a

{WWII

5
1

l
l

I“

I
I
l

IIHIHIH

N—|
_x_\

 

-.—I
I
mooco

THE EFFECTS Q? ETQSJ E T {ME 3N THE

ERRQR C?” CENTREL TEENDENCV {N NEG

T4103“? LOt' {4'13 DGQE‘QQ 3% EA. ’5‘"
EELMETEGEN STETE “MEET-ET?
Caroiyn \E/EIczynmi Mui‘ Ea: y
1967

E‘-
t
‘1

Hub

NT

"it".

 

ABSTRACT

THE EFFECTS OF EXPOSURE TIME ON THE
ERROR OF CENTRAL TENDENCY IN JUDGMENT

by Carolyn Walczynski Mullally

An experiment was designed to investigate the effects of expo-
sure time on the central tendency of judgment. It was proposed that
the manifestations of central tendency are evidence of regression to-
ward the mean occurring as a result of imperfect correlation between
two variables; therefore central tendency was considered synonymous with
regression.

Since the magnitude of the regression effect is measured by:

(a) the inverse of the correlation between the stimulus material and the
judgments, (b) the frequency with which the middle and end judgment cate-
gories are used, (c) the mean judgment and (d) the standard deviation

for each individual stimulus, hypotheses were formulated such that as

the exposure time is reduced, these effects would increase. The design
of this research consisted of two parts: (I) to determine the relation-
ship between exposure time and correlation over a decreasing range of
exposure time, and (2) to study in greater detail the regression effect
for the exposure times located at the extremes of the range.

The stimulus material consisted of eleven photographic slides
composed of 36 dots printed in two colors to form a graded series such
that as one color gradually increased by two dots from card to card, the
other color gradually decreased by the same amount. In order to avoid

the confounding effects of end anchoring, the stimulus set chosen did not

 

Carolyn Walczynski Mullally

contain an anchor or salient stimulus to which the 55 could readily
attach a response value. The slides were judged on an eleven-point
scale ranging from zero (least amount of judged color) to ten (most
amount of judged color).

A total of 97 55 were recruited from introductory psychology
classes. TWenty of these were used in the standardization of the stim-
ulus material and another 25 were used for the first part of the ex-
periment. The remaining 52 55 were randomly assigned to one of the
four experimental conditions of part two: (I) one-second exposure,
slides judged on blueness, (2) one-second exposure, slides judged on
greenness, (3) .04-second exposure, slides judged on blueness, and (h)
.Oh-second exposure, slides judged on greenness.

An inspection of the correlations between the judgments and the
slides, the means, and the standard deviations (of the data of part 2)
indicated that the data gathered under the two color conditions (blue
and green) could be combined. This accomplished, the data were then
analyzed in terms of the correlation between judgments and the stimuli,
the scale value (mean) and standard deviation of each individual stim-
ulus in the series, and a frequency Count for the middle and end judg-
ment categories. An inspection of the graphic and tabular presentation
of the data indicated that the regression increased as a function of de-
creasing exposure time as predicted.

Three possible explanations for the increase in regression as a
function of a decrease in exposure time were offered, but it was noted
that these were not necessarily independent of each other.

The first of these interpretations, that there is a general

Carolyn Walczynski Mullally

tendency for regression to occur in all judgment situations in which

the stimulus objects are not clearly perceived, accounts for the re-
sults in a very simple manner: as the exposure time is reduced the 55
are less and less able to discriminate between the stimuli and hence
make more errors in judgment. The second of these, that the response
scale becomes subjective, cannot be overlooked either. However, it is
the third interpretation, that the regression may be an artifact of the
construction of the stimulus and response scales, which appears to ac-
count for the largest part of the regression in the present research.
Which of these interpretations is the mast applicable most certainly can

be determined by future research.

[,/ , 7 1/} Q /.
Approved -':’/£I{M/ (:1. #131 ‘:7:::a4i//‘;Z¢./

b

THE EFFECTS OF EXPOSURE TIME ON THE

ERROR OF CENTRAL TENDENCY IN JUDGMENT

By

Carolyn Walczynski Mullally

A THESIS

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

MASTER OF ARTS

Department of Psychology

1967

 

ACKNOWLEDGEMENT

The author wishes to acknowledge the kind assistance of
Dr. Donald M. Johnson, chairman of her committee, whose contribution
of his experience and advice made this thesis possible.

The author is also grateful to Dr. Terrence M. Allen and
Dr. William T. Stellwagen for their helpful suggestions and criticisms

in the preparation of the manuscript.

 

TABLE OF CONTENTS

 

Page

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . I
THE PROBLEM . O O . 0 C C 0 O O . O . O O C O . . O . . . 8
Hypotheses 8

METHOD 0 O O O O O O O O O O O O O O O O O O O O O O O O 0 lo

Stimulus Material lO
Judgment Scale ll
Apparatus l2
Exposure Time l2
Subjects l3
Procedure l3

RESULTS 0 O O O O O O O O O O O O O O O O O O O O O O O O '5

DISCUSSION 0 O O O O O O O O O O O O O O O O O O O O O C O 24

SUMMARY 0 O O O O O O O O O O O O O O O O O O O O O O O O 30

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . 33

Table

LIST OF TABLES

Page

Statistics describing judgments on two color
conditions at l.00 and .Oh-second exposure . . . . . 17

Statistics describing judgments for combined data
at l.OO and .Oh-second exposure . . . . . . . . . . l8

Scale (mean) value and standard deviation for each
slide at 1.00 and .ON-second exposure . . . . . . . . 2i

 

Figure

LIST OF FIGURES

Page

Correlation of judgments and stimuli as a function
Of exposure time O O O O O O O O O O O O O O O O O O '6

Regression of judgments on stimuli for l.OO-second
exposure . . . . . . . . . . . . . . . . . . . . . . l9

Regression of judgments on stimuli for .04-second
exposure O O O O O O O O O O O O O O O O O O O O O O 19

Scale value of each slide for l.OO and .Oh-second
exposure . . . . . . . . . . . . . . . . . . . . . . 22

 

Appendix

A.

LIST OF APPENDICES

Judgment Recording Sheet . . . . . . . . . . . . .

vi

Page

35

 

INTRODUCTION

Although the judgment process has been studied in several ways
(including subjective, affective or esthetic, and psychophysical ex-
perimentation), the judgment process itself is generally left undefined.
Johnson (I955, p. 282) stated that ''In general judgment is a conclusive
or decisive process, not a productive one, that brings a thoughtful
episode to an end.” In order to clarify the meaning of judgment, he
compared and contrasted it with several related functions. Judgment
may be thought of as an overt communicative response; that is, a judg-
ment communicates the result of some other process. A judgment is
different from problem solving in that a judgment task involves choos-
ing between two or more responses or courses of action which have been
specified for the individual, whereas the problem-solving task involves
various activities which produce the solution. Or, in judgment 3 choice
is made between alternatives clearly specified; in problem solving
responses are produced to fit specified conditions.

Johnson continued: ”Judgment is, of course, not the same as
perception. If one perceives the situation clearly enough that the
activity under way may proceed, no distinguishable act of judgment oc-
curs. But if the perceptual field is not clear, if the first perception
is not rewarded, choosing one of the alternatives may be called an act
of judgment.” (I955, p. 28h) If the definition is left at this point,
the distinction between judgment and perception seems clear; but, it is
a distinction which becomes less clear upon closer examination. The

I

 

conditions which give rise to judgment are not clearly distinguishable
from those which permit immediate perception. One may differentiate
between the two on a quantitative basis, as did Hollingworth (I925),

by making the assumption that as confidence (in the judgment) increases
and as the judgment time approaches the reaction time, judgment is
minimized and the act approaches direct perception.

Research in the judgment process can be categorized as either
simple or complex, based on the complexity of the stimulus material.
Simple judgment situations typically involve a single aspect of the
stimulus object, such as the length of a line or the pitch of a tone.
Experiments on simple judgments have chiefly been concerned with three
characteristics of the stimulus: perceptual, affective, and abstract.
Complex judgment situations usually involve one or more aspects of the
stimulus situation, such as in selecting a candidate for a public of-
fice or grading an essay. The stimulus material is thought of as
heterogeneous, with no single dimension to which the judge may easily
respond. Research in complex judgment involves all aspects of the
stimulus; generally different stimulus aspects will be salient to dif-
ferent 85. Although there is a great deal more to be said about complex
judgment, it is beyond the sc0pe of this paper, which is primarily con-
cerned with simple judgment.

Some years ago the error of central tendency in judgment was
discovered and named by Hollingworth (I909, l9l0). In an investigation
of the reproduction of hand movements over linear distances, it was
found that within each series of standards used, the 55 developed a

system of positive and negative constant errors. The system had a neu-

tral point at which no constant error occurred; above this point under-
estimation occurred and below it overestimation. That is, the large
distances were underestimated and the small distances were overestimated.
The term ”central tendency” refers to the fact that estimates at the
extremes of the range of stimuli are shifted toward the neutral point.
Many investigators since l9l0 have reported the same phenomenon occur-
ing in judgment experiments using a variety of stimuli under many dif-
ferent conditions and methods. For example, in I926 Ipsen reported this
phenomenon for judgments of lines (cited by Johnson, I955), Woodrow
(I933) for judgments of weights, and Turchoie (I948) for reproduction
of temporal intervals. Johnson, in a I952 paper, stated:
The central tendency of judgment, named by

Hollingworth in I909, is manifested (a) as a

double truncation of the response scale, when

this scale is described in terms of category

thresholds, (b) as an overestimation of stimu-

lus magnitudes below the center of the stimulus

series and an underestimation of stimulus mag-

nitudes above the center of the stimulus series,

when stimulus ratings are averaged.
In addition, Johnson proposed that the central tendency of judgment is
but another manifestation of the regression toward the mean that always
occurs when two variables are imperfectly correlated. In other words,
the averages of judgments of stimuli in a series show some regression
toward the central tendency of that series because of imperfect dis-
crimination. The 5 begins with the stimulus series arranged in succes-
sive categories on a particular physical dimension and judges the
response categories. When the correlation between the stimuli and the

judgments (responses) is less than l.0, the central tendency or regres-

sion effect is the obvious statistical consequence. Hereafter, then,

the central tendency of judgment will be considered synonymous with
Johnson's term ”regression effect.”

Philip (I9N7a) investigated judgments of color mass in the
following manner:

Seven 0's [subjects] made a prolonged series
of judgments of colour mass on material which was
tachistosc0pically exposed for brief controlled
periods. The material consisted os six sets of
cards, each of which comprised eleven cards prin-
ted in two colours to form a graded series such
that, as one colour gradually increased from card
to card the other diminished. Each card held
thirty-six evenly spaced and randomly distributed
dots in either of two colours, blue and green.

On one card, there were twenty-three blue dots

and hence thirteen green dots; on the next card

in the series, twenty-two blue dots and fourteen
green dots; on the next card, twenty-one blue and
fifteen green; the end card of the series had
thirteen blue dots and twenty-three green dots.

As each card was exposed, 0 had to determine the
relative amount of a previously designated colour,
and locate on a scale from one to eleven the point
which, in his Opinion, represented the relative
amount of the designated colour. Hence, the psy-
chometric procedure used was that of the absolute
judgment. Preliminary practice was given before
the experiment proper, and also before each ses-
sion. A total of l,650 judgments was obtained
from each 0, yielding some ll,550 judgments in all.

 

Johnson (I955) suggested that with a response scale consisting of as
many categories as there are stimuli so that it can be related to the
stimuli in a one-to-one fashion and used for repeated judgments of
evenly spaced stimulus objects, the correspondence between the response
scale and the series of stimuli could be perfect. In fact, Philip's
data indicated that the correspondence was a good deal less than per-
fect. With perfect correspondence, the first stimulus in the series

(#I3) would be judged a ”one”, the second stimulus (#IN) would be judged

a “two”, and so on to the last stimulus (#23) which would be judged an
”eleven.” Stimulus #l3 recejved an average rating of 3.3, instead of
”one.“ Stimulus #lh was rated 3.7 instead of I'two." The same regres-
sion toward the center of the scale occurred at the opposite end also.
Clearly, then, the central tendency of judgment or the regression effect

is evidenced by Johnson's illustration of Philip's data on the judgment

of color mass.

 

In a recent study by Johnson and King (I963) the effects of end
anchoring were investigated using slides constructed from Philip's
stimulus series. 55 were asked to judge a set of II slides on a scale
of zero to ten. Anchoring occurred when the end slide was monochroma-
tic (and therefore a salient stimulus) in accord with the subjective-
standard hypothesis of Eriksen and Hake (I957). Anchoring of the other
slides, when it occurred, was interpreted as due to the effect of the
monochromatic slide, even though it was not presented. No anchoring
occurred when neither end was salient (monochromatic). However, re-
gression toward the mean (central tendency of judgment) occurred at
the unanchored end or ends. It was concluded that end anchoring re-
duces the central tendency or regression effect; but at the unanchored
end or ends of a scale of judgment, the judgments regress toward the
mean judgment and the end categories are used with less than average
frequency.

In addition, another study by Philip (I9h7b), in which he
examined the problem of the relationship between exposure time and
accuracy of judgment in a perceptual task, is important to the deveIOp-

ment of the present research. Specifically, Philip was looking for an

answer to the question: ”Does accuracy, in a perceptual task, diminish
according to some continuous function with increasing speed of perfor-
mance of that task or is there an optimal time interval, somewhere be-
tween the highest and the lowest speeds, at which accuracy is at a
minimum?” (p. I78). The stimulus material was similar to that used in
the discrimination of color mass experiment previously reported. The
advantages of using this type of material are: (I) the influence of
prior experience and schooling is negligible, (2) practice effects are
slight, (3) scoring is objective, and (h) judgments may be made quickly.
Philip found that the sigmoid curve was representative of the relation-
ship between accuracy (in terms of errors) and exposure time. It was
noted, however, that Optimal points might occur in the initial stages
of the experiment (during the first couple of trials), but that as the
experiment progressed this effect was lost. Therefore Philip cautioned
that the continuously decreasing function of exposure time and accuracy
applies only to judgments of well-practiced 85. Furthermore, this
sigmoid relation between exposure time and errors may have very narrow
limits beyond which there is slight variation, since at the slowest
speed (near the upper limits of the function) accuracy may decrease
very slowly with increasing Speed. As a logical consequence of this
sigmoid relationship, there was a continuous drop in errors from the
fastest to the slowest speeds. However, Philip also noted that there
was a tendency to assign more judgments to the central positions rather
than to the terminal positions of the judgment scale, i.e., the central
tendency effect, which ”distorted” the results. However, Philip does

not comment on how the presence of the central tendency effect distorts

 

the results. That is, did the central tendency increase, decrease, or
remain constant with decreasing exposure time and thereby increase,
decrease, or not affect the number of errors? Further, is it possible
that it is the exposure time which increases the central tendency ef-
fect which in turn results in increased error or perhaps accounts for
all the error?

It is this question which serves as an impetus for the present

research.

 

THE PROBLEM

Philip's data on judgment of color mass have not only illus-
trated the regression effect, but also have raised questions regard-
ing the nature of the relationship between exposure time and central
tendency (regression). The purpose of this study then, is to inves-
tigate the effects of exposure time on the central tendency of judg-
ment.

The central tendency effect is manifested in two ways: (I)

a truncation of the response scale at both ends, and (2) an underesti-
mation of the stimulus size above, and an overestimation of the stim-
ulus size below the center of the stimqus series. Further, it has
been proposed that these manifestations are simply evidence of regres-
sion toward the mean occurring as a result of imperfect correlation of
two variables (Johnson, I952). The magnitude of the regression effect
is measured by the inverse of the correlation between the stimulus
material and the judgments, the frequency with which the middle and
end judgment categories are used, the mean judgment (scale value) and

the variance or standard deviation for each stimulus.

mm

It is hypothesized that as exposure time is reduced:

I. The regression effect is increased (that is, the correla-
tion is decreased), and concomitantly:

2. The variability of the judgments (indicated by the standard

deviation) of each individual stimulus in the series is

8

 

increased.

The overestimation of stimuli (measured by the scale
value) below the center of the stimulus series and the
underestimation of stimuli (by scale value) above the
center of the stimulus series becomes more pronounced.
The middle judgment categories are used more frequently

than the end categories.

METHOD

The design of this research consists of two parts:

I. To provide a foundation for Part Two by determining the rela-
tionship between exposure time and correlation over the follow-
ing decreasing range of exposure times: l.00, .50, .20, .IO,
.04 second. The results of this phase are reported in terms of
the correlation plotted against the exposure time. The results
of this part will permit choosing specific exposure times for
further analysis and comparison in Part Two.

2. To study in greater detail the regression effect for the ex-
posure times located at the extremes of the aforementioned

range of exposure times.

Stimulus Material

In order to avoid the confounding effects of end anchoring, the
stimulus set chosen did not contain an anchor or salient stimulus to
which the 55 could readily attach a response value. The set was, of
course, patterned after Philip's stimulus material as described earlier
in this paper. However, it was anticipated that the correlation ob-
tained using a set of photographic slides constructed after Philip's
stimuli might be so low, at shorter exposure times, as to make compari-
son of these correlations inferentially unreliable. A preliminary
study to investigate this possibility was undertaken, using four groups
of five 55 each judging the blueness or greenness of the series of

IO

ll

slides at a one-second exposure for each slide. The obtained correla-
tion between the slides and the judgment scales was very low--ranging
from .30 to .40.

Such low correlations and verbal reports from the 55 indicated
that discrimination between slides was poor; therefore, the stimulus
material was standardized to insure discrimination between the slides.
This was accomplished by extending the stimulUs series in a systematic
fashion so that each slide differed from the preceeding slide by two
dots instead of one. The extended stimulus series consisted, then, of
slide #8 (8 green, 28 blue dots), #l0 (l0 green, 26 blue dots), #l2
(I2 green, 24 blue dots), and so on to #28 (28 green, 8 blue dots) for
a total of eleven slides. While this extended set allowed for more
adequate discrimination, it also did not contain a salient stimulus
(anchor) which could increase the accuracy of judgment and thereby re-
duce the regression effect. The set of slides was presented a total

of eight times: two practice trials and six test trials.

Judgment Scale
The 55 were instructed to judge the slides on an eleven-point
scale ranging from zero to ten. The stimulus dimension to be judged
was color; hence, ”0” indicated the least amount of the stimulus color
and l'IO” indicated the greatest amount of the stimulus color. One half
of the Ss judged the slides on ”greenness,” and the other half judged

the slides on ”blueness;” thus there were two judgment conditions.

l2

Apparatus

The slides were projected manually by a Viewlex slide pro-
jector equipped with an Alphax variable-speed, manually-operated
tachistosc0pic lens-shutter. The shutter speed ranged from one to
.0l second. The projector was placed fifteen feet from the projec-
tion screen, and produced an image three by four feet on a beaded
screen. The illumination level in the windowless room was controlled
by a rheostat to insure sufficient light for the $5 to record their
judgments, but not enough to interfere with the image on the screen.

The rheostat setting remained constant throughout the experiment.

Exposure Time

Although the choice of exposure times was fairly arbitrary,
it was dictated to some degree by the limits of the apparatus as de-
scribed earlier. The following exposure times were used in the first
part of this experiment: l.OO, .50, .20, .IO, and .04 second. The
extremes of this range, l.00 and .04 second, were used in the second
part.

The tachistoscope was calibrated at the end of the experiment
and it was found that with constant cable pressure a positive error
was present and without constant pressure the error was negative. At
none of the exposure times did the error approach an overlap with the
adjoining exposure time. In addition, since the experimental procedure
involved maintaining a constant pressure on the exposure-time cable, it
was concluded that the error remained constant and positive throughout.

Moreover, since the actual numerical representation of the exposure

 

l3

time (in milleseconds) did not enter into any of the arithmetic calcu-
lations of the data, there should be no question regarding the statis-

tical results.

Subjects

A total of 97 55 were recruited from introductory psychology
classes. Color-deficient students were requested not to participate
in the experiment. TWenty of these 85 were randomly assinged to four
groups of five 55 each and were used in the standardization of the
stimulus material. This accomplished, twenty-five of the remaining
77 55 were randomly assigned to five groups of five 55 each; each group
subsequently was presented with one of the following exposure times:
l.OO, .50, .20, .IO, .04 second, judging the standardized set of slides
on ”blueness.” This concluded Part One of the experiment.

The remaining 52 55 were then randomly assigned to one of the
four experimental conditions of Part TWo: l) l.OO-second exposure,
slides judged on blueness; 2) l.OO-second exposure, slides judged on
greenness; 3) .04-second exposure, slides judged on blueness; and 4)

.04-second exposure, slides judged on greenness.

Procedure
Each 5 was given a judgment recording sheet containing in-
structions and space to record judgments (responses). After S read
the instructions, E answered any questions raised by S. The stim-
ulus set was then presented twice for practice at the assigned exposure
time; the interslide interval was approximately 8 seconds. The material

was presented by the method of single stimuli or absolute judgment. In

I4

Part One the slides were judged on the blue color dimension only for
each of the five exposure times. In the second part of this experi-
ment one grOUp judged Vblueness” and another ”greenness” for each of
the two exposure times, hence there were four experimental groups.
Following each of the two practice presentations, E asked if there
were any questions and also checked the responses of each 5 to be cer-
tain that the correct procedure was being followed. Since these two
presentations were considered practice trials they were not included
in the analysis of the results.

After the practice trials, the set of slides was presented six
times, in a consistent, but random order. In addition, the orienta-
tion of each slide in the slide holder was chosen at random from one
of the following: forward and backward, and then up, down, left, or
right in combination with the front or back of the slide for a total
‘of eight possible orientations. The randomized order of presentation
and the re-orientation of each slide was used to minimize extraneous
cues, such as pattern of the dots, which might appear in a slide. The
entire presentation took about 30 minutes.

Disregarding the practice trials, the data collection resulted
in the following:

Part I. Each stimulus in the series of eleven slides was
presented six times for judgment by five 55; therefore 30 judgments
were available for each of the eleven slides.

Part 2. Each stimulus in the series of eleven slides was
presented six times for judgment by I3 35; therefore, 78 judgments

were available for each of the eleven slides.

RESULTS

Part I: An inspection of Figure l reveals that as the ex-
posure time is reduced, the correlation between the slides and the
judgments is likewise reduced. The obtained curve is negatively
accelerated; thus an increase in exposure time beyond l.00 second
would result in little increase in accuracy and consequently little
increase in the correlation. 0n the other hand, decreases in expo-
sure time below .04 second would most likely yield greater inaccura-
cies in judgments and therefore the correlation would approach zero.
This concluded Part I of the experiment since, as predicted, the
correlation between the judgments and stimuli was decreased as the
exposure time was decreased. Therefore the decision was made to use
the extremes of the range of exposure times (l.OO and .04 second)

for Part 2.

Part 2: Since there was no interest in the color condition
of judgment per se, it was first determined whether or not the data
collected under the two color conditions for each exposure time could
be combined. This was desirable for several reasons: (I) the amount
of data for each exposure time would double, and (2) the hypotheses
were concerned with the effects of exposure time, not the specific
color being judged. Table l contains the correlations between the
judgments and the stimuli, the means, and the standard deviations for
both color conditions (blue and green) at the two exposure times: l.00

IS

l6

.oE_u ogsmoaxo mo co_yoc:m m mm __:E_um ocm mucmEmozm mo co_um_mLLOu ._tmczm_m

Amazouu: mz_k ”mamomxm
3 a. a. p. .g. N. t a. u. a.

_ . . _ A _ . d . ..

 

t0.

nix/138303

 

NO

I7

and .04 second. Examination of the statistics reveals that they are
sufficiently similar to warrant combining the data without regard to
the color upon which the judgments were based. Thus the data are com-

bined for the remaining analyses.

Table l. Statistics describing judgments on two color conditions at
l.OO and .04-second exposure. r is the correlation between
the judgments and the stimuli, T is the mean and S.D.y is
the standard deviation of the judgments for I3 55 in each
color condition under both exposure times.

 

 

 

l.OO-second exposure .04-second exposure
color _ _

r Y S.D.Y r Y S.D.Y
Blue .72 5.3 3.l .32 5.6 3.3
Green .73 5.3 3.2 .27 5.5 2.9

 

An examination of Table 2 also confirms the first hypothesis:
the regression effect is increased (that is, the correlation is de-
creased) as the exposure time is reduced. Moreover, since the magni-
tude of the regression is inversely related to the size of the correla-
tion, it follows that the regression effect is greater at the .04-
second exposure time than at the one-second exposure time. Another
way to illustrate the central tendency or regression effect is to ex-
amine the regression coefficient or sl0pe constant, which is the product
of the correlation coefficient and the ratio of the standard deviations
of the response scale and the stimulus series. The regression coeffi-
cient for the l.OO-second exposure is .72 and for .04-second exposure
it is .29; thus, it is concluded that the accuracy of the judgments or

the general correspondence between the response scale and the stimulus

l8

series is decreased when the exposure time is decreased. Further-
more, the slope of the regression line for the .04-second exposure
shifts farther away from the sl0pe of the line of perfect correspon-
dence, r = l.0, than does that of the l.OO-second exposure.

Eta, computed for each correlation, indicates that the data
are linear. The correlation ratio, or eta, is a general index of
correlation particularly adapted to data in which there is a curvi-
linear regression. However, the eta coefficient assumes no specific
type of functional relationship between the two variables, therefore,
the eta coefficient is the maximum correlation index for any set of
data. A comparison of the eta and r for each exposure time indicates
that the data must, in fact, be linear since the difference between
the two coefficients is minute.

Table 2. Statistics describing judgments for combined data at l.00 and
.04-second exposure. There are 26 Ss in each exposure-time

 

 

 

group.
Exposure Regression Frequency of Middle and
Time r Eta Equation End Judgment Categories
0 5 l0
l.00 sec. .75 .753 Y=l.7l+.72X l2l I62 I44
.04 sec. .29 .303 Y=4.07+.29X 86 I78 I40

 

As another consequence of the correlation between two variables,
the regression effect increases as the variance (or the standard devia-
tion) of the judgments of each stimulus increases. Figures 2 and 3,

which give an indication of the over-all variability of the judgments

JUDGMENTS

JUDGMENTS

l9

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

/ _I
’1' I I l T 1 If ‘[ I I s
,’ 8 12 16 .10 39- :5
A" S L I D E S
Figure 2. Regression of judgments on stimuli for
I.00-second exposure. The vertical line for
each stimulus value illustrates the mean judg-
ment (scale value) and one standard deviation
on both sides of the mean. Line ”A'' is the
~ line of perfect correlation (r = l.0) and Line
”B” is the regression line. i
A .
It '
e~ " /
. ‘I I , / ‘ /B
I ‘L/Z’
L r’
# .
43/.- II
t'/ I ‘l d»
I I r— I I I T I 1 I I T
/ 8 I2 (6 20 24 26

S L I D-E S

Figure 3. Regression of judgments on stimuli for
.04-second exposure.

20

for each stimulus, support the second hypothesis. Table 3 lists the
specific statistics for each stimulus under both time conditions.
Clearly, the variability of the judgments of each individual stimulus
in the series is increased as the exposure time is reduced. At the
same time the variability of the judgments of the stimuli across the
series remains homogeneous. However, the distributions of the judg-
ments of the end stimuli are skewed. The reader is thus cautioned
not to conclude from Figures 2 or 3 that judgments greater than l0 or
less than zero occurred. Comparing Figures 2 and 3 reveals that for
each stimulus category the standard deviation is larger at the .04-
second exposure than at the I.00-second exposure. Reference to Table
3 indicates that at the .04-second exposure, the standard deviations
range from 2.4 to 3.3; at the l.OO-second exposure, the standard de-
viations range from l.7 to 2.2.

An examination of the mean judgment (scale value) for each
stimulus, under both time conditions, reveals that as the exposure
time is reduced the overestimation of the stimuli below the center of
the series and the underestimation of the stimuli above the center of
the series becomes more pronounced. For example, reference to Table 3
indicates that at l.OO-second exposure stimulus #8 received an average
rating of l.4 and stimulus #28 received an average rating of 8.6; on
the other hand, at .04-second exposure, stimulus #8 received an average
rating of 3.9 and stimulus #28 received an average rating of 6.5.
Figure 4 also illustrates the dispersion of the scale (mean) values.

These results are, of course, in accord with the third hypothesis.

2l

Table 3. Scale (mean) value and standard deviation for each slide
at l.00 and .04-second exposure.

 

 

 

Scale (Mean) Value Standard Deviation

Slide Number 1.00 sec. .04 sec. 1.00 sec. .04 sec.
28 8.6 6.5 1.9 2.7
26 7.9 6.9 2.0 2.7
24 7.8 6.6 1.7 2.8
22 6.7 6.0 2.2 2.4
20 6 3 5.9 1.9 2.6
18 5.5 5.9 1.9 2.7
I6 4.7 5.5 l.9 2 4
I4 4.0 4.8 l.9 2.6
12 3 1 4.3 1.8 2.7
I0 2.3 4.4 l.9 2 8

 

(MEAN) VALUE

S C A L E

‘22

,0 q . w

9 ~ _ 9

8 -~\\\ 2€3 ~ 8
\

7 - 24 26\:- 7
1.\ 22 ,

 

 

5- ~5

4.1/ *4-
12

.3 .::::;’/’1O _ 3

Z. 8 2

/

I“ . LI
SLIDE NUMBER

0- ' LO

l.OO SECOND .04 SECOND

Figure 4. Scale value of each slide for l.OO and
.04-second exposure.

23

Table 2 also gives a tabulation of the frequency of judgments
in the end categories, 0 and l0, and in the middle category, 5, which
indicates the relationship between the central tendency of judgment
and exposure time. If all the judgment categories were used equally
often, the frequency for each category would be l56. In fact, the
central tendency or regression effect affects the judgments for both
exposure times, for categories 0 and ID are used with less than aver-
age frequency and category 5 is used with greater than average fre-
quency. A comparison of the category frequency across exposure times
indicates that the regression effect is greater for the .04-second
exposure, since the and categories are used less frequently and the
middle category more frequently, than the frequencies for the I.00-
second exposure time. Although these frequencies are in the expected
direction and lend support to the fourth hypothesis, it seems apparent

that they are not significantly different across exposure time.

DISCUSSION

It has been suggested earlier in this paper that the mani-
festations of the central-tendency phenonemon are simply evidence
of regression toward the mean occurring as a result of imperfect
correlation of two variables. In the present research the variables
are the stimulus series (slides) and the response scale (judgments).
When these variables are imperfectly correlated, the stimulus ratings
(scale values) regress toward the center of the rating scale and the
response scale is extended and truncated, resulting in the regression
effect. The magnitude of the regression is inversely related to the
correlation between these variables; that is, as the correlation de-
creases the regression effect increases, and vice versa. As the re-
gression effect is increased (correlation decreased), three concomi-
tant statistical consequences would be expected:

I. The variability (or standard deviation) of the judgments
of each individual stimulus is increased.

2. The overestimation of stimuli below and underestimation
of stimuli above the center of the stimulus series, as
measured by the mean judgment (scale value) is increased.

3. The middle judgment (response) categories are used more
frequently than the end categories.

These three statistical consequences are dependent upon the correlation

and upon each other and can be easily noted in the 55' performance.

24

 

25

The present research was designed to investigate the effects
of exposure time on the central tendency of judgment or regression
effect. The independent variable, exposure time, was systematically
decreased and its effects on the dependent variable, the correlation
between the stimulus and response scales, were noted. It has been
shown that as the exposure time is reduced the correlation is decreased
and therefore the central tendency or regression effect is increased.
Another way to state these results is that the discrimination accuracy
(as measured by the correlation) is decreased as a function of de-
creased exposure time. It is concluded that in general the results
support the hypothesis that a reduction in exposure time increases the
central tendency of judgment or regression effect.

There appear to be three possible explanations for the increase
in regression as a function of decreased exposure time, although it
should be noted that these are not necessarily independent of each
other.

First, there is a general tendency for regression to occur in
all judgment situations in which (a) the stimulus objects are not
clearly perceived, (b) the response scale is not well defined, or (c)
the stimulus and response scales are not easily brought into corres-
pondence. In the present research (a) seems to account for the results
in a most straightforward and simple manner: as the exposure time is
reduced the S is less and less able to discriminate between the stimuli
and hence makes more errors in judgment. Therefore the correlation be-
tween the stimulus and response scales decreases, and the regression

effect.increases. It is possible, however, for the regression effect

26

to be found in judgment situations which do not fit the above requi-
sites. In a recent study of end anchoring by Johnson and King (I963)
discussed earlier in this paper, it was reported that regression oc-
curred in all stimulus sets, but that the effects were reduced in
those sets which contained an end anchor (defined as a salient stim-
ulus). In the present research the independent variable (exposure
time) was manipulated in such a way as to make the stimuli less dis-
criminable and thereby increase regression. Conversely, in the
Johnson and King study the independent variable (salience of the end
stimulus in the set) was manipulated in such a way to increase the
discrimination between the stimuli and hence decrease the regression
effect. Although the effect of the anchor (salient stimulus) was to
decrease the regression effect, it must be noted that the regression
effect was not eliminated completely. Therefore, it may be concluded
that a general tendency exists for regression to occur in all judg-
ment situations and that certain experimental conditions may either
increase or decrease the amount of regression.

While the first explanation for regression relied on control-
lable experimental conditions, the second explanation depends upon the
$5; i.e., it is subjective and not easily controlled.

The second possible explanation is this: the response scale
becomes subjective; specifically, the Ss tend to shorten the scale by
extending the category intervals. The 55 have been instructed that
every category in the response scale is represented by one of the stim-

uli in the series and that on each trial all categories will be pre-

27

sented. In any given trial, the 55 are presented with the series of
slides in a predetermined random order. In judging each stimulus,
the 55 are hampered because the end (extreme) stimuli are not identi-
fied, there is no standard stimulus against which to compare the
stimulus at hand, and the exposure time reduces the discrimination
between slides. Therefore, the Ss judge conservatively, ”saving” the
end categories in hopes of using them for the appropriate stimuli.
However, these stimuli never seem to appear; that is, the trial is
over before the Ss realize that they haven't used the end responses.
After responding in this manner for several trials the Ss begin to
use the end categories, but not with as much frequency as that with
which the middle categories are used. Further research is indicated
here, to determine at which point in the sequence of trials the Ss
begin to use the end response categories more frequently.

In the Johnson and King study the 55 did not shorten the re-
sponse scale in those stimulus sets which contained an anchor, at
least not at the end of the scale which contained an anchor. That the
scale was shortened at the end not containing the anchor is evidenced
by the regression occurring at that end.

The third possibility is this: it appears that regression may
be an artifact of the stimulus and response scales, for at the ends of
the stimulus series there is only one direction (toward the middle) in
which judgments may vary. Therefore, since erroneous judgments may be
made only in the direction of the center of the response scale, regres-

sion is a most obvious consequence. In other words, both the stimulus

28

and the response scales are finite, and, in addition are related in a
one-to-one manner. These conditions combine in such a way as to lit-
erally force the $5 to make judgment errors in the direction of the
middle of the scale. An examination of the distributions of each
individual stimulus in the series (regardless of the exposure time)
illustrates this point more clearly. The distributions of the end
stimuli are very highly skewed (the tails pointing in the direction of
the center of the scale), while the distributions of the middle stim-
uli are more nearly normal. It seems fairly obvious that this should
be the case, for the judgments of the end stimuli may vary in only one
direction whereas the judgments of the middle stimuli may vary in two
directions. In a certain sense, then, an interaction between judgments
and the position of the stimuli on the physical scale exists which
accounts for the regression effect.

While it may be possible that any one of these explanations
would account for the experimental results, this seems unlikely to
this experimenter. 0n the other hand, it is the Opinion of this re-
searcher that the third explanation--the construction of the stimulus
and response scales--is more likely to explain the largest part of
the regression. However, all three of these possibilities should be
the subject of further research.

The following questions were raised by the present research
and give rise to possibilities for future research:

I. What happens (to central tendency) when the end stimuli

are identified (as containing the least and greatest

29

amount of the designated color), but are not monochromatic?
What happens (to central tendency) when both exposure time
and judgment time are limited (varied)?
Is there a point in the sequence of trials at which 55 be-
gin to use the end judgment categories with average fre-
quency?
(a) Are the and categories used only in “desperation‘I
--i.e.--at the end of the trial when S realizes
that he hasn't used the categories?
(b) Is the subjective shortening of the response scale
related to (a)? How?
What happens (to central tendency) when the response scale
is unbounded (that is, a center point such as I00 is iden-
tified and S is free to extend the scale as far as he wishes

on both sides of the center)?

SUMMARY

An experiment was designed to investigate the effects of ex-
posure time on the central tendency of judgment. It was proposed that
the manifestations of central tendency are evidence of regression to-
ward the mean occurring as a result of imperfect correlation between
two variables; therefore central tendency was considered synonymous
with regression.

Since the magnitude of the regression effect is measured by:
(a) the inverse of the correlation between the stimulus material and
the judgments, (b) the frequency with which the middle and end judg-
ment categories are used, (c) the mean judgment (scale value) and (d)
the variance or standard deviation for each individual stimulus, hypo-
theses were formulated such that as the exposure time is reduced,
these effects would increase. The deéign of this research consisted
of two stages: first, to determine the relationship between exposure
time and correlation over a decreasing range of exposure time, and
second, to study in greater detail the regression effect for the expo-
sure times located at the extremes of the range.

The stimulus material consisted of eleven photographic slides
composed of 36 dots printed in two colors to form a graded series such
that as one color gradually increased by two dots from card to card,
the other color decreased by the same amount. In order to avoid the

confounding effects of end anchoring, the stimulus set chosen did not

30

3l

contain an anchor or salient stimulus to which the 55 could readily
attach a response value. The slides were judged on an eleven-point
scale ranging from zero (least amount of judged color) to ten (Most
amount of judged color).

A total of 97 55 were recruited from introductory psychology
classes. TWenty of these 55 were used in the standardization of the
stimulus material and another 25 were used in the execution of the
first part of the experiment. The remaining 52 55 were randomly as-
signed to one of the four experimental conditions of part two: (I)
one-second exposure, slides judged on blueness, (2) one-second expo-
sure, slides judged on greenness, (3) .04-second exposure, slides
judged on blueness, and (4) .04-second exposure, slides judged on
greenness.

An inspection of the correlations between the judgments and
the slides, the means, and the standard deviations (of the data of
part 2) indicated that the data gathered under the two color condi-
tions (blue and green) could be combined. This accomplished, the
data were then analyzed in terms of the correlation between judgments
and the stimulus series, the scale value (mean) and standard devia-
tion of each individual stimulus in the series, and a frequency count
for the middle and end judgment categories. An inspection of the
graphic and tabular presentation of the data indicated that the re-
gression increased as a function of decreasing exposure time as pre-
dicted.

Three possible explanations for the increase in regression as

32

a function of a decrease in exposure time were offered, but it was
noted that these were not necessarily independent of each other.

The first of these interpretations, that there is a general
tendency for regression to occur in all judgment situations in which
the stimulus objects are not clearly perceived, accounts for the re-
sults in a very simple manner: as the exposure time is reduced the
55 are less and less able to discriminate between the stimuli and
hence make more errors in judgment. The second of these, that the
response scale becomes subjective, can not be overlooked either.
However, it is the third interpretation, that the regression may be
an artifact of the construction of the stimulus and response scales,
which appears to account for the largest part of the regression in the
present research. Which of these interpretations is the most appli-

cable most certainly can be determined by future research.

l0.

ll.

I2.

l3.

l4.

BIBLIOGRAPHY

Eriksen, C. W. and Hake, H. W. Anchoring effects in absolute
judgments. J. exp. Psych., I957, 53, l32-l38.

Garrett, H. E. A study of the relation of accuracy to speed.
Arch. Psych., I922, No. 56.

Guilford, J. P. Psychometric Methods. New York: McGraw-Hill,
l954.

Guilford, J. P. Fundamental Statistics in Psycholoqy and
Education. (4th. ed.) New York: McGraw-Hill, I956.

Hollingworth, H. L. The inaccuracy of movement. Arch. Psych.,
I909, No. l3.

Hollingworth, H. L. The central tendency of judgment. J. Philos.
Psych. and Sci. Method., l9l0, Z, 46l-469.

Hollingworth, H. L. The definition of judgment. Psych. Rev.,
1925. 32. 337-361.

Johnson, D. M. A systematic treatment of judgment. Psych. Bull.,

I945: [4'29 '93'224-

Johnson, D. M. The central tendency of judgment as a regression
phenomenon. Amer. Psych., I952, Z, 28I.

Johnson, D. M. The Psycholoqy,of Thought and Judqment. New York:
Harper and Brothers, I955.

Johnson, D. M. and King, C. R. Systematic study of end anchoring
and central tendency of judgment. J. exp. Psych., I964, 62,
501-5060

Kellog, W. N. Time of judgment in psychometric measures. Amer. J.

PSYChO, '93], 53’ 65-860

Philip, B. R. Generalization and central tendency in the discri-
mination of a series of stimuli. Canad. J. Psych., l947a, 1,
196-20“.

Philip, B. R. The relationship of exposure time and accuracy in

a perceptual task. J. exp. Psych., l947b, 31, I78-I86.

33

IS.

l6.

l7.

l8.

I9.

20.

34

Rundquist, E. A. Response sets: a note on consistency in taking
extreme positions. Educ. Psych. Meas., I950, 19, 97-99.

Tresselt, M. E. The influence of amount of practice upon the
formation of a scale of judgment. J. exp. Psych., I947, 31,
251-2600

Turchoie, R. M. The relation of adjacent inhibitory stimuli to
the central tendency effect. J. gen. Psych., I948, 39, 3-l4.

Wever, E. G. and Zener, K. E. The method of absolute judgment in
psychOphysics. Psych. Rev., I928, 35, 466-493.

Woodrow, H. Weight discrimination with a varying standard. Amer.
JO PSYCh.’ I933, £5, 391-hl6o

Woodworth, R. S. and Schlosberg, H. Experimental Psycholoqy.
(Rev. ed.) New York: Holt, Rinehart, and Winston, I954.

JUDGHSIT RECORDIIG snas

 

”Greenness”
Instructions: You will be shown a series of sliles projectinn blue ani
green dots on the screen. Phe task before you is to julre the ”greennezs”
of each sliie presented, usiee a scale from O to 10. In other woris, you
would juiee the greenest slile as a 10, the ”next rreen” slile a 9, 311 s
on, with the ”least-cream” slide iuleed as a 0. However, if you feel t.
2 or more sliles have the same anouqt of greenness, you may use the same
number for each. _—
"leastLgreen"
.

 

(+0

 

I! A , I!
most ¢reee
l 1 I 1 1 1 1 1 I

/0
Please do not skip a judgment. Also, do QDC comoere your resooqses with
those of the person next to you. He will basin with two 0r1ctice trials

i
followed by six test trials. Record your juizmeqts for both practice uni
test trials. Are there any questions?

 

 

 

 

 

Practice ifl Test 51 East 73 To3+ I;
E 3 2 B E 3 2 3
1 ___ 23 ___ 45 .__ 67 ___
2 ___ 24 ___ 46 ___ 68 ___
3 ___ 25 ___ 47 ___ 59 ___
4 ___ 26 ___ 48 ___ 70 ___
5 ___ 27 ___ ho ___ 71 ___
*3 __ 2‘3 __ 5o __ 72 __
7 _ 29 __ 51 ___. 75 __
° ___ 30 ___ 52 ___ 51 ___
9 __ 31 __ 55 __ 75 _
10 (___ 32 ___ 54 ___ 75 _._
11 __ 55 __ 55 __ 77 __
Practice 52 Test :2 TESt 2E Test 5
12 ___ 54 ___ 56 ___ 7? ___
13 __ 3, __ 57 __ 7'9 ___.
14 __ 36 __ 5% __ 80 __
15 __ 37 _ 59 __ .91 __
16 ___ 58 ___ 61 ___ 92 ___
l7 _ 39 __ él __ 85 __
1Q ___ 4O ___ 62 __‘ 04 -__
19 ‘___ 41 ___ 63 -__ 3% __—
20 ___ “2 ___ 64 ___ sg ___
21 ___ 45 ___ 65 ___ 97 .._
22 _— ’+’+ ___. 66 __ 8'1, __
Name w or p_‘__.
Class E 3-Set ____
Section Color

 

I
II
II
I
I
I
I
I
I
I
I
I
I
III
|
|
I
|
I
I
III:
I
III
I
I
III

03142 9354

3 129