a stimulus series is an

anchor only when the end stimulus is salient.

THE PROBLEM

The purpose of this study is to investigate end
anchoring effects in stimulus sets selected from a para-
metrically lawful stimulus series, using consecutive inter-
vals to express judgment. The stimulus series was constructed
in such a manner that the physical, discriminable aspect,

color mass, varied systematically from stimulus to stimulus

 

along a continuum of blue and green. One of the end stimuli
contained the maximal amount of blue mass, while the other
end stimulus contained the maximal amount of green mass.

Any number of stimulus sets could be selected
from the series such that color mass would be controlled by
the E. Hence, it was possible to study anchoring when the

end stimuli possesses varying degrees of salience.

Hypotheses

 

It was hypothesized that:

l. Anchoring effects will be maximal in those
stimulus sets which include either of the end stimuli
of the complete series since these are the most
salient stimuli.

11

12

The effectiveness of the end stimuli as anchors
will decrease as their salience decreases.

The anchoring effect will influence judgments of
adjacent stimuli, with the influence decreasing
as the distance of the judged stimuli from the
end stimuli increases.

The influence of the anchor upon judgments of
adjacent stimuli will decrease as the salience of
the anchoring stimuli decreases.

Overall, it is the salience of the end stimulus
which determines anchoring, not merely its position

at the end of the stimulus set.

METHOD

Stimulus-Series

 

The complete stimulus series was composed of thirty-
seven 1" x l" photographic slides. Each slide was constructed
of blue and green dots arranged in a naively random pattern,
with a total of thirty-six dots on each slide. The ratio of
blue to green dots was manipulated in such a way that a
continuous series of stimuli was formed running from all blue
dots to all green dots. That is, slide #0 contained no
green dots and 36 blue dots: slide #1 contained 1 green dot
and 35 blue dots; the middle slide, #18, contained 18 green
dots and 18 blue dots: and so forth, with slide #35 con-
taining 35 green dots and 1 blue dot and slide #36 containing
36 green dots and no blue dots. Hence, the slides were
coded in terms of the number of green dots. The code number
subtracted from 36 gives the number of blue dots. This
stimulus construction followed Philip (1947).

The advantages of such a stimulus series are readily
apparent: (l) The stimuli differ only in the desired para-

meters, the color mass of blue and green. This difference

13

14

is systematic and easily controlled throughout the series.

(2) It contains two salient stimuli (the monochromatic slides)
as anchoring stimuli. (3) It contains gradations in stimuli
such that anchoring effects could be observed in the absence
of the monochromatic slides. (4) The range of stimuli
represented allows for the selection of sections of the
stimulus series for presentation in such a fashion that the
salience of the end stimulus can be controlled. For purposes
of this research, the stimulus series was divided into seven

stimulus sets (see Table 2), each composed of eleven conse-

 

cutive slides. This selection permits the observation of
anchoring effects when the salience of the anchor decreases
while the number of stimuli in the series is constant.

The middle set was included to determine if anchoring would
occur when the stimulus set does not contain any salient

stimuli.

Judgment Scale

An eleven—point or eleven—category judgment scale
was selected ranging from O to 10, one category for each
judged stimulus. The stimulus property to be judged was,
of course, stimulus color. Ss were asked to judge the

"blueness" or "greenness" of each slide presented, expressing

15

their judgments numerically. The "bluer" or "greener" the
stimulus, the higher the S's numerical response: hence, the
bluest or greenest slide would be called a 10, and the least
blue or the least green slide would be called a 0. Half

of the $5 judged "blueness,' and half judged "greenness,"

hence there were two judgment conditions.

Apparatus

 

The apparatus used included a Viewlex manually-
operated slide projector equipped with a Alphax variable speed,
manually—operated tachistoscopic lens shutter. The shutter
was set for a one-second presentation. With the aid of an
assistant an inter-trial interval of approximately eight
seconds was maintained.

The projection screen was placed at approximately
20 feet from the projector on each trial so that the size of
the image remained constant. Although the viewing room was
darkened, enough light remained to permit Ss to record their

judgments.

Subjects

Seventy 83 were recruited from elementary psychology
courses. Each S was randomly assigned to one of the fourteen

experimental conditions (seven stimulus sets judged under

16

two color conditions); thus there were five 58 in each
condition. All five 88 were run simultaneously. The only
criterion for recruitment was a verbal request that no

color-blind students participate.

Procedure

 

Each S was given a mimeographed instruction and
recording sheet. E read the instructions aloud to S, and
then answered any questions that were raised. The stimulus
set selected for presentation was then presented eight
times. The slides were randomized for each presentation:
orientation of the slide was changed at random for each
presentation. The latter technique was introduced to
minimize any extraneous cues that might appear in a slide,
particularly any intrinsic pattern among the dots. There
were eight possible orientations: forwards and backwards,
and in either of those orientations the slide could be
presented up, down, side-left, or side-right.

Following each of the first two presentations, E
asked if there were any questions. Thereafter, no assistance
was given. The first two presentations were considered to
be practice trials and are, therefore, not included in the

analysis. It was assumed that at least two trials would be

17

necessary for S to establish some knowledge of what was
expected of him and of the series of stimuli before him.
An inspection of the data bears out this assumption.

Each stimulus was presented for one second, followed
by an eight-second inter—trial interval. Approximately one
minute intervened between set presentations. The whole
procedure took about 40 minutes.

Discounting the first two presentations of the
selected set, the amount of data gathered was as follows:
each stimulus in a set was presented six times for judgment
by five 85. Hence, thirty judgments are available for

each stimulus in a set.

RESULTS

It was first determined whether or not the data
gathered under the two color conditions could be combined
into a single body of data. This was desired because
(1) it would double the amount of data, and (2) no hypotheses
were concerned with the specific color being judged. A gross
determination of the feasibility of combining data was
obtained by running an overall analysis of variance of errors.
Error, the dependent variable in this as well as all subse-
quent analyses of variance, is defined as any incorrect
association of a response category to a stimulus. Since
there was a judgment category for each stimulus in a set,
errors could quickly be ascertained merely by subtracting
the number of correct judgments from sixty, the total number
of judgments given for any stimulus.

The results of the overall analysis of variance of
errors are presented in Table 1. No significant difference
exists between the two color conditions. Consequently, the
data will be combined in all remaining analyses.

However, the differences among sets are highly

significant. This offers a gross, but nonetheless important

18

19

contribution to an understanding of anchoring: manipulation
of the salience of the end anchor of a set does influence

the number of errors made in judging stimuli within a set.

 

 

 

Table 1. Overall analysis of variance of errors.

Source df 33 ms F
Set 6 4215.03 702.51 18.99*
Color 1 14.34 14.34 0.39
Set x Col 6 390.34 65.06 1.76
Within 140 5176.95 36.98

Total 153 9796.66

 

*Significant at .01 level.

The first statistical manipulation of the combined
data was the calculation of medians and interquartile ranges
for judgments of each stimulus of each set. These values,
when plotted for judgments of each stimulus, provide a
graphical overview of the data. The values were calculated
by formulae set forth by Edwards (1958). Medians and inter-
quartile ranges were used to give a better graphical
description of the distribution of judgments since the latter
were skewed at the anchored ends of the absolute scale. These
medians and interquartile ranges are presented in Figure 1,

with one graph for each combined stimulus set.

20

.ommum>mu mm3 mm mnu mam: Mom mason ucmﬁmcsn on» mmmomssm

Hmownmmum mom mmsoum mcﬂnﬁoo OB .ucmﬁmbsn uomuuoo mpcmmmummu maﬁa

HmcomMHo one .nmn mmouu wnu an pmumowncﬂ we assume one .mGHH amoeunm>

mnu an pmuwuaccw ma mucmﬁmosn mo unmohwm human macmﬁﬁ was .umm sumo CH
ocean sumo mo musmﬁmnsn mo soamuommﬂo may no mcﬂuososm cam mo muommmm 039 .H musmﬂm

MQJM uojm Uij unjm
w»....n.«w.mn...«+o.a mm an .n on 5 mm 1n «n on s. 3 ..M an a I 5 m. m.

d I d d d

       

     

1 w 4 ‘ 1‘ d‘J q 1

   

x. hum w ham m ham

1 4 L L 1 L4

lNZINOOnf‘

1 L4 L A A A l A A L

 

 

 

 

 

«Lenora :.v>m0.. Emoruo

‘ d‘ I d 1 1 I d I ‘ 1

0 Fun .

t”

 

.LNE‘A‘SCIHP
at
9

 

 

 

 

21

This gross overview of the data indicates that
anchoring occurred as predicted. The judgment scale is
anchored to the salient stimulus in each set, with the
exception of the middle set which did not contain a salient
stimulus. The anchoring stimulus also exerts an influence
upon the variability of judgments in adjacent categories,
this influence being determined primarily by the salience
of the anchoring stimulus and the distance of the judged
stimulus from the anchor. Furthermore, it is not the end
stimuli per se which anchor the judgment scale; had this
been true, anchoring would have occurred at both ends of
the sets and in the middle set. The salience of the stimulus
appears to be the primary factor in anchoring.

The combined data were then analyzed as follows:
each set was subjected to a Slides X Subjects Analysis of
Variance (after Lindquist's Treatment X Subjects Design).
The numerator of the F ratio was the mean square for slides
and the denominator was the interaction mean square. The
results are presented in Table 2. The mean errors for
these data are shown in Table 3, which includes the critical
difference values for making multiple comparisons among these

means within sets (after Lindquist, p. 166).

22

 

 

 

Table 2. Analysis of variance of errors for each set.
Set Source df 38 ms
Set 1: Slides 10 298.76 29.876 22.858*
Slides Subjects 9 82.37 9.152
0-10 81 x Subj 29 117.63 .1.307
Total 109 498.76
Set 2: Slides 10 212.16 21.216 10.791*
Slides Subjects 9 46.85 5.206
1—11 81 x Subj _29 176.95 1.966
Total 109 435.96
Set 3: Slides 10 43.07 4.307 3.059*
Slides Subjects 9 225.04 2.782
2-12 81 x Subj ‘29 126.76 1.408
Total 109 194.87
Set 4: Slides 10 7.97 0.797 1.014 ns.
Slides Subjects 9 8.04 0.893
13,23 81 x Subj 99_ 70.72 0,786
Total 109 86.73
Set 5: Slides 10 42.22 4.222 3.707*
Slides Subjects 9 21.48 2.387
24-34 81 x Subj 90 102.52 1.139
Total 109 166.22
Set 6: Slides 10 79.10 7.910 4.287*
Slides Subjects 9 125.99 13.998
25-35 81 x Subj 29 166.01 1.845
Total 109 401.35
Set 7: Slides 10 220.05 22.005 20.799*
Slides Subjects 9 86.07 9.563
26-36 81 x Subj .99 95.23 1.058
Total 109 401.35

 

*Significant at .01 level.

23

Table 3. Mean errors (i) for each slide and critical

difference values (CDV) for each set.

 

 

 

 

Set Slides
1: 0 1 2 3 4 5 6 7 8 9 10 cnv
i. 0.2 1.1 1.6 1.8 3.4 3.8 4.7 4.8 4.4 5.1 4.5 1.35
2: 1 2 3 4 5 6 7 8 9 10 11 cnv
i: 0.4 1.7 2.9 3.5 4.0 3.7 4.7 4.2 4.8 5.1 4.8 1.66
3: 2 3 4 5 6 7 8 9 10 11 12 cnv
i: 3.0 4.4 4.6 4.7 4.9 5.1 4.7 5.2 5.0 5.2 5.5 1.40
4: 13 14 15 16 17 18 19 20 21 22 23 CDV
x: 5.2 5.0 4.9 4.7 5.3 5.4 5.0 4.6 4.9 5.0 4.5 1.05
5: 24 25 26 27 28 29 30 31 32 33 34 CDV'
i: 4.6 5.0 5.1 4.9 5.4 4.8 4.6 5.1 4.8 4.1 3.0 1.26
6: 25 26 27 28 29 30 31 32 33 34 35 cnv
i: 4.8 5.1 5.1 4.4 4.1 4.6 4.0 4.1 3.6 3.0 2.2 1.60
7: 26 27 28 29 30 31 32 33 34 35 36 cnv
i: 2.4 3.9 3.3 3.5 2.7 1.6 0.9 0.6 0.2 0.0 0.0 1.21
Several conclusions, supportive of the hypotheses, can
be drawn from an inspection of the data. The dependent variable
in all cases is errors. Anchoring, or the influence of it,
is thus indicated by a selective decrease in the frequency
of errors made in judgment.

Anchoring is maximal in the two sets containing the

end stimuli of the complete series. Errors are fewest in

24

judging those two stimuli. Using the critical difference
values in Table 3, the magnitude of difference in errors

of judgment becomes significant at a point two slides re—
moved from the most salient stimulus in Set 1, and five
slides removed from the most salient stimulus in Set 7. The
first hypothesis is supported.

As the salience, measured in color mass, of the end
stimuli decreases, so does anchoring. That is, slide #1,
the most salient stimulus in Set 2 is less effective in
terms of errors than in slide #0, the most salient stimulus
of Set 1. Slide #2, the most salient stimulus in Set 3, is
less effective than either of the above anchors. A comparison
of the most salient stimuli in Sets 5-7, in which green mass
predominates, reveals similar findings. Hence, the effective—
ness of the end stimuli as anchors, in terms of errors,
decreases as their salience decreases. This supports the
second hypothesis.

The third hypothesis is concerned with the effect
anchoring has upon adjacent stimuli in the set. As the
distance of the judged stimulus from the anchor increases,
the number of errors in judging that stimulus will increase.
This hypothesis is supported. An inspection of Table 3

indicates that errors increase as the distance of the judged

25

stimulus from the anchor increases. At some point in
reference to distance from the anchor, the number of errors
becomes significantly greater than those made in judging
the anchoring stimulus. The only set in which this does
not hold true is the middle Set 4, in which there is no
salient stimulus, and consequently, no anchoring.

The above hypothesis is intimately related to the
fourth hypothesis which predicts that as the salience of
the anchoring stimulus decreases, the number of errors
occurring in judgments of adjacent stimuli increases. The
data support this hypothesis. Not only do errors increase
in judgments of adjacent stimuli as the salience of the
anchor decreases: in the middle Set 4, in which no
salient stimulus is present to anchor the judgment scale,
errors are fairly evenly distributed across judgment cate—
gories and no significant differences exist among judgments
of the stimuli.

Finally, the data support the proposal that salience
determines anchoring. If anchoring occurred merely as a
function of the end stimuli, without consideration for the
salience of the end stimuli, then it should have occurred
at both ends of all sets with approximately equivalent

effectiveness. The data indicate, however, both that the

26

number of errors increase as the salience of the end stimuli
decreases, and that the significance of the differences in
errors of judgment decreases as salience decreases. The
middle Set 4 indicates that when there is little difference
among the salience of the stimuli in a set, no anchoring
will occur. These results indicate that the presence of
anchoring and its influence upon adjacent stimuli hinges

upon the salience of the stimuli.

DISCUSSION

While a considerable body of research literature
exists on anchoring, most of it is descriptive in content,
rather than interpretive. Hewever, three interpretive
hypotheses have been offered (Eriksen and Hake, 1957).

1. Response Attenuation Effect: This is considered to
be an artifact of the absolute method. It is based upon
the fact that errors in judgment of the end stimuli can
occur only unidirectionally, while errors in responding to
midrange stimuli can occur in two directions. Hewever, this
hypothesis is insufficient in itself, for the decrease in
judgment variability at the anchoring stimuli is greater
than that expected solely on the basis of unidirectional
errors.

2. Discrimination and Stimulus Generalization: This
interpretation assumes that for each value in the stimulus
series a relatively symmetrical generalization gradient is
set up in terms of the particular response that is assigned
to that stimulus. For the mid-range stimuli, a number of
overlapping generalization gradients would be formed.
Consequently, the presentation of a stimulus in the middle

27

28

of the series would tend to elicit several responses besides
the correct one. These competing responses would reduce

the habit-strength of the correct response. However, for
the end stimuli, the generalization gradients would be asym-
metrical with a reduction in competing responses, and hence,
fewer errors in judgment.

3. Subjective-Standard Hypothesis: This hypothesis is
based upon the observation that judgments are always made
relative to a standard or reference level that is subjectively
present. This subjective standard is derived from a few
selected stimuli that the S uses as standards for judging
the remaining stimuli. Accordingly, then, the S transforms
the task into a comparative judgment using the recalled value
of the selected stimulus as a comparison stimulus.

This hypothesis assumes that, in the case of end
anchoring, the S selects the end stimuli to use as his
subjective standards. He frequently tries to recall the
value of the end stimuli, and this increases the accuracy
of recognition of these stimuli when they do appear. Further-
more, the accuracy of recognition of the other stimuli is
directly related to the distance of these stimuli from the
end stimuli. This relationship exists because the effective-

ness of a standard is inversely related to the degree of

29

similarity between the standard stimulus and the comparison
stimulus.

Eriksen and Hake present a study which appears to
support the subjective-standard hypothesis. They constructed
a stimulus continuum of hues without any obvious end points.
There could be no response attenuation because the subject
could err in either direction. Stimulus generalization
gradients would be approximately symmetrical for all stimuli:
hence response competition would be about the same for all
stimuli. But 88 would still be expected to select a few
stimuli which they would use as standards to anchor their
judgments. The data of the experiment support the subjective-
standard hypothesis.

This hypothesis explains why a stimulus is selected
as the standard. Herein we must introduce the concept of
salience. For a stimulus to be selected as the subjective
standard, it must be noticeable, conspicuous, or salient.

If a stimulus is salient, then it will easily be associated
with a particular response category.

This has been demonstrated in the experiment. In
those stimulus sets containing a stimulus of maximum color
mass, the fewest errors in judgment were made in judging

that stimulus. Hewever, as the salience of the end stimulus

30

decreased so that it was less easily discriminated from other
stimuli in the set, errors increased. When a set was intro—
duced which contained no salient stimulus relative to the
other stimuli in the set, no anchoring occurred. Hence, a
stimulus must be noticeable if anchoring is to occur, and
the more easily a stimulus is recognized, the more influential
it is as an anchor.

Not only is there a significant decrease in errors
in judging the most salient stimulus of a set, but there
is also a decrease in errors in judging stimuli adjacent to
the anchor. The more remote a stimulus is from the salient
end stimulus, the more errors occur in judgment, until at
some point along the stimulus continuum there ceases to be
a significant decrease in errors. Moreover, the more
salient the end stimulus, the further the significant decrease
in errors extends along the stimulus continuum. Thus, the
salient end stimulus influences judgments of adjacent stimuli,
and the extent of this influence is determined by both the
distance of a given stimulus from the end stimulus, and the
degree of salience of the end stimulus.

When no salient stimulus appears in a set, errors
are distributed rather evenly across all judgment categories.

No anchoring occurs. Furthermore, when one of the end stimuli

31

of a set is not readily noticeable, that is, when it cannot
easily be distinguished from other stimuli, anchoring will
not occur at that end of the stimulus set. Consequently,
the generally accepted notion of end anchoring must be
modified, for anchoring does not always occur at one or both
ends of a stimulus set. For end anchoring to occur, one

or both end stimuli must be easily identified as such.
Moreover, a modified definition should also note that anchoring
influences judgments of adjacent stimuli relative to the
salience of the end stimuli and the distance of the judged
stimulus from the end stimuli.

It should be noted that a regression effect is found
in all seven stimulus sets. Regression occurs when the
judgment scale and the stimulus series are not perfectly
correlated. It is observed by regression in judgments toward
the center of the judgment scale. In the data of this
experiment, when perfect correspondence exists between the
anchoring stimulus and the associated judgment category, all
other judgments tend to be underestimated, and the size of
the underestimation depends upon the distance of the judged
stimulus from the anchor. As the salience of the anchor
decreases, judgments of the lower end of the series tend

to be overestimated, while those judgments of the upper end

32

of the series tend to be underestimated. In the middle
set in which there is no anchoring, overestimations are
relatively similar in extent to underestimations. This
supports Johnson's (1955) theory of regression.

Since anchoring is a function of the salience of
the end stimulus in end anchoring, it might then be predicted
that as the salience of an anchor decreases, the correlation
between the stimulus series and the judgment scale would
also decrease. That is, regression would also depend upon
anchoring. The relationship between regression and anchoring
would be an interesting topic for future investigation.

The reader will notice that relatively fewer errors
were made in judging Set 7 than in judging its counterpart
Set 1, although both sets contained a maximally salient
end stimulus. This might indicate a greater ease in judging
the green end of the series. However, if this was true,
then Set 6 should also have been judged with fewer errors
than Set 2. The only reasonable interpretation for this
occurrence is to attribute it to a group of exceptionally

acute judges.

SUMMARY

An experiment was designed to investigate and
clarify the concept of anchoring. Anchoring was defined
as the special way in which certain response categories
appear tied to certain stimuli.

Seven stimulus sets were selected from a systematically
varied stimulus series. The complete stimulus series was
composed of thirty-seven photographic slides, each contain—
ing thirty-six blue and green:dots such that the entire series
varied systematically from an all—blue slide to an all-green
slide. This permitted the E to control the color mass
of the stimuli selected fer presentation.

An eleven point scale was used to express judgments.
Seventy 38 were recruited from introductory psychology
courses, and were randomly assigned to groups of five 85.

Half of the SB judged greenness, and half judged blueness.

The main argument of the experiment was that the
salience of a stimulus, rather than its placement in the
series, determined anchoring.‘ A salient stimulus was de-
fined as a stimulus that is easily noticeable and readily
associated with a response category.

33

34

Other hypotheses were offered relating the effective-
ness of anchoring to the salience of the end stimuli,and the
influence of anchoring upon judgments of adjacent stimuli to
the salience of the end stimulus and to the distance of the
judged stimuli from the end stimulus.

The data were analyzed in terms of errors, an error
being defined as any incorrect association of a response
category to a.-stimu1us.n. An overall analysis of variance
of errors indicated that the data gathered under the two
color conditions could be combined into a single body of
data. An overview of the data was then presented graphically
by calculating the median and the interquartile range of
judgments for each stimulus in a set. This analysis indi-
cated that anchoring effects occurred as predicted.

Subsequently, an analysis of variance of errors for
each set was run on each of the seven_stimulus sets to pro-
vide a more refined analysis of the magnitude and extent of
anchoring effects. Significant F—ratios were obtained
on all sets with the exception of the mid-range set where no
anchoring was predicted to occur. The F—ratios obtained
decreased in significance as the salience of the end

stimulus decreased, supporting the hypotheses.

35

It was suggested that current interpretations of
anchoring would have to be revised to incorporate the con-
cept of salience. The most acceptable interpretation is

the subjective-standard hypothesis, which embraces salience

as the key concept for eXplaining the selection of a stimulus
as the subjective standard. The data of the experiment
indicated that anchoring is not primarily a function of the
placement of a stimulus in a series. Rather, for a stimulus
to act as an anchor, it must be easily identified and

associated with a particular response category.

10.

BIBLIOGRAPHY

Edwards, A. L. Statistical Methods for the Behavioral
Sciences. New York, Rinehart and Company Inc., 1958.

 

Eriksen, C. W. and Hake, H. W. Anchoring effects in
absolute judgments. J. exp. Psychol., 1957, 5;,
132-138.

 

Fitts, D. M. and Deininger, R. S-R compatibility:
correspondence among paired elements within stimulus
and response codes. J. exp;_Psychol., 1954, ﬁg,
483-492.

 

Johnson, D. M. The Psychology of Thought and Judgment.
New York, Harper and Brothers, 1955.

Lindquist, E. F. Design and Analysis of Experiments
in Psychologygand Education. Boston, Houghton
Mifflin Company, 1956.

Needham, J. G. Ratecﬂfpresentation in the method of
single stimuli. Amer. J. Psychol., 1935, 31, 275-284.

 

Onley, J. W. and Volkmann, J. The visual perception
of perpendicularity. Amer. J. Psychol., 1958, 11,
504-516.

Philip, B. R. The relationship of exposure time and
accuracy in a perceptual task. J. exp. Psychol.,
1947, 31, 178-186.

 

Rogers, S. The anchoring of absolute judgments. Arch.
of Psychol., 1941, No. 261.

Secord, P. F., Bevan, W. and Katz, B. The Negro
stereotype and perceptual accentuation. J. abnorm.
and soc. Psychol., 1956, 5;, 78-83.

 

36

11.

12.

13.

14.

15.

16.

37

Sherif, M. and Hovland, C. I. Socigl Judgment.
New Haven, Yale University Press, 1961.

Sherif, M , Taub, D., and Hovland, C. I. Assimilation
and contrast effects of anchoring stimuli on judgments.
J. exp. Psychol., 1958, 55, 150-155.

Volkmann, F C. and Engen, T. Three types of anchoring
effects in the absolute judgment of hue. J. exp.
PSXChOIo I 1961! Q: 7‘17.

VOlkmann, J. Scales of judgment and their implications
for social psychology. In J. H. Rohrer and M.
Sherif, eds. Social Psychology at the Crossroads.
New York, Harper and Brothers, 1951.

VOlkmann, J. The anchoring of absolute scales.
Psychol. Bull., 1936, §§, 742-743.

Wever, E. G. and Zener, K. E. The method of absolute
judgment in psychophysics. Psychol. Rev., 1928,
§§, 466-493.

APPENDIX A

JUDGMENT RECORDING SHEET

JUDGMENT RECORDING SHEET

 

Instructions: You will be shown several series of slides
projectii ng blue and green dots on the screen. The task
before you is to judge the "blueness" or ngeenness of
each slide presented, using an 11-point scale. In other
words, you would judge the "bluest slide as aIO, the "next-
bluest" slide all, and so on, with the "least-bluest“ slide
judged as an II. There are 11 slides in each series, so try
to use all 11 numbers.

PTease do not skip a judgment; make a judgment for each slide

presented. Please make no audible responses. Are there
any questions?

".- "I
I

 

MICHIGAN STATE UNIVERSITY LIBRARIES

III'II III III

3 1293 030715

5

55