(9J1 nu‘. 23*»:-

‘9 19...; y“,

1 v vr-r‘ v" v

t! .

"km". wx. v- -‘n‘u‘mrn .

mu

 

 

Lgs.

n v
«In
13!:
r- g
4
1‘1.

9 A.” .
x._ :‘f

,i

o. a:

‘~

at

”val—4
1 y

 

:
.J‘ Lrn
., a“. , "1- u
ivr'h

.4! “
3:2‘;

Univ.)
”.3” ._V
. 4”

R: St cult“
autumn
, ,. ,

715'
r"r1:v\ nnrw ru
ls. tn- m.

If

In. ..
a: r;...,.:(. ..
" ,m. ..: n
nor-x N‘-

., ‘fiq v
.- c

J. finfiz'r? . 0‘ r__,
“'. "x .1 r; . v-Jg: '— nuzéw_:~-H: r

..:J&-.,.;,.,,,....m- , ' ' v- o3; F‘t (
r . V 1-~ifi ’¢:,!:,J.J..»,.rm, ,. '
~ v J; w- ..
'7 ' P‘”?

. .r‘» 23?.

M; 'f,’

- vvnn L.
.m n“; u- an?"

' ”u . ‘
:.."

v"
A;,'L.: m

M . a .- I: «J v 4
8y “ , . ‘ ..- .. .5 .. .. a ‘. 2,, '2.
. . . r - an a 5 5g." 1 , ‘ .. J r . . ..- -. .; ...
""“Q’nvé‘u‘: - e .. ' f3 “z.{,z;v..:.r ”2"???" by? .. J." ‘ m . .4. a. "" r‘ 7:— ”Ma.
73?}. J}; H 'v J J ff. "J" I! i" ‘1 l a o nu- - .37 31.. «
a .‘r. L . I . - r. ,—- Xv- -.‘ 1. .u .7 . 0
5,1,. $5 m. .. 'r . . .fi « ‘1‘; .u..:, ...::‘.r.;;. - u.
. . . pl » - ....
w ”ntiJJIm‘ «15 f a}. "’3 »
3“ ,w 'I'..H':,..m fix; . ’ , :7 V.
3.2.: .3,......«...._ ,..... ‘,,, A
4.. J, . .. pnitwmuan-b .. l . ”—42-" ,',,;.’,
. 4'. . 9 . , ,, ”no..-“ ..
" W... ,J': ' 12::- ""
-

.: nit-'7
7:23 a? 3.3”;- A
7-" 1 ' .._.. " If: 5;."- ....
fax... 1,. ' .rxr.-..,xx.-..,,.r.,”WV...
. .J, J "mu“ ‘ -. 4.“... ”3""“1?” ..

I

H»- ' Jan-"Va. ‘

u.-." '7. :‘a.._:—~.'1Q
.9-

.-. —cv J
2' J!" n 145'!"

.,. ”v.4 “TIMI..."

"' “til;

~~M:.L.. '- .3-
,Jr’x'. H‘IMV '{w .
' ‘ {4:411 ‘52 25::1’: 11'3”,

J nn ,3“. fl, ,
Y H «‘0 \ " l
....¢,‘.....'1,._.'. "1.3?
31:1"? ..r. .3. «Ba

:‘1 11.~\\<-A

 

(HESS

ARIES

iiiiiiiiiiini ii iii

This is to certify that the
thesis entitled

The Effects of Processing Coal and
Level of Information Inconsistency on the
Processing of Performance Information

presented by

Lisa Wood

has been accepted towards fulfillment
of the requirements for

M. A . degree in Psychology

 

 

flkfiwfi/

Major professor

Date November 8, 1990

 

0-7639 MS U is an Aflirmative Action/Equal Opportunity Institution

 

 

LIBRARY
Michigan State
University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.

TO AVOID FINES return on or betore date due.

 

 

 

 

 

DATE DUE DATE DUE DATE DUE
W ,w i; .
_—T
L AW

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

fi—F:

 

 

 

 

 

MSU Is An Affirmative Action/Equal Opportunity Inetitflion

cmmhx

 

THE EFFECTS OF PROCESSING GOAL AND LEVEL OF
INFORMATION INCONSISTENCY ON THE

PROCESSING OF PERFORMANCE INFORMATION

By

Lisa Wood
A THESIS

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

MASTER OF ARTS

Department of Psychology
1990

ABSTRACT
THE EFFECTS OF PROCESSING GOAL AND LEVEL OF
INFORMATION INCONSISTENCY ON THE
PROCESSING OF PERFORMANCE INFORMATION
By

Lisa Wood

The effects of an individual’s goal when processing
performance information and the inconsistency of subsequent
information on performance ratings and recall of performance
information were examined. Undergraduate psychology
students (n = 178) viewed slides containing behavioral
statements about the performance of two police officers, and
were asked to recall the information and to make performance
ratings based on the information. Interaction effects
between processing goal, level of information inconsistency,
and initial performance level on recall and ratings were
found, providing support for the hypotheses dealing with
recall of the performance information. However, support for
the hypothesis dealing with performance ratings was mixed.
Theoretical and practical implications of these results are
discussed, and directions for further research are

suggested.

ACKNOWLEDGEMENTS

This thesis is a very nice study that I hope will make a
contribution within the field of Industrial/Organizational
Psychology. It is an accomplishment that I am proud of.

Of course, I could not have written this thesis without the
support of many important individuals. First, I am
grateful to my committee chair, Kevin Ford, who worked so
hard to keep me focused and on track, and to the other
members of my committee, Neal Schmitt and Steve Kozlowski,
whose comments provided valuable insight. I would like to
thank my friends and colleagues at Michigan State
University who are always willing to share ideas, answer
questions, or just listen. I would also like to thank my
family for their continued love and encouragement.

Finally, I especially wish to thank my husband, Galen, who
brings love and laughter into my life, who has never
doubted me even when I’ve doubted myself - who makes all of

this meaningful.

iii

TABLE OF CONTENTS

Page
LIST OF TABLES . . . . . . . . . . . . . vi

LIST OF FIGURES O O C O O C O O O O O O O Vii

INTRODUCTION . . . . . . . . . . . . . . 1
Processing Goal . . . . . . . . . . . 5
Level of Information Inconsistency . . . . . 14
Interaction between Processing Goal and

Level of Information Inconsistency . . . 25
Summary . . . . . . . . . . . . . . 29

Hypotheses . . . . . . . . . . . . . 31
Effects on recall of subsequent

information . . . . . . . . . 34
Effects on recall of initial

information . . . . . . . . . 35
Effects on ratings . . . . . . . . . 37

METHOD 0 O O O O O O O O O O O O O O O 4 3

Subjects . . . . . . . . . . . . . . 43
Design Parameters . . . . . . . . . . . 43
Processing goal . . . . . . . . . . 43
Level of information inconsistency . . . 45
Initial performance level of
targets . . . . . . . . . . . 46
Stimulus Material . . . . . . . . . . . 46
Procedure . . . . . . . . . . . . . 47
Dependent Measures . . . . . . . . . . 50
Performance ratings . . . . . . . . 50
Recall measure . . . . . . . . . . 50
RESULTS . . . . . . . . . . . . . . . 53
Recall . . . . . . . . . . . . . . 55
Subsequent information . . . . . . . 55
Initial information . . . . . . . . 57
Performance Ratings . . . . . . . . . . 63

iv

DISCUSSION . . . . . . . . . . .

Summary of Results . . . . . . .
Effects on recall of subsequent
information . . . . . .
Effects on recall of initial
information . . . . .
Effects on performance ratings
Interpretation of Results . . . .
Implications of Results . . . .
Limitations of the Study . . . .

LIST OF REFERENCES . . . . . . . . .

APPENDIX A: Instructions for Processing Goal
Conditions . . . . . . .

APPENDIX B: Stimulus Material: Performance
Information . . . . . . .

APPENDIX C: Consent Form . . . . . .
APPENDIX D: Interference Task: Questionnaire
Dealing with Different Modes of
Processing Information . . .
APPENDIX E: Measures of Dependent Variables
APPENDIX F: Debriefing Statement . . . .

APPENDIX G: Results of Analyses with Testing
as an Independent Variable . .

84

86

90

91
93
95

96

LIST OF TABLES

Page
Table 1: Means, Standard Deviations and
Intercorrelations between Dependent
Variables . . . . . . . . . . . . 54

Table 2: ANOVA Summary for Recall Variables . . . 56

Table 3: Means and Standard Deviations for Amount
of Subsequent Information Recalled . . . 58

Table 4: Means and Standard Deviations for Amount
of Initial Information Recalled . . . . 60

Table 01: ANOVA Summary for Recall Variables
(with Testing Order as an Independent
Variable) . . . . . . . . . . . 96

Table 62: ANOVA Summary for Rating Variables

(with Testing Order as an Independent
Variable) . . . . . . . . . . . 97

vi

LIST OF FIGURES

Page

Figure 1: Predicted Interaction between Processing

Goal and Level of Inconsistency on

Recall of Subsequent Information . . . . 36
Figure 2: Predicted Interaction between Processing

Goal and Level of Inconsistency on

Recall of Initial Information . . . . . 38
Figure 3: Predicted Effect of Processing Goal and

Level of Inconsistency on Ratings when

Initial Performance is Low . . . . . . 40
Figure 4: Predicted Effect of Processing Goal and

Level of Inconsistency on Ratings when

Initial Performance is High . . . . . . 41
Figure 5: Experimental Design . . . . . . . . 44
Figure 6: Processing Goal X Initial Performance

Level X Level of Inconsistency Interaction

for Recall of Initial Information . . . . 62
Figure 7: Observed Interaction of Processing Goal

and Level of Inconsistency on Ratings when

Initial Performance is Low . . . . . . 64
Figure 8: Observed Interaction of Processing Coal

and Level of Inconsistency on Ratings when

Initial Performance is High . . . . . . 66

vii

INTRODUCTION

Much of the current work in the area of performance
appraisal has been dominated by theory and research
concerned with the cognitive processes involved in making a
performance evaluation (e.g. Ilgen & Feldman, 1983; DeNisi,
Cafferty, & Meglino, 1984; DeNisi & Williams, 1988). It has
been suggested that, in order to understand the errors made
in performance ratings, one must first understand the
operations involved in processing performance information
(DeNisi et al., 1984). To achieve this, researchers have
borrowed from the social cognition literature, which has
been concerned with an individual’s ability to process
information about others (Ilgen & Feldman, 1983).

DeNisi and his colleagues (1984) outlined a model of
performance appraisal that views appraisal as a series of
social cognitive operations: beginning with (1) the
observation and acquisition of performance information,
followed by (2) the encoding and storage of this information
into memory, and (3) its retrieval from memory, ending with
(4) the integration of the retrieved information in order to
assign a formal evaluation. DeNisi et a1. (1984) also

indicated that each of these steps in the appraisal process

2
is affected by a number of personal or contextual factors.
By examining both the social cognition and performance
appraisal literature, they were able to identify a number of
factors that influence how performance information is
processed. Based on the premise set forth by DeNisi and his
colleagues, the present research investigates how two
contextual factors (processing goal and level of information
inconsistency) affect the encoding and storage of
performance information which, in turn, affect the retrieval
of the information and the judgments that are made.

A distinction has been made between an impression goal
(incoming information is organized into an impression), and
a memory goal (incoming information is organized so that it
can be recalled at a later time). It has been suggested
that one’s processing goal determines the way in which
information is encoded and stored in memory as well as the
information that is recalled (Hamilton, Katz, & Leirer,
1980; Srull, 1983). More recently, researchers have used
processing goal to try to explain the lack of correspondence
between a judgment made of a person and specific information
that can be recalled (Hastie & Park, 1986; Lichtenstein &
Srull, 1987).

The differentiation between impression and memory goals
of processing information has implications for performance
appraisal. A common view is that performance information is

stored in memory and is recalled when a performance rating

3
is made (memory-based). The process of performance
appraisal is most likely more complex. In fact, it is
likely that individuals also form an impression as
information is acquired (impression-based). It is also
likely that individuals switch from goal to goal throughout
the process of acquiring performance information. Recent
research in the area of performance appraisal has examined
the relationship between memory and judgment as a way of
identifying the conditions under which performance
evaluations are impression-based (i.e. when the correlations
between the information recalled and ratings are non-
significant) and when they are memory-based (i.e. when
correlations are significant) (DeNisi, Robbins & Williams,
1989; Williams, Cafferty & DeNisi, 1990).

The social cognition literature has established that an
individual’s processing goal affects the retrieval of
information through its effect on the encoding and storage
of the information. The proposed research intends to
examine whether processing goals differentially affect
performance ratings.

The second contextual factor examined in this study
deals with the processing of information that is
consistent/inconsistent with previously acquired
information. This variable, identified by DeNisi et a1.
(1984) as a factor that can influence the encoding, storage

and retrieval of performance information, has also been

4
extensively researched in the social cognition literature.
It has been suggested that information is differentially
processed based on its level of consistency (Hastie, 1980).
A model proposed by Bastie (1980) offers a theoretical
explanation for differential processing. Research examining
the effects on recall have supported the model (e.g. Hastie
& Kumar, 1979; Srull, 1981).

The theory and research involving the effects of level
of information inconsistency on the processing of
information about others has implications for research on
performance appraisal. Performance appraisal is an on-going
process. New information about ratees is constantly being
acquired. The present study investigates how new
information about a ratee is considered in relation to
previously acquired information. The social cognition
research has established that level of information
inconsistency affects the retrieval of information through
its effect on the encoding and storage of this information.
This study examines how new information that is inconsistent
with previously acquired information affects subsequent

performance ratings.

Therefore, this paper examines two variables from the
social cognition literature to increase our understanding of
the processing of performance information. Specifically,
the present research has three purposes. First, it

integrates the research and theory concerned with the

5

effects of processing goal and level of information
inconsistency on the processing of information. Second, the
limitations of the existing research are described, and this
study makes extensions to the existing research by focusing
on the limitations. Finally, the present research hopes to
make a contribution within the area of performance
appraisal. The variables of processing goal and level of
information inconsistency have not been simultaneously
examined in the context of performance evaluation.

The next two sections review the literature on
processing goals and level of information inconsistency.
The third section reviews the literature that has
simultaneously researched both processing goals and level of
information inconsistency. This section also integrates the
literature dealing with these two variables and identifies
the limitations of the existing research. This sets the
stage for the development of the hypotheses for this study.
The last section contains the development of hypotheses as
well as a review of the literature that lend support to
these hypotheses.

Processing Goal

Several researchers (Hamilton, Katz & Leirer, 1980;
Lichtenstein & Srull, 1987; Srull, 1981, 1983) have
distinguished between the effects of impression and memory
goals on the processing of behavioral information.

According to this perspective, an individual given the goal

6
of forming an impression about a target person will actively
organize incoming behavioral information about the target
person. As the individual tries to develop a coherent
representation of the target person, he/she will engage in a
process of relating new information to information already
processed (Hamilton et al., 1980). That is, the individual
does not encode each new piece of information independently
but processes it in relation to what is already known about
the target person (Srull, 1983). The resulting
representation of the information in memory consists of many
associative paths between individual pieces of information.
This perspective also suggests that these associations
between items facilitate an individual’s later retrieval of
information about the target person (Hamilton et al., 1980,
Srull, 1983). It is also suggested that, under an
impression goal, individuals use the incoming information to
make evaluations as the information is acquired (referred to
as an on-line judgment), causing the information to be
stored separately from the evaluation that is formed
(Lichtenstein & Srull, 1987).

On the other hand, if an individual is instructed to
remember incoming information about a target person (memory
goal), there should be no attempt to organize and
interrelate pieces of information. Instead, efforts are
directed towards other strategies, such as rehearsal of each

individual piece of information (Hamilton et al., 1980). In

7
addition, evaluations will not be made during information
acquistion but are computed on the spot from the specific
information stored in memory (Lichtenstein & Srull, 1987).

Individuals under an impression goal have a greater
number of associative paths between pieces of information
(which facilitates recall of the information) than
individuals under a memory goal. Researchers suggest that
the two processing goals result in differential recall of
information about a target person (i.e., under an impression
goal, individuals recall more information about a target
person than under a memory goal).

Hamilton, Katz and Leirer (1980) conducted one of the
earliest investigations of the effects of different
processing goals on recall. In three experiments, they
manipulated processing goals when presenting information
about a target person. In all three experiments they found
greater recall of information under an impression goal
rather than a memory goal.

Srull (1983) manipulated processing goals when
presenting information about a target person and examined
the effect on recall and whether recall was moderated by
situational conditions. This was examined by manipulating
the length of delay before recall (immediate or 12 minutes)
and the presentation format of the behavioral information.
Presentation format was manipulated by presenting

information "blocked" by person (i.e., presenting all

8
information about the first target person and then
presenting all information about the second target person)
or in random order. It was hypothesized that in the block
condition, individuals should easily be able to form person
categories from the information presented. This
organization should facilitate recall.

It was also hypothesized that recall under an
impression goal would not be affected by presentation
format. That is, these subjects should organize information
in relation to a target regardless of presentation format.
The associative paths formed for each target in the random
order condition compensate for the lack of natural
categories (which are found in the block condition).

Lastly, Srull (1983) left constant the total number of
behaviors presented, but varied the ratio of target persons
to behaviors per target persons. (These ratios being 2/64,
4/32, 8/16 or 16/8.) This was also expected to moderate the
relationship between processing goal and recall.

The results of this experiment conformed to the
hypotheses. In line with the findings of Hamilton et al.
(1980), this experiment showed that individuals recalled
more information under an impression goal than under a
memory goal. In addition, there was an interaction between
processing goal and presentation format. Recall under an
impression goal was not affected by presentation format, but

recall under a memory goal was greater when information was

9
presented in a block format than when presented in random
order. A three way interaction between processing goal,
presentation format and delay indicated that this difference
was larger when there was no delay before recall. Lastly,
there was an interaction between processing goal and the
ratio of targets to behaviors per target. Recall under an
impression goal increased as the number of targets increased
(and the number of behaviors per target decreased). Recall
under a memory goal increased as the number of targets
decreased (and the number of behaviors per target
increased). Lichtenstein and Srull (1987) manipulated
impression, memory and comprehension processing goals. A
comprehension goal is the most like "natural" observation.
There is no attempt to organize and interrelate incoming
information (as in an impression goal), nor is there an
attempt to rehearse information (as in a memory goal).
Recall under a comprehension goal should be lower than

recall under an impression or memory goal.

Lichtenstein and Srull (1987) were also concerned about
the lack of correspondence between judgments made of a
person and specific information that can be recalled about
that person. They suggested that the processing goal would
moderate the relation between recall and judgment. The
processing goal would affect whether an evaluation is made
at the time of information acquisition (referred to as an

on-line judgment) and prestored in memory or whether an

10
evaluation is computed on the spot from specific information
stored in memory.

For instance, when an individual is under an impression
goal, any information in memory would be stored separately
from the evaluation that is formed. The result would be a
weak relation between the evaluation made and the specific
pieces of information recalled. At the other extreme,
individuals under a comprehension goal would not make an
evaluation at the time of information acquistion, but would
need to compute an evaluation when they are asked for one.
They would have to make this evaluation with whatever
information they can recall; the result being a high
correspondence between the evaluation and the specific
information recalled. The relationship between the
evaluation and the specific information recalled under a
memory goal should be higher than with an impression goal,
but should be weaker than with a comprehension goal.

Lichtenstein and Srull (1987) manipulated three
variables. The first, processing goal, included a
comprehension goal as well as impression and memory goals.
The second, list composition, consisted of whether positive
or negative information was presented first in the list of
behavioral statements. The third, testing order, indicated
whether recall or a judgment was made first. The dependent
variables were the amount of information recalled, the

judgment (likability rating) of the target, and the

11
correlation between judgment and mean scale value of the
information recalled.

As expected, Lichtenstein and Srull (1987) found that
recall under an impression goal was greater than recall
under a memory goal which was, in turn, greater than recall
under a comprehension goal. Examination of the likability
ratings showed a main effect for processing goal. The
target was rated more favorably under an impression goal
than under a memory goal or a comprehension goal. An
interaction effect between processing goal and list
condition on judgments was also found. When positive
information was presented first in the list of behavioral
statements, impression and memory goal individuals rated the
target more favorably, indicating a primacy effect of
information on judgment. When positive information was
presented last, comprehension goal individuals rated the
target more favorably, indicating a recency effect of
information on judgement.

The last dependent variable examined was the
correlation between the judgment and mean scale value of
information recalled. Only under a comprehension goal was
this correlation statistically significant. Testing order
did not have an effect on any of the dependent variables,
nor did it enter into any interactions.

Hastie and Park (1986) were also interested in how

processing goals affect the relationship between memory and

12
judgment. They presented five different models, each of
which tries to explain the relationship between memory and
judgment. These models range from an availability model
(what is available in memory is used to make a judgment) to
a set of three models that assume that judgment causes
memory, to an independence model (assumes no relationship
'between memory and judgment).‘ Hastie and Park's (1986)
examination of these five models led them to conclude that
there are conditions under which one may or may not find a
relation between memory and judgment. A condition they
identified as important was an individual’s processing goal,
and in a series of experiments, they tested how processing
goals affect the relation between memory and judgment. They
found that correlations between memory and judgment were
significant under a memory goal, but the correlations were
not significant under an impression goal. This suggests
that judgments made under a memory goal are memory-based,
whereas judgments made under an impression goal are made as
the information is acquired.

Sherman, Zehner, Johnson and Hirt (1983) also examined
the effects of processing goals on memory and judgment.
After presentation of initial information about two football
teams, individuals were asked to explain a hypothetical

victory by one team. These researchers were interested in

 

1Readers are referred to Hastie and Park (1986) for a
detailed discussion of these models.

13
whether this task would affect a later judgment and the
relationship between this judgment and the information
recalled. It was predicted that one’s processing goal would
influence whether this task would lead to the reconstruction
of the initial information. It was expected that, under an
impression goal, information would be organized and
integrated so that it would resist change from a subsequent
explanation task. Under a memory goal, the information is
not organized, and the subsequent explanation would likely
have reconstruction effects on later judgment and
information recalled. The data supported the predictions.
Under a memory goal, judgments were biased in favor of the
team whose victory was explained. This bias did not occur
under an impression goal. The correlation between memory
and judgment was significant under a memory goal but not
significant under an impression goal.

Recent research in the area of performance appraisal
has used the previous literature to identify when appraisal
decisions are impression-based and when they are memory-
based (DeNisi et al., 1989; Williams et al., 1990). These
researchers varied the the appraisal situation (e.g. the
appraisal purpose differed or the incidence of diary-keeping
differed) and examined the correlations between the amount
of information recalled and the performance ratings. They
concluded that under certain conditions, ratings were

primarily impression-based (the correlations were not

14

significant), while under other conditions, ratings were
primarily memory-based (the correlations were significant).

In summary, the social cognition research indicates
that an individual’s processing goal affects processing of
behavioral information which, in turn, affects the retrieval
of the information. This factor also influences the
relationship between what is recalled and the judgments that
are made. The performance appraisal research indicates that
there are conditions when performance evaluations are
primarily impression-based, and conditions when they are
primarily memory-based. Future research should examine how
one’s processing goal affects the actual performance
ratings.

Level of Information Inconsistency

Social cognition researchers have recently examined the
way in which information about persons is organized in
memory2 (Srull, 1981). Some of the research on person
schemata has been concerned with trying to account for how
923 behavioral information acquired about a target person is
actively integrated with what is already known about a
target person. A main concern has been how an individual
processes behavioral information that is either consistent
or inconsistent with an existing person schema.

Several early studies (Cantor & Mischel, 1977;

 

2The cognitive organization of information about a
person is commonly referred to as a person schema.

15
Rothbart, Evans & Fulero, 1979; Snyder & Uranowitz, 1978),
have reported that behavioral information that is consistent
with an existing person schema is remembered better than
information that is inconsistent. It is suggested that
these findings may be due to an individual’s resistance to
change initial beliefs. Thus, information consistent with
one’s belief will be preferentially encoded and better
remembered than other information (Cantor & Mischel, 1977).

Nevertheless, a series of experiments by Hastie and
Kumar (1979), reported that free recall of information
inconsistent with an existing person schema was greater than
recall of information consistent with the existing person
schema. This effect was found when individuals were
presented with an equal number of consistent and
inconsistent behavioral statements. The effect was even
greater when the number of inconsistent statements within
the list was reduced.

Hastie (1980) proposed an associative network model to
account for these results. Hastie assumed that behavioral
information inconsistent with an existing person schema is
difficult to encode and remains in working memory longer.
This information is thought of in relation to other
information known about the target person in an attempt to
explain its occurence. Consequently, inter-behavior
associations tend to be formed between each inconsistent

behavior and the others (of. Wyer & Srull, 1986).

16
Information consistent with an existing person schema is
easily encoded and integrated with information known about
the target person. Thus, none of the associative paths that
are formed for inconsistent information are formed for
consistent information. This model suggests that the
greater number of associative paths formed with inconsistent
information facilitates its recall. That is, the
probability of retrieving a particular item should be a
function of the number of associative paths that have been
formed during encoding (of. Srull, 1981). This would explain
why Hastie and Kumar (1979) found that a greater proportion
of inconsistent information than consistent information was
recalled.

Srull (1981) has offered a criterion-based explanation
for why research results have been conflicting. Two of the
early studies (Cantor & Mischel, 1977; Snyder & Uranowitz,
1978) tested for recognition rather than free recall of the
information. The retrieval process underlying free recall
is assumed to traverse the associative paths laid down in
the network (of. Srull, 1981). Use of the recognition
procedure bypasses this retrieval process with its
associative paths, which eliminates the advantage that
inconsistent information has in being remembered better than
consistent information. In the early study by Rothbart et
al. (1979), because there were more consistent than

inconsistent behaviors, the absolute number of consistent

17
behaviors recalled was greater. In fact, a greater
proportion of inconsistent information was recalled (Srull,
1981). Higgins and Bargh (1987) have also suggested that
the Rothbart et al. (1979) study’s differing results are
attributable to terminology. That is, the Rothbart et al.
(1979) study reported that "expected behaviors" were better
recalled than "unexpected behaviors". These "unexpected
behaviors" included both behaviors irrelevant and
inconsistent with the expectancy. It is important to make a
distinction between behaviors that are "unexpected" because
they are irrelevant to the expectancy and behaviors that are
"unexpected" because they actually violate the expectancy
(Higgins & Bargh, 1987).

Much of the recent research has offered support for the
Hastie (1980) associative network model. In a series of
experiments, Srull (1981) showed that a greater proportion
of inconsistent than consistent information is recalled. In
two experiments, Srull (1981: Experiments 2 & 3) also
manipulated the composition of the list of information and
examined the effect on recall of type of information. In
one experiment, Srull manipulated list composition by
holding the number of consistent items constant, but varied
the number of inconsistent items. He found that the
proportion of consistent items recalled increased as the
number of inconsistent items increased. In the other

experiment, Srull manipulated list composition by holding

18
the number of inconsistent items constant, but varied the
number of consistent items. He found that varying the
number of consistent items had no effect on the recall of
inconsistent items. Srull (1981) explained how the model
would account for these results. Increasing the number of
inconsistent items facilitated the recall of consistent
information because it led to more associative paths linked
to the consistent information. But increasing the number of
consistent items had no effect on the number of paths
emanating from the inconsistent information and,
consequently, there was no effect on the recall of
inconsistent information.

As Hastie and Kumar (1979) have noted, there are
criterion-based factors that affect whether consistent or
inconsistent information is remembered at a higher rate.
Srull (1981: Experiment 4) made an additional elaboration on
the Hastie (1980) model by identifying another factor that
affects what type of information is remembered. In this
experiment Srull manipulated whether individuals completed
an irrelevant task during the time they are processing the
consistent/inconsistent information. Information was
presented sequentially and individuals rehearsed aloud the
item that was being presented. It was hypothesized that
this irrelevant task would affect the subject’s ability to
form the associative paths between pieces of information and

lead to the recall of inconsistent information losing its

19
advantage over consistent information. The results were in
the hypothesized direction. When individuals did not engage
in the irrelevant task, recall for inconsistent information
was greater than for consistent information. But when
individuals did engage in the irrelevant task, recall of
inconsistent information declined to match the level of
recall of consistent information. Although Srull varied the
number of times (1-3) subjects were to rehearse the items
aloud, this manipulation did not make a difference.

Stern, Marrs, Miller and Cole (1984) have identified an
additional factor that affects whether inconsistent
information is recalled at a higher rate. They hypothesized
that the information will be differentially recalled based
on whether the target is a single individual or a group.
They reasoned that a person would not expect members of a
group to behave consistently, but they do expect an
individual to behave consistently. When presented with
information that is inconsistent, only those individuals who
are forming an impression of an individual will deeply
process this information. Because the recall advantage of
inconsistent information is due to its being deeply
processed, these researchers expected to find greater recall
of inconsistent information for the person who is forming an
impression of a single individual. There should be no
difference in recall of consistent/inconsistent information

for the person who is forming an impression of a group.

20

In one experiment, these researchers manipulated both
level of information consistency and type of target
(individual vs. group), and examined the effect on recall.
The results were as hypothesized. There was a greater
tendency for those individuals who formed an impression of a
single individual to recall inconsistent behaviors. Recall
of consistent and inconsistent behaviors did not differ for
those individuals who formed an impression of a group.

In a second experiment, Stern et al. (1984) included an
examination of the time it took to process the behavioral
information. They reasoned that if the greater recall of
inconsistent information was due to its being deeply
processed, then processing time should be higher for
inconsistent information. The results indicated that
individuals who formed an impression of a single individual
did recall inconsistent information at a higher rate. There
was also a tendency for inconsistent information to take
longer to be processed. This difference between processing
time for consistent and inconsistent information was even
greater when the target was an individual.

O’Sullivan and Durso (1984) have examined the
associative network model from a different perspective.
According to O’Sullivan and Durso (1984), if the issue in
question is how the consistent/inconsistent information
affects the existing schema, then the variable of interest

should be memory of the initial information used in

21
establishing the schema g2; memory of the information that
is consistent or inconsistent with this schema. In two
experiments, O’Sullivan and Durso (1984) held constant the
initial information used in establishing the schema, varied
the subsequent consistent/inconsistent information, and
examined recall of the initial schematic information. It
was predicted that recall of this initial schematic
information should be greater when it is followed by
inconsistent information than when it is followed by
consistent information. It would be assumed under this
model that the subsequent inconsistent information would
cause the initial information to be reconsidered, during
which time associative paths between the inconsistent and
initial information should be formed. Presentation of
subsequent consistent information should not lead to the
reconsideration of this initial information.

In the first experiment, O’Sullivan and Durso (1984)
manipulated three variables: (1) the level of consistency of
subsequent items, (2) the relevance of the subsequent items
("core" or very relevant items vs. "peripheral" or less
relevant items), and (3) the length of delay (none vs. two
days) after presentation of subsequent information, but
before recall of initial information.

In terms of the relevance of information, it was
hypothesized that the facilitating effect of inconsistent

items would be limited to the core-inconsistent items. As

22
suggested by the Hastie (1980) model, it is only those items
that are distinctive (i.e. core-inconsistent items) that are
deeply processed. It is this deep processing that provides
linkages to schematic information and benefits its recall.
Inconsistent items that are less relevant (i.e. peripheral-
inconsistent items) should be considered uninformative; they
should be "filtered out", and therefore, not affect the
schema. Core-consistent items should be remembered well
because "quintessentially characteristic" items (Hastie,
1980) also receive deep processing (of. O’Sullivan & Durso,
1984). However, these items should not benefit recall of
initial schematic information because they should not lead
to the reconsideration of the initial information.

In terms of the length of delay before recall, it was
hypothesized that associative paths would decay at the same
rate across all conditions. In other words, length of delay
should not interact with either level of information
consistency or relevance of items. This hypothesis is also
consistent with the Hastie (1980) model.

O’Sullivan and Durso (1984) used two types of
information to prime the initial person schema. First, they
described the target person in terms of a label
stereotypical of college students (e.g. "jocks", "greeks"),
then they provided two supporting facts attributed to that
target person. Dependent variables of interest were recall

of this initial label and recall of the two supporting

23
facts.

The results of this experiment showed that recall of
the label was not significantly affected by any of the
independent variables. Examination of the recall of the two
supporting facts showed that this information was recalled
at a greater rate when followed by an inconsistent item that
was highly relevant. Length of delay before recall only
entered into a significant main effect (i.e. recall of the
two supporting facts was better after no delay than after a
two day delay).

In the second experiment, O’Sullivan and Durso (1984)
were interested in the effects of Egg highly inconsistent
items on the recall of the initial schematic information.
Therefore, in addition to manipulating level of consistency
of subsequent items (consistent vs. inconsistent), they also
manipulated number of subsequent items (one vs. two).

These researchers offered two hypothesized outcomes.
The first, based on intuition, suggests that two
inconsistent items would disrupt schematic processing,
leading to the schematic information being recalled least
well under this condition. The second, based on the Hastie
(1980) model, suggests that two inconsistent items would
result in the highest recall of schematic information. The
two highly inconsistent items would be deeply processed,
resulting in even more linkages to initial schematic

information. This greater number of linkages will, of

24
course, facilitate the recall of schematic information,
leading to it being recalled best.

As in the first experiment, dependent variables of
interest were recall of the initial stereotypical label and
recall of the two supporting facts. The results of this
second experiment showed greater recall of the label when
followed by inconsistent information than when followed by
consistent information. No other effects that involved
recall of the label were found.

As expected, the results showed greater recall of the
two supporting facts when followed by inconsistent
information than when followed by consistent information.
However, the number of subsequent items did not have a main
effect on recall of the two supporting facts, nor did it
enter into an interaction with level of information
consistency. This result does not conform to either of the
hypothesized outcomes. These researchers do not offer an
explanation for why neither of these hypothesized outcomes
were supported. This author suggests that the lack of
support for the Hastie (1980) model may be due to a ceiling
effect. That is, O’Sullivan and Durso (1984) were examining
recall of only two supporting facts. One piece of
subsequent inconsistent information may have had all of the
possible facilitating effects for recalling this small
amount of information. Additional research which increases

the number of initial items may very well find the greater

25
facilitating effect for two subsequent inconsistent items as
predicted by the Hastie (1980) model.

The research by O’Sullivan and Durso has taken us one
step forward in terms of understanding how the initial
person schema is affected by subsequent consistent and
inconsistent information. At first glance, the Hastie
(1980) model and the early research that has tested it (e.g.
Hastie & Kumar, 1979) seem to suggest that an initial schema
may be easily modified (due to the high availability of
inconsistent information). Yet the results of the preceding
research (that inconsistent information improves
availability of schematic information) showed us that the
model may actually account for the failure of inconsistent
information to modify the initial schema.

Interaction between Processing Coal and
Level of Information Inconsistency

One conclusion reached earlier in this paper was that
there are conditions under which the Hastie (1980)
associative network model does not apply. There are
conditions in which inconsistent information may not be
recalled at a higher rate than consistent information. For
instance, one of these conditions, the use of recognition
versus recall for a memory task, has previously been
addressed. Another one of these conditions seems to be
related to an individual’s processing goal when he/she is

presented with behavioral information about the target

26
person. This leads to an investigation of the theory and
research on the interaction of our two variables of
interest, level of information inconsistency and processing
goal.

It has been suggested that individuals must be under an
impression goal for inconsistent information to be
remembered at a higher rate than consistent information
(Higgins & Bargh, 1987; Srull, 1981). Individuals under a
memory goal should not differentially recall consistent and
inconsistent information. The theoretical explanation, in
terms of the Hastie (1980) model, is that inconsistent items
only undergo additional processing and establish linkages
with other information when individuals are attempting to
form an impression of the target person (Srull, 1981).

Srull (1981: Experiment 1) has examined the effects of
both these variables, level of information inconsistency and
processing goal, in the same study. In this experiment,
Srull also manipulated length of delay before recall (none
vs. two days). It was hypothesized that length of delay
would not interact with either level of information
inconsistency or processing goal. The model assumes that
all associative paths decay at a constant rate across
conditions.

Consistent with predictions, it was found that
inconsistent items were recalled at a higher rate than

consistent items. More items were recalled under an

27
impression goal than under a memory goal. However, there
was also an interaction between processing goal and level of
information consistency. While the recall of consistent and
inconsistent items did not differ under a memory goal, more
inconsistent items were recalled under an impression goal.
In terms of the underlying theoretical model, this suggests
that items inconsistent with a prior impression undergo
additional processing when an individual is attempting to
form an impression (of. Srull, 1981).

An interaction between length of delay and processing
goal was also found. Recall was greater under an impression
goal when there was no delay, but recall was slighty greater
under a memory goal when there was a long delay. Two
explanations can account for these results. Individuals
with a memory goal may have intentionally rehearsed the
items during the two day delay. This would have given them
a spurious advantage over individuals with an impression
goal who did not rehearse the items because they did not
anticipate a recall test after the delay period (Srull,
1981). This delay may have also caused the individual under
a memory goal to form an impression of the target.

In a series of experiments, Srull, Lichtenstein, and
Rothbart (1985) examined both processing goal and level of
information inconsistency. One experiment (Srull et al.,
1985; Experiment 7) was a replication of Srull’s (1981;

Experiment 1) earlier examination of these two variables,

28
however, a delay condition was not included. The results
from this study parallelled those found in the earlier
study.

In two other experiments, Srull et al. (1985;
Experiments 5 and 6) did not prime individuals with an
initial expectancy, but allowed individuals to develop their
own expectancies. Information was presented in blocks
consisting of two consistent, one inconsistent, and two
irrelevant items. It was reasoned that, at first,
individuals under an impression goal would not have a clear
impression of the target, and inconsistent information would
not be recognized as being inconsistent. However, at some
point during information acquisition, a general impression
will be formed and inconsistent information will be
recognized as inconsistent information. On the other hand,
individuals under a memory goal should never distinguish
between consistent and inconsistent information.

Srull et al. (1985; Experiments 5 and 6)3 examined
recall of information across the successive presentation of
blocks of information. As expected, recall of consistent
and inconsistent information was equal across blocks under a
memory goal. Recall of consistent and inconsistent

information was equal for the first few blocks under an

 

3Experiments 5 and 6 differed only in that consistent
information was of a negative nature for experiment 5,
but was of a positive nature for experiment 6. Results
were very similar across experiments.

29
impression goal. But at a certain point, recall of
consistent and inconsistent information began to differ.
Inconsistent information was recalled at a greater and
greater rate from the fifth to the tenth (and last) block of
information. It was suggested that this point of divergence
occurred at the time individuals had formed a clear
impression of the target.
Summary

A review of the research and theory involving
processing goals and level of information inconsistency has
been presented. First the research dealing with processing
goals was reviewed, followed by the research involving level
of information inconsistency. Finally, the research that
dealt simultaneously with both variables was reviewed. A
major limitation within each of the three areas of research
is that it has only focused on how the variables affect
recall of behavioral information. For example, three
studies have examined the effects of processing goal on
recall (Hamilton et al., 1980; Srull, 1983; Lichtenstein &
Srull, 1987), whereas only recent research has begun to
examine the effects of processing goal on variables other
than recall (e.g. Hastie & Parks, 1986; Lichtenstein &
Srull, 1987). The review of the research on level of
information inconsistency indicates that five studies have
examined the effect on recall of the consistent/inconsistent

information (Hastie & Kumar, 1979; Srull, 1981 - Experiments

30
2, 3 and 4; Stern et al., 1984 - Experiment 1), whereas one
study (O’Sullivan & Durso, 1984) has examined the effects on
recall of initial schematic information. Only one study
(Stern et al., 1984; Experiment 2), which examined the
effects on the time it took to process the consistent/
inconsistent information, has focused on a dependent
variable other than recall. The research involving both
processing goals and level of information inconsistency has
pply examined the effects on recall. In order to understand
how these two variables combine to affect processing of
behavioral information, variables other than recall must be
examined.

The intentions of this study are to extend the research
on the interaction between processing goals and level of
information inconsistency by examining their effects on
variables other than recall. The model of performance
appraisal by DeNisi et al. (1984) suggests that factors that
affect the encoding, storage, and retrieval of performance
information should also affect the formal evaluations
assigned to a ratee. The following section presents a
detailed discussion of how an individual’s subsequent

performance evaluations may be influenced by the interaction

between processing goals and level of information

inconsistency.

31
Hypotheses

Research indicates that subsequent information
inconsistent with prior information is recalled at a higher
rate than consistent information when individuals are under
an impression goal. However, it is not clear how this
greater recall of inconsistent information affects an
initial impression. This study explores the effects of
information inconsistency on recall and judgment.

It has been suggested that inconsistent information is
more memorable because it has been more deeply processed.
This suggests that the greater availability of inconsistent
information could contribute to the initial impression being
modified or abandoned. However, O’Sullivan and Durso (1984)
have shown that subsequent inconsistent information
facilitates recall of initial information used to establish
an impression. That is, the inconsistent information causes
the individual to reconsider the information used in forming
the initial impression. These results suggest that
inconsistent information may actually fail to modify the
initial impression. Other research supports this conclusion
(e.g. Srull 1981; Experiment 2).

Previous studies have only made inferences about the
effects of inconsistent information on an existing
impression by examining the effects on recall. A more
direct way of determining whether inconsistent information

does or does not cause an individual to modify an existing

32
impression is by examining the performance ratings that are
made after inconsistent information has been presented.

It is possible to make predictions about how
inconsistencies affect subsequent ratings of a target. As
was stated above, some research has indicated that
inconsistencies may pp; lead to a modification in the
existing impression (O'Sullivan & Durso, 1984; Srull, 1981).
Although it was shown that recall is higher for the
inconsistent information, it was also shown that
inconsistent information facilitated recall of consistent
information. In addition, a number of studies from the
social cognitive literature dealing with the perserverance
of initial impressions lend support to the prediction that
an existing impression is not easily modified (Ross, Lepper,
& Hubbard, 1975; Anderson, Lepper, & Ross, 1980). Although
these studies do not directly examine the effects of
inconsistent information on an existing person schema, they
do examine how individuals’ judgments are affected by
contradictory information. Following is a review of these
studies.

The first study (Ross, Lepper & Hubbard, 1975) to
demonstrate the "perserverance effect" of initial
impressions dealt with the perserverance of self-perceptions
and social perceptions. In this study, individuals were
given false feedback information indicating their success or

failure on a task. At a later point in the experiment, this

33
feedback information was negated through a "debriefing
procedure". Although individuals completely accepted this
subsequent information, when they were later asked to rate
their success or failure on the task, their ratings
continued to be heavily influenced by the initial false
feedback information. This effect was also shown for the
social ratings made by outside observers. That is, these
individuals were rated by others who witnessed both the
initial session when false information was given and the
later debriefing session. The ratings of these observers
were also heavily influenced by the initial false
information.

A second study (Anderson, Lepper & Ross, 1980) extended
the research of the previous study by demonstrating that the
"perserverance effect" occurs even when an individual’s
initial information is minimal and logically inadequate. In
the first experiment, individuals were presented with two
case studies which led them to believe that there was a
relation between risk taking and success or failure as a
firefighter. All individuals also performed a task in which
they offered an explanation for why this relation existed.
In a debreifing session that followed, individuals were
informed that the relationship they had discovered and
explained was actually fictitious. They were also told that
other individuals had been presented with information that

described the opposite relationship. However, when

34
individuals were later asked to make predictions that
involved this relationship, their predictions were heavily
influenced by the initial information. Even though the
information from which individuals formed their initial
belief was minimal and was later negated, individuals still
did not make the appropriate modifications in their beliefs.
This "perserverance effect" also held in the second
experiment in which half of the individuals did not even
participate in the explanation task.

Another study (Sherman, Zehner, Johnson & Hirt, 1983),
reviewed earlier, supported this idea of a "perserverance
effect" but also showed that processing goals moderated the
effect. It was found that initial information was modified
by a subsequent explanation task for individuals under a
memory goal, but the initial information resisted change for

individuals under an impression goal.

Effects on Recall of Subseguent Information
The formation of the hypotheses dealing with the

effects of processing goal and level of inconsistency on
ratings of performance are based on research that examined
the effects of these variables on recall. Therefore, it is
necessary to examine the effects of processing goal and
level of information inconsistency on recall. One goal of
this research is to replicate the results of the studies
that manipulated processing goal and examined the

differential recall of subsequent consistent/inconsistent

35

information (e.g. Srull, 1981: Experiment 1; Srull,
Lichtenstein, & Rothbart, 1985: Experiments 5-7).
Therefore, the following hypothesis is posited:

Hypothesis 1: Recall of inconsistent information

will be higher than recall of subsequent

consistent information for individuals under an

impression goal. There will be no differential

recall for individuals under a memory goal.
Figure 1 illustrates the proposed interaction effect of
processing goal and information inconsistency on the recall
of subsequent information.
Effects on Rgpgll of Initigl Inforpption

It is also important to examine the recall of initial

impression—forming information. The study by O’Sullivan and
Durso (1984) pointed out the importance of examining the
recall of this initial information. They suggested that
subsequent inconsistent information was causing the initial
impression-forming information to be reconsidered. It is
also suggested that the reconsideration of the initial
information results in an individual failing to modify the
initial impression of the target even when presented with
subsequent inconsistent information. To investigate this
possibility, it is necessary to examine the effects of
processing goal and level of information inconsistency on

the recall of initial information. This leads to the

following hypothesis:

36

FIGURE 1

Predicted interaction between Processing Goal and Level
of Inconsistency on Recall of Subsequent Information

Recall of Subsequent Information

HI

 

LO

 

l 1

Consistent Inconsistent

Level of Inconsistency

 

‘9‘ Impression Goal + Memory Goal

 

 

 

37
Hypothesis 2: Recall of initial information will
be higher for the target whose subsequent
information contained inconsistencies than for the
target whose subsequent information did not
contain inconsistencies, for individuals under an
impression goal. There will be no difference in

recall of initial information across targets for
individuals under a memory goal.

Figure 2 illustrates the interaction effect of processing
goal and level of information inconsistency on recall of
initial information proposed by Hypothesis 2.
Effects ongfigtings
The findings which support the existence of a
"perserverance effect" of initial impressions (Ross et al.,
1975; Anderson et al., 1980; Sherman et al., 1983), and the
findings which indicate that inconsistencies may actually
cause initial information to be reconsidered (O’Sullivan &
Durso, 1984) suggest the following hypothesis regarding the
effects of processing goal and information inconsistency on
subsequent judgments:
Hypothesis 3: Ratings of a target whose subsequent
information contains inconsistencies will differ
based on the processing goal of the rater. That
is, inconsistencies will lead to less of a
modification in ratings under an impression goal
than under a memory goal. There will be no
difference in ratings across processing goals
for the target whose subsequent information does
not contain inconsistencies.
This hypothesis posits that a "perserverance effect"
will occur when a target’s subsequent information contains
inconsistencies and when raters are under an impression

goal. This "perserverance effect" is not expected to be as

strong when the raters are under a memory goal. Thus,

38

FIGURE 2

Predicted interaction between Processing Goal and Level
of Inconsistency on Recall of Initial Information

H008" 0f IIIIIISI IMOI‘II‘IEIIOII

 

 

 

III
as
LO
L J
Consistent Inconsistent

Level of Inconsistency

 

‘9‘ Impression Goal + Memory Goal

 

 

 

39
ratings of the target with inconsistencies will differ
according to processing goal. For example, assume that the
initial information for a target indicates low performance.
Subsequent inconsistent information about this target would
indicate high performance. Hypothesis 3 posits that ratings
under an impression goal will be 123;; than ratings under a
memory goal (or, if initial information indicates high
performance, ratings under an impression goal will be higher
than ratings under a memory goal). Failure to modify an
existing impression should be reflected in the ratings of
individuals under an impression goal. This effect will not
occur for the target whose subsequent information does not
contain inconsistencies. Figure 3 illustrates the proposed
effect of processing goal and information inconsistency on
ratings when initial information indicates low performance.
Figure 4 illustrates the corresponding predictions when
initial information indicates high performance.

As is indicated by Figures 3 and 4, the initial
performance level of the targets was varied in the present
study (e.g. reflecting either good or poor performance).
Previous studies have not varied performance level in one
study. Thus, these studies could not determine if
performance level has differential effects on the processing
of performance information. Research on employment
interviews suggests that interviewers place more weight on

negative information than on positive information (e.g.

40

FIGURE 3

Predicted Effect of Processing Coal and Level of
Inconsistency on Ratings when Initial Performance is Low

Performance Ratings

HI

LO

 

L I

Consistent Inconsistent

Level of Inconsistency

 

-9- Impression Goal + Memory Goal

 

 

 

41

FIGURE 4

Predicted Effects of Processing Coal and Level of
Inconsistency on Ratings when Initial Performance is High

Performance Ratings

HI

 

LO

I I

Consistent Inconsistent

Level of Inconsistency

 

-9- Impression Goal + Memory Goal

 

 

 

42
Hollmann, 1972). This may result in a larger "perserverance
effect" when initial information is negative, or less of an
effect when initial information is positive (and/or greater
recall of the negative information). This independent
variable was included in order to determine if performance
level has differential effects on either performance ratings
or amount information recalled. However, given the
exploratory nature of this variable, no formal hypotheses

are proposed.

METHOD

Subjects

A power analysis was conducted with the assumptions of
a medium effect size and an alpha level of .05. It
indicated that a sample size of 160 would result in power of
.80 (Cohen, 1988). One hundred seventy-eight (122 females,
56 males) introductory psychology students at a large
midwestern university participated in the experiment.
Subjects received nominal course credit for their
participation. Five of the 178 subjects did not complete
the recall measures correctly, and two did not complete the
rating measures correctly. The data from these subjects
were not included in the final analyses.
Design Parameters

The research design is a 2 X 2 X 2 factorial with two
between-subject factors (processing goal and initial
performance level of targets) and one within-subject factor
(level of information inconsistency). Figure 5 presents an
illustration of the cells of the proposed design.

Processing goal. Stimulus information was presented in
the context of one of two processing goals. In the

impression goal condition, subjects were told that the

43

44
FIGURE 5

Experimental Design

Processing Goal

 

Memory

(2)

Impression

(1)

 

 

:
I
I
I
Initial Consistent Target I Consistent Target
Performance High I
Level of Inconsistent TargetIInconsistent Target
Targets I
(3) I (4)
Consistent Target I Consistent Target
Low I

Inconsistent TargetIInconsistent Target

45
purpose of the experiment was to examine how people form
impressions of others, and how different modes of presenting
information affect one’s impression. They were told that
they would be presented with several statements regarding
the performance of two police officers, and that they should
use this information to form an impression of the
effectiveness of each officer’s performance. They were told
that they would be asked to make performance ratings based
on this information. In the memory goal condition,
subjects were told that the purpose of the experiment was to
examine how people memorize information, and how different
modes of presenting information affects what is memorized.
They were presented with a series of statements regarding
two police officers and were told to try to memorize the
information as best as they can. They were told that they
would be asked to recall these sentences as it was
presented. A copy of the specific instructions given under
each condition is in Appendix A.

Lgvel of inforpption inconsistency. Subjects were
presented with performance information on two targets.
Subjects were initially presented with 4 pieces of
information for each target. All of this initial
information was consistent for each target. In the
subsequent information that was presented, level of
information inconsistency differed across targets (i.e.

subsequent information for each target either did or did not

46
contain inconsistencies). Subjects were presented with 6
pieces of subsequent information for each target. For one
target, all six pieces of subsequent information were
consistent with the initial information. For the other
target, one-half (3 pieces) of the subsequent information
was consistent with the initial information, and one-half (3
pieces) was inconsistent.

Initial performance level:pf targets. The initial

performance information that was presented indicated that
the targets were either good performers or poor performers.
Half of the subjects received initial information that
suggested that both targets were good performers, and half
of the subjects received initial information that suggested
that both targets were poor performers.
Stimulus Material

In developing the stimuli for this research, an
occupation, police officer, was chosen for which subjects
would have a general level of knowledge sufficient to
distinguish successful from unsuccessful performance, but
not an extensive job domain knowledge. Research has
indicated that job domain knowledge affects the rating
process (Kozlowski, Kirsch, & Chac, 1986) and the recall of
consistent/inconsistent information (Fiske, Hinder, &
Larter, 1983).

Performance information was in the form of critical

incident statements. Each statement addressed a single

47

behavior corresponding to one of four performance
dimensions: communication, demeanor, work attitude, and
dealing with others. These statements, selected from Landy
and Farr (1975), were scaled on a nine-point metric by a
student sample (Kozlowski a Ford, in press). Incidents with
mean ratings exceeding seven and falling below three were
used to indicate high and low performance, respectively.
For each of the two targets, a set of 10 statements (4
initial and 6 subsequent) were presented. In addition,
because initial performance level of targets was
manipulated, two sets of statements were complied for each
target - one set when initial performance was high and one
set when initial performance was low. These four sets of
performance statements (2 for each target), along with the
mean scale value and standard deviation of each statement,
are presented in Appendix B. Previous research indicates
(e.g. Srull, 1981; 1983; Stern et al., 1984) that the amount
of information (20 behavioral statements) presented to
individuals is a sufficient amount to allow for variability
in recall, yet is not overburdening to subjects.
Procedure

Subjects were run in one of 16 groups (4 groups for
each of the 4 experimental conditions). Groups consisted of
8-14 subjects. All subjects in each group were run under
the same experimental condition. Each group was randomly

assigned to one of the four experimental conditions (see

48
Figure 5).

Upon arrival to the experiment, subjects were asked to
read and sign a consent form. A copy of the consent form is
presented in Appendix C. The experimenter then explained
the purpose of the experiment, and subjects were presented
with both written and verbal instructions. The instructions
explained to the subjects differed depending on whether
subjects were assigned to an impression goal condition or to
a memory goal condition. The specific instructions for each
condition are found in Appendix A. In the impression goal
conditions, subjects were instructed to use the information
presented to form an evaluation of each police officer’s
performance, with the ultimate goal of completing
performance ratings for each target. In the memory goal
conditions, subjects were instructed to memorize the
information that was presented, with the ultimate goal of
recalling as much information as possible. All subjects,
regardless of condition, were told that information would be
presented on slides, and that they would have 6 seconds to
view each slide. (Previous research indicates that 6
seconds is sufficient time to process these statements -
e.g. Lichtenstein & Srull, 1987)

All subjects saw two example slides before the
presentation of actual stimulus material. (EX: (1) Officer X
is frequently charged with making false arrests. (2) Officer

X behaves in a "just do the job" manner.) Individuals under

49
an impression goal were asked to use the example statements
to make a rating of Officer X. Individuals under a memory
goal were asked to recall these statements.

If subjects had no questions, the presentation of the
initial stimulus information occurred. This information was
presented one statement at a time through the use of a slide
projector. Subjects had 6 seconds to view each slide.

Order of presentation was randomized. Subjects were
presented with initial information for two police officers.
After presentation of this initial information, subjects in
the impression goal conditions were asked to make initial
ratings of the effectiveness of each target’s performance.
(This was to insure that these subjects have formed an
impression before the presentation of inconsistent
information.)

At this point in the experiment, subjects were asked to
fill out a questionnaire relating to different modes of
presenting information. This task was intended to act as an
interference task between presentations of stimulus
material. The interference task was included so that the
presentation of the initial information would be distinct
from the presentation of the subsequent information. This
task took approximately 10 minutes to complete. A copy of
this questionnnaire is presented in Appendix D.

Subjects were then presented with the subsequent

stimulus information. Subjects were instructed to attend to

50
the information as they had earlier. The procedure for
presenting this information was the same as for the initial
information. Following the presentation of the second set
of information, all subjects completed two measures. The
order in which subjects filled out these measures was
counterbalanced across subjects. Order of presentation was
not expected to have a differential effect on ratings or
recall. The measures described below are presented in
Appendix E.
Dependent Measures

Performance ratings. Subjects were asked to use all

the information that was presented to rate the effectiveness
of each officer’s performance. The rating form consisted of
a 9 point Likert type scale (1=very ineffective performer,
9=very effective performer).

Recall measure. Subjects were asked to recall all of
the information they could remember about each officer.
They were instructed to try to recall this information as it
was presented, but that they should also record what they
could recall even if it was only the general meaning of the
statement. Responses on the recall measure were coded
according to a "general meaning" criterion by two judges
blind to the experimental conditions. Coding consisted of
determining whether each statement recalled by the subjects
matched one of the statements that was presented. Because

each performance statement described a very distinct

51
behavior, matching those recalled with those presented was
self—evident. However, discretion was used by the judges to
determine if the statement recalled actually portrayed the
"general meaning" of what was presented. That is, a
statement did not have to be word for word to be classified
a match, it had to exhibit the same idea as the statement
that was presented. For example, one statement that was
presented was "Officer Red insults and bullies a father in
front of his family". A statement recalled by one subject
that was identified as a match to the previous statement was
"Officer Red puts down a man in front of his family". One
statement that was recalled which could not be matched to
any of the statements that were presented was "Officer Blue
doesn’t take his job seriously". Although this idea was
implicit in many of the statements that were presented, this
statement could not be directly matched to one specific
statement.

Of the total of 556 statements that were recalled, the
two judges initially agreed that 507 statements matched one
of the statements that was presented (interjudge agreement
of 91%). Disagreements occurred when one judge adopted a
more stringent decision rule than the other judge (i.e. was
less willing to classify the recalled statement as a
"match"). Subsequent consensus reflected the more

stringent decision rule.

The statements that were recalled were classified into

52
one of six categories: (1) target 1-initial, (2) target 2-
initial, (3) target l-subsequent, (4) target 2-subsequent/
consistent, (5) target 2-subsequent/inconsistent, and (6)
mismatch (statements attributed to the wrong target) or
intrusion. The two judges agreed completely (100%) on the
classification of the recalled statements into one of these
six categories.

Completion of the measures above marked the conclusion
of the experiment, at which time subjects were fully
debriefed regarding the nature of the experiment, and any
questions they had about the experiment were answered. A
copy of the debriefing statement given to subjects at the

completion of the experiment can be found in Appendix F.

RESULTS

Initial analyses were conducted to determine if testing
order (recall-ratings vs. ratings-recall) had any effects on
the dependent variables. This factor, which was
counterbalanced across subjects, did not have a significant
effect on any of the dependent variables nor did it enter
into any interactions. A summary of the analyses with
testing order as an independent variable is presented in
Appendix G. Subsequent analyses were made without testing
order as an independent variable.

To determine if a separate analysis of variance (ANOVA)
for each dependent variable was justified as a research
strategy, a test for multicollinearity was conducted by
examining the correlations between dependent variables.
These correlations are presented in Table 1. Also included
in Table 1 are the means and standard deviations of the
dependent variables. (The rating variables are further
defined according to initial performance level since
collapsing over this variable would yield means and standard
deviations that are uninterpretable.) The correlations
between dependent variables are quite low suggesting that

separate analyses are warranted.

53

54
Table 1

Means, Standard Devigtions and Intercorrelgtions

Between Dependent Variables

 

 

Correlations

Eeriehls Mean SD 1 E i i i Q
1. Ratingl' -

HI perf 7.77 1.27

LO perf- 2.20 2.11
2. Rating2 .21 -

HI perf 5.24 2.06

LO perf 4.48 1.28
3. Initiallb .84 .85 -.05 -.19 -

4. InitialZ 1.01 1.08 -.18 -.23 .34 -
5. SubCon° 1.40 .90 .10 -.01 .18 .21 -

60 SUbIDOOD 1087 092 038 -005 012 012 023 -

 

'Ratingl = Performance ratings of target with no
inconsistent information (Target 1/Red)

Rating2 = Performance ratings of target with inconsistent
information (Target 2/Blue)

bInitiall = Initial information recalled for target with
no inconsistent information (range = 0-3)
Initialz = Initial information recalled for target with
inconsistent information (range = 0-3)

cSubCons = Subsequent/consistent information recalled for
target with inconsistent information
(range = 0-3)
SubIncon = Subsequent/inconsistent information recalled for
target with inconsistent information
(range 0-3)

55
Recall

To test hypotheses 1 and 2, the amount of subsequent
consistent/inconsistent information recalled and the amount
of initial information recalled were analyzed with a 2 X 2 X
2 analysis of variance (two between-subject factors:
processing goal and initial performance level and one
within—subject factor: level of inconsistency). The results
of the ANOVA summarized in Table 2 are described below.

Sppgequent inforpption. Because this analysis was
concerned with the differential recall of subsequent
consistent/inconsistent information, only recall regarding
the target whose subsequent information contained
inconsistencies (Target Z/Officer Blue) was included in this
analysis. That is, the variable "level of inconsistency"
refers to how subsequent information inconsistency varied
within a single target (pieces of information were either
consistent or inconsistent).

The results of the analysis regarding the amount of
subsequent consistent/inconsistent information recalled are
summarized in column 3 ("Subsequent Information") of Table
2. The results revealed significant main effects for
processing goal, initial performance level, and level of
inconsistency. Two interactions were also significant: the
processing goal by level of inconsistency interaction, and
the initial performance level by level of inconsistency

interaction. Means and standard deviations for amount of

ANOVA Summary for Recall Variables

56

Table 2

 

Source
Processing Goal (G)

Initial Performance
Level (P)

G X P

Within (S/GP)

Level of
Inconsistency (I)

G X I
P X I

G X P X I

Within (Is/GP)

169

169

E-values for Recall Variables

 

Subsequent
Information

13.05sss

26.7533:

1.27

(.86)

33.42***
3.951
10.041!

.00

(.58)

Initial

Interestisa

7.9588

3.53
.87

(1.17)

3.24
4.431
2.85
6.36ss

(.61)

 

 

8p < .05
832 < .01
.001

Values in parentheses represent mean squares terms.

57
subsequent information recalled are reported in Table 3.

Hypothesis 1 predicted an interaction between
processing goal and level of inconsistency. The results
indicated that this interaction was significant and was in
the direction predicted. More subsequent inconsistent
information was recalled under an impression goal (M = 2.11)
than under a memory goal (Ll = 1.63) [$41,111, = 3.48, p <
.01). However, recall of subsequent consistent information
did not differ across processing goals (fl = 1.49 and 1.31
for impression and memory goals, respectively, EU,HH =
1.30, p > .05).

No hypotheses were made regarding the effects of
initial performance level. However, results revealed a
significant interaction between initial performance level
and level of inconsistency. The results indicated that when
subsequent information was consistent, recall was not
significantly different across performance level (fl = 1.52 h
1.28 for high and low initial performance, respectively,
EUIUU = 1.78, p > .05). However, when subsequent
information was inconsistent, information indicating poor
performance (M = 2.25) was recalled at a higher rate than
information indicating good performance (M;= 1.51) [Luann =
5.79, p_< .01].

Initial information. This analysis was concerned with
how recall of initial information differed when a target’s

performance information contained inconsistencies versus

58
Table 3

Means and Standard Deviations' for Amount of
Subseguent Inforpption Recalled

 

 

Initial Performance Level

 

Processing Goal High Low Total

 

Inconsistent Information

Memory 1.96(.93) 1.29(.89) 1.63(.97)
Impression 2.50(.50) l.70(.81) 2.11(.81)
Total 2.25(.82) 1.51(.87) 1.87(.92)

 

Consistent Information

Memory 1.38(.91) 1.24(.99) l.31(.96)
Impression 1.69(.80) 1.32(.88) 1.49(.84)
Total 1.52(.87) 1.28(.92) 1.40(.90)

 

'Values in parentheses are standard deviations

59
when a target’s performance information contained only
consistent information. That is, the variable of "level of
inconsistency" refers to how information inconsistency
varies across targets (a target’s subsequent information
does or does not contain inconsistencies).

The results regarding the amount of initial information
recalled are summarized in column 4 ("Initial Information")
of Table 2. The results revealed a significant main effect
for processing goal. Two interactions were also
significant: the processing goal by level of inconsistency
interaction, and the processing goal by initial performance
level by level of inconsistency interaction. Means and
standard deviations for the amount of initial information
recalled are reported in Table 4.

Hypothesis 2 predicted an interaction between
processing goal and level of inconsistency. The results
indicated that this interaction was significant and was in
the direction predicted. That is, under an impression goal,
recall of initial information was higher for the target with
inconsistent information (M = 1.28) than for the target with
only consistent information (M_= .92) [EUJU = 2.73, p <
.01]. However, under a memory goal, recall of initial
information for the target with inconsistent information (M
= .77) did not differ from the target with only consistent

information (M = .75) [EUJU = 0.21, p > .05].

60

Table 4

Means and Standard Deviptions; for Amount of

Initial Information Recglled

 

Initial Performance Level

 

 

 

Processing Goal High Low Total
Target with Inconsistencies
Memory .62 (.81) .88(1.11) .75 (.98)
Impression 1.03(l.11) 1.49(1.10) 1.28(1.12)
Total .81 (.82) 1.20(1.14) 1.01(1.08)
Target with No Inconsistencies
Memory .57 (.72) .97 (.90) .77 (.83)
Impression 1.05 (.86) .81 (.88) .92 (.87)
Total .79 (.82) .89 (.89) .84 (.85)

 

'Values in parentheses are standard deviations

61

The results also indicated a three-way interaction
between processing goal, initial performance level, and
level of inconsistency. The form of this interaction is
depicted in Figure 6. When initial performance level was
low, the interaction between processing goal and level of
consistency was as predicted. That is, recall of initial
information differed for the target with inconsistent
information such that more information was recalled under an
impression goal (M = . 1.49) than under a memory goal (M =
0.88) [pan = 2.59, p < .05]. Recall for the target with
no inconsistent information did not differ across processing
goal (M = 0.81 and 0.97 for impression and memory goals,
respectively, LUJU = - 0.89, p > .05).

When initial performance level was high, results
differed slightly from what was predicted. As was
predicted, more initial information was recalled under an
impression goal (M = 2.01) than under a memory goal (M =
1.20) [£UJU = 2.60, p < .05]. However, contrary to
predictions, individuals under an impression goal did not
recall more initial information for the target with

1.03) than for the target with no

inconsistencies (M

inconsistencies (M 1.05) [LUJU = - 0.18, p > .05], rather

initial information was recalled at the same rate for both

targets.

62

FIGURE 6

Processing Goal x Initial Performance Level X Level of
Inconsistency Interaction for Recall of Initial Information

Recall of Initial Information

1.6 '

1.4 r

 

0.6 r

0.2 "

 

 

O I I
Consistent Inconsistent

Level of Inconsistency

 

‘9- lmpression/Low part “I" Memory/Low perf
‘0' Impression/High pert '*' Memory/High pert

 

 

 

63
Perforppnce Rgtipgg

To test hypothesis 3, performance ratings were analyzed
with a 2 X 2 analysis of variance (one between-subject
factor of processing goal and one within-subject factor of
level of inconsistency). Two separate ANOVAs were used to
analyze the effects of performance ratings: the first when
initial performance level was low, the second when initial
performance level was high. These analyses were conducted
separately because of the effect that initial performance
level has on ratings (i.e. when initial information is low,
targets should be rated lower overall, and when initial
information is high, targets should be rated higher
overall).

When initial performance level was low, the results
revealed significant main effects for both independent
variables: processing goal (EUJH = 8.09, p < .01) and level
of inconsistency (EUJU = 119.70, p < .001). The
interaction between processing goal and level of
inconsistency was also significant (EUJU = 14.11, p <
.001). This observed interaction is illustrated by Figure
7. Comparison of Figure 7 (observed results) with Figure 3
(predicted results) shows that the observed results mimic
the predicted results. That is, when initial performance
level was low, the performance ratings of the target whose
information was consistent were not affected by processing

goal (M = 2.13 and 2.16 for impression and memory goals,

64
FIGURE 7

Observed Interaction of Processing Goal and Level of
Inconsistency on Ratings when Initial Performance Is Low

Performance Ratings

 

l l

Consistent Inconsistent

Level of Inconsistency

 

F—e- Impression Goal + Memory Goal

 

65
respectively, EUJU = - 0.13, p > .05). However, the
performance ratings of the target with inconsistent
information was affected by processing goal. The target was
rated lower under an impression goal (M = 3.62) than under a
memory goal (M = 5.21) LENA“ = - 3.77, p < .01].

When initial performance level was high, the results
also revealed significant main effects for both independent
variables: processing goal (EUJU = 9.52, p < .01) and level
of inconsistency (EUJH = 94.35, p < .001). The interaction
between processing goal and level of inconsistency was also
significant (EUJU = 9.84, p < .001). This observed
interaction is illustrated by Figure 8. Comparisons of
Figure 8 (observed results) with Figure 4 (predicted
results) shows that the observed results did not completely
conform to the predicted results. The performance ratings
of the target whose information was consistent were as
predicted (the ratings were not affected by processing goal,
M = 7.80 and 7.73 for impression and memory goals,
respectively, EUJU = 0.25, p > .05). However, the observed
ratings of the target with inconsistent information were
opposite to what was predicted. That is, it was
hypothesized that this target would be rated higher under an
impression goal than under a memory goal. The results
indicate that the reverse occurred. The target actually was
rated higher under a memory goal (M = 6.05) than under an

impression 8081 (M. = 4061) [It—(1’84, = - 3074, E < 001].

66
FIGURE 8

Observed Interaction of Processing Goal and Level of
Inconsistency on Ratings when Initial Performance Is High

Performance Ratings
10 r

 

 

o l l
Consistent Inconsistent

Level of Inconsistency

 

-€- Impression Goal + Memory Goal

 

DISCUSSION

The purpose of this research was to examine the effects
of processing goal and level of information inconsistency on
the processing of performance information. The first
objective of this research was to replicate previous
findings concerned with the effects of processing goal and
level of inconsistency on recall. The second objective was
to extend the research to the area of performance appraisal
by examining the effects of these variables on performance
ratings. These objectives are reviewed in the following
discussion. First, the results of the present study are
examined and in relation to the hypothesized results. The
implications of the results for theory and research in
performance appraisal are then discussed. Finally, the
potential limitations of the study and directions for

further research are presented.

Summary of Results

It was proposed that processing goal and level of
information inconsistency would interact to affect three
different dependent variables (recall of subsequent
information, recall of initial information, and performance

ratings). Three separate hypotheses were made regarding

67

68
each dependent variable. The following sections examine the
extent to which the observed results support each of these
hypotheses.

Effectg on recgll of subseguent inforpption. One goal
of this research was to replicate the results of the studies
that manipulated processing goal and examined the
differential recall of subsequent consistent/inconsistent
information (e.g. Srull, 1981: Experiment 1; Srull, et al.,
1985: Experiments 5-7). It was hypothesized that recall
would be greatest when subsequent information was
inconsistent and when individuals were under an impression
goal (Hypothesis 1). The data support this hypothesis,
thus, lending even greater support to the previous
literature by extending the results from the area of social
cognition to the area of performance appraisal.

An additional finding of this research concerned the
interactive effect of initial performance level and level of
inconsistency on recall of subsequent information. The
results showed that when subsequent inconsistent information
indicated poor performance, it was recalled at a higher rate
than when it indicated good performance. However,
subsequent consistent information was recalled at the same
rate across performance levels. This finding suggests an
elaboration of the Hastie (1980) associative network model.
This model states that inconsistent information is recalled

at a higher rate because the difficulty in encoding

69
inconsistent information leads to a greater number of
associative paths being formed during the encoding process.
The results of the present research suggest that
inconsistent information that is negative is even more
difficult to encode than positive inconsistent information.
Thus, the greatest number of associative paths are formed
(and recall is the greatest) when information is
inconsistent Eng when it indicates poor performance. This
idea is also consistent with the research on employment
interviews which suggests that interviewers place more
weight on negative information than on positive information
(e.g. Hollman, 1972). However, the present study further
qualifies the research by suggesting that only when this
negative information is inconsistent with the initial
positive information will it be weighted higher.

Effects on recall of initigl information. As was
suggested by O’Sullivan and Durso (1984), it was
hypothesized that subsequent inconsistent information would
cause individuals under an impression goal to reconsider the
initial impression-forming information. Hypothesis 2
proposed that recall of initial information would be
greatest for the target whose subsequent information
contained inconsistencies and when individuals were under an
impression goal. The support for this hypothesis is
qualified since a third variable (initial performance level)

interacted with processing goal and level of inconsistency

 

70
to affect recall of initial information.

The data show that when initial performance level was
low, the results were as hypothesized. However, when
initial performance level was high, the initial information
for both targets (one with inconsistencies and one without
inconsistencies) was recalled at the same high rate by
individuals under an impression goal. These results suggest
that, when individuals are under an impression goal,
subsequent inconsistent information (which indicates poor
performance by an initially good performer) possibly leads
to a reconsideration of all initial information. These
results compliment the discussion above which suggests that
the greatest number of associative paths are formed for
inconsistent information indicating poor performance.
Subsequent information indicating poor performance by an
initially good performer may be so difficult to encode that
p_l initial information is reconsidered, leading to more
associative links between the initial information and the
subsequent negative information. The greater number of
links facilitated recall of both the initial information and
the subsequent negative information.

The effects of processing goal and level of
inconsistency on recall of initial information were
examined, in part, to provide support for the hypothesis
dealing with the effects of these variables on performance

ratings. It was suggested that reconsideration of the

71
initial impression-forming information would cause
individuals under an impression goal to fail to modify their
initial impression. Thus, the mixed support for Hypothesis
2 has implications for the following hypothesis and will be
discussed in further detail below.

Effects on performance ratings. Hypothesis 3 focused
on the effects of processing goal and level of inconsistency
on performance ratings. Based on the research on
"perserverance effects" in social judgments (Ross et al.,
1975; Anderson et al., 1980; Sherman et al., 1983), and the
research on the reconsideration of initial impression-
forming information (O’Sullivan & Durso, 1984), it was
hypothesized that individuals under an impression goal would
fail to modify their initial impression of a target whose
subsequent information contains inconsistencies.

Support for Hypothesis 3 was mixed. When initial
performance level was low, the data supported the
hypothesis. Because initial impressions influenced the
ratings under an impression goal (perserverance effect), the
target with inconsistent information was rated lower under
an impression goal than a memory goal. However, when
initial performance level was high, the results were
opposite to what was predicted. Due to a perserverance
effect, it was predicted that the target would be rated
higher under an impression goal than under a memory goal.

The results indicated that the target actually was rated

72

lower under an impression goal than under a memory goal.
Interpretation of Results

The results involving performance ratings as well as
the results involving recall indicate that performance level
is clearly having an impact on the processing of performance
information. When initial performance level was low, the
results for all three dependent variables were as
hypothesized. Only when initial performance level was high,
did the results differ from the hypothesized results.
Results from all three dependent variables suggest that the
processing of the performance information is different than
hypothesized when initial performance indicates good
performance. One explanation for these findings is that,
when faced with inconsistencies, raters under an impression
goal are weighting good and poor information differently
than raters under a memory goal. It is important to point
out, however, that these differences only occurred when the
target’s information contained inconsistencies. When the
target’s information was consistent no differences in
weighting occurred. This target was rated exactly the same
by all raters when performance level was high and when
performance level was low.

The results of this research suggest that information
indicating poor performance is given much more weight by
raters under an impression goal. When the initial

information indicated poor performance, subsequent good

73

performance did not lead to a change in the initial
impression (perserverance effect of initial poor
performance). However, when initial information indicated
good performance, subsequent poor performance led to the
initial impression being modified. Subsequent poor
performance also led to the reconsideration of all initial
information (perhaps as an attempt to verify one’s change in
initial impression).

The results also suggest that information indicating
good performance is given much more weight by raters under a
memory goal. This is attested by the fact that when initial
performance was both high and low, the target was rated
higher by raters under a memory goal than by raters under an
impression goal. As hypothesized, when initial performance
was low, raters under a memory goal were more willing to use
subsequent inconsistent information indicating good
performance. The result being a higher rating than under an
impression goal. However, when initial performance was
high, raters under a memory goal were less willing to use
subsequent inconsistent information indicating poor
performance. The result being that ratings were much higher
than was predicted. In conclusion, it appears that if
individuals thought the purpose of the task was to memorize
the information, when they were later asked to make ratings,
they gave less weight to negative information, resulting in

more lenient ratings. However, when individuals thought the

74
purpose of the task was to use the information to make
ratings, they gave more weight to negative information,
resulting in less lenient ratings. The implications of
these findings and the other findings are discussed in
detail in the following section.
lpplicgtiong of regplpp

The most striking result of the present research is
that, even though all raters were presented with the same
information upon which to base their ratings, the target
with inconsistent information was rated differently by
raters under different processing goals. In addition, the
amount and type of information that was recalled differed
greatly across processing goals. These results have a
number of important implications to the research and
practice of performance appraisal.

The results can be reviewed in terms of the performance
appraisal model proposed by DeNisi et al. (1984). This
model suggested that there are factors which affect the
processing of performance information. This study provides
support for this model. Results indicate that three
contextual variables (processing goal, level of information
inconsistency, and initial performance level) interact to
affect retrieval of performance information and the
evaluations that are based on this performance information.
In addition, the social cognition literature suggests that

the link between the contextual variables and subsequent

75
recall and judgments is actually mediated by the processing
variables of encoding and storage. That is, the contextual
variables lead to differences in encoding and storage which,
in turn, lead to the observed differences in recall and
judgments.

A second implication of the present research concerns
the many different situations in which performance
appraisals may occur. Researchers should be aware that
performance ratings will differ depending on whether they
are made on-line (impression-based) or computed from the
information in memory at the time of the evaluation (memory-
based). Efforts should be taken to identify what conditions
cause ratings to be memory-based and which conditions cause
ratings to be impression-based. As was indicated earlier,
recent research has examined the correlations between recall
and ratings to determine if the ratings are impression or
memory-based (DeNisi et al., 1989; Williams et al., 1990).
DeNisi et al. (1989) were able to identify a number of
situations in which evaluations were impression-based (e.g.
when the purpose of evaluation was to indicate how deserving
each ratee was for a raise), and a number of situations in
which evaluations were memory—based (e.g. when the purpose
was to determine which ratee would receive the raise). The
results of the present research suggest that, if the ratings
were based on the exact same information and this

information contained inconsistencies, the ratings would

76
differ based on the conditions under which they are made.
In addition, the information that is recalled will also
differ (leading to the information used in performance
feedback to be different).

If researchers were able to identify whether the
ratings were impression or memory-based, then they would be
able to determine the way in which the performance ratings
are being biased. The present research indicates that very
little information was recalled when raters were under a
memory goal, and that any distortions in ratings were most
likely the result of memory biases. The results also seem
to suggest that raters under a memory goal gave less weight
to negative information, in fact, it appeared as if they
gave more weight to positive information. A possible
explanation for this finding concerns the amount and type of
information that was recalled by these individuals. For the
target with inconsistencies, even less initial information
(M = .74) than subsequent information (M = 2.94) was
recalled. If the ratings of those under a memory goal are
based only on the information that is recalled, then these
ratings are based primarily on subsequent information. This
subsequent information, half of which indicated poor
performance and half of which indicated good performance
regardless of initial performance level, was not
differentially recalled based on its inconsistency. Thus,

it can be concluded that a recency effect in recall may have

77
caused the ratings under a memory goal to appear more
lenient. That is, biases in memory (raters are only able to
recall recent information) may have possibly led to
distortions in the performance ratings. The implication is
that, under conditions when evaluations are memory-based,
efforts should be taken to insure that raters are using more
than the most recent performance upon which to base their
ratings.

Results showed that recall of subsequent inconsistent
information was higher under an impression than under a
memory goal, suggesting that memory (or retrieval) biases
are not occurring under an impression goal. It is most
likely that biases under an impression goal occur when
information is encoded and stored in memory. For instance,
when initial performance level was low, less modifications
in the initial impression were made by raters under an
impression goal. Efforts can be taken to make raters aware
of the biases that can occur when they are faced with
inconsistent information (e.g. perserverance of initial
impression causing them to disregard the inconsistent
information).

Another implication of these findings concern how
raters deal with information that indicates good versus poor
performance. When faced with inconsistencies, raters under
an impression goal gave more weight to information

indicating poor performance. This finding was consistent

78

with the research on employment interviews which also found
that more weight was given to negative information
(Hollmann, 1972). It is suggested that raters believe
negative information has greater diagnostic value, and thus
is given more weight. However, the belief that negative
information is more important may lead to very large
distortions in ratings. Not only is negative information
weighted more in evaluations, but, as is indicated by
onlowski and Ford (in press), raters may also actively
search out incidents of poor performance. The implication
is that the raters’ processing of performance information
will be inefficient because they spend so much time
searching for and processing the negative information, while
disregarding information indicating good performance.
Limitationg of the Stpgy

There are limitations to the generalizability of the
present research which have implications for future
research. First, the conditions under which evaluations were
memory-based versus impression-based were artifically
created by the researcher. They were also manipulated so
that the condition was either one or the other. It is
likely that these different conditions do occur in the
context of performance appraisal. In fact, because
supervisors usually observe their subordinates over a long
period of time (six months to a year), it is likely that

they switch from goal to goal throughout the process of

79
acquiring performance information. As was indicated
previously, future research should try to identify when
evaluations are primarily impression-based and when they are
primarily memory-based. For instance, when a supervisor’s
job is very complex and other duties take up a large portion
of their time, time pressures may cause them to have to make
evaluations that are primarily memory-based (i.e. they would
not have time to form an impression as information is
acquired). In addition, future research should examine how
the amount of delay between the acquisition of (1) initial
and subsequent information and (2) the information and the
ratings affects the processing of performance information.

A second limitation of the present research is that the
performance information was given to the raters. Raters
were not actively able to acquire information about each
target. It has been suggested that assigning the rater to
the role of passive receiver limits the generalizability of
the present research (Kozlowski a Ford, in press). Recent
research has employed the process-tracing methodology so
that raters may be allowed to actively search for
performance information. If biases in performance ratings
are occurring at the acquisition and encoding stages as was
suggested above, then future research should be directed
towards examining how processing goal, inconsistent
information, and performance level affect the search for

subsequent performance information.

80

Another limitation on the generalizability of the
present research concerns the operationalization of
"inconsistency". The initial information that was presented
unequivocally portrayed the target as either a very good or
a very poor performer. Subsequent inconsistent information
was made to be evaluatively opposite to the initial
information (i.e. if initial information indicated good
performance then inconsistent information indicated poor
performance). However, this is only one way that
inconsistency may be operaticnalized (e.g. initial
information could indicate both good and poor performance,
and inconsistent information would indicate only poor
performance). Further research is needed to determine if
the effects found in the present research will generalize to
other operationalizations of inconsistency.

Finally, the present research only examined how a
constant number of inconsistent pieces of information
affected performance ratings. Results showed that three
pieces of inconsistent information caused raters under an
impression goal to modify their initial impression when the
inconsistent information was negative, but to fail to modify
when the inconsistent information was positive. Future
research should be directed at determining how little
negative information causes a modification in initial
impression, and how much positive information is needed

before a modification is made in the initial impression.

LI ST OF REFERENCES

LIST OF REFERENCES

Anderson, C.A., Lepper, M.R., & Ross, L. (1980).
Perserverance of social theories: The role of
explanation in the persistence of discredited
information. Journal of Personality and Social
Psychology, pp, 1037-1049.

Bernardin, H.J., & Walter, 0.8. (1977). Effects of rater
training and diary keeping on psychometric errors in
ratings. Journal of Applied ngcholcgy, pg, 64-69.

Cantor, N., & Mischel, W. (1977). Traits as prototypes:
Effects on recognition memory. Journal of Personality
gnd Social ngchology, l5, 38-48.

Cohen, J. (1988). Statlgtical Power Analysis for the
Behavioral Sciences. Hillsdale, NJ: Laurence Erlbaum.

DeNisi, A.S., Cafferty, T.P., & Meglino, B.M. (1984). A
cognitive view of the performance appraisal process: A
model and research propositions. Organizational
Behavior and Human Performance, pp, 360-396.

DeNisi, A.S., Robbins, T., 8 Williams, K.J. (1989).
Retrieval vs. computational memory models: What we
remember vs. what we use. Paper presented at the 4th
Annual Conference of the Society for Industrial and
Organizational Psychology, Boston, MA.

DeNisi, A.S., & Williams, K.J. (1988) Cognitive approaches
to performance appraisal. In G.R. Ferris a K.M. Rowland
(Eds.), Research in Personnel and Human Resources
Management (Vol. 6). Greenwich, CT: JAI Press.

Fiske, S.T., Kinder, D.R., & Larter, W.M. (1983). The novice
and the expert: Knowledge-based strategies in political
cognition. Journal of Experimentgl,Soclgl Ps cholo ,
lg, 381-400.

Hamilton, D.L., Katz, L.B., & Leirer, V.O. (1980). Cognitive
representation of personality impressions:
Organizational processes in first impression formation.
Jougngl of PersonalitygllSocigl Ps cholo , pp, 1050-63

81

82

Hastie, R. (1980). Memory for behavioral information that
confirms or contradicts a personality impression. In
R. Hastie, T.M. Ostrom, E.B. Ebbesen, R.S. Wyer, Jr.,
D.L. Hamilton, & D.E. Carleton (Eds.), Pepgon Megory:
The Cognitive Basis of Socigl Perception; (pp. 121-
153). Hillsdale, NJ: Erlbaum.

Hastie, R., a Kumar, P.A. (1979). Person memory: Personality
traits as organizing principles in memory for
behaviors. Journal of Personality and Social
Psychology, 31, 25-38.

Hastie, R., a Park, B. (1986). The relationship between
memory and judgment depends on whether the judgment

task is memory-based or on-line. Ps cholo ical Review,
33, 258-268.

Higgins, E.T., & Bargh, J.A. (1987). Social cognition and
social perception. Annppl Review of ngchology, 33,
369-425.

Hollmann, T.D. (1972). Employment interviewers’ errors in
processing positive and negative information. Journal
9: Applied ngcholpgy, 33, 130-134.

Ilgen, D.R., & Feldman, J.M. (1983). Performance appraisal:
A process focus. Research in Organizational Behavior,
3, 141-197.

Kozlowski, S.W.J., a Ford, J.K. (in press). Rater
information acquisition processes: Tracing the effects
of prior knowledge, performance level, search
constraints, and memory demand. Organizational Behavior
and Human Decision Processes.

Kozlowski, S.W.J., Kirsch, M.P., & Chao, G.T. (1986). Job
knowledge, ratee familarity, conceptual similarity and
halo error: An exploration. Journal of Applied

Psychology, ll, 45-59.

Landy, F.J., a Farr, J.L. (1975). Policg Performance
Appraisal. University Park, PA: Pennsylvania State
University.

Lichtenstein, M., & Srull, T.K. (1987). Processing
objectives as a determinant of the relationship between
recall and judgment. Journal of Experimental Social

Psychology, 33, 93-118.

83

O’Sullivan, C.S., & Durso, F.T. (1984). Effect of schema-
incongruent information on memory for stereotypical
attributes. Journpl70f Pergonality 9nd Socigl

PSIChOIOEI’ 11, 55-700

Ross, L., Lepper, M.R., a Hubbard, M. (1975). Perserverance
in self-perception and social perception: Biased
attributional processes in the debriefing paradigm.
Journgl of Pepgonplityypnd Socigl ngchology, 33, 880-
892.

Rothbart, M., Evans, M., & Fulero, S. (1979). Recall for
confirming events: Memory processes and the maintenance
of social stereotypes. Journal of Experipental Socigl

Psychology, l3, 343-355.

Sherman, S.J., Zehner, R.S., Johnson, J., & Hirt, E.R.
(1983). Social explanation: The role of timing, set,
and recall on subjective likelihood estimates. Journal
of Pergonality and Social ngchology, 33, 1127-1143.

Snyder, M., & Uranowitz, S.W. (1978). Reconstruction the
past: Some cognitive consequences of person perception.
Journal of Perponplity 3,Sccial Psycholpgy, 33, 941-50

Srull, T.K. (1981). Person memory: Some tests of associative
storage and retrieval models. Journal of Experimental
Psycholqu: Human Learning and Mgpory, 1, 440-463.

Srull, T.K. (1983). Organizational and retrieval processes
in person memory: An examination of processing
objectives, presentation format, and the possible role
of self generated retrieval cues. Journal of
Personality 9nd Social ngchology, 11, 1157-1170.

Srull, T.K., Lichtenstein, M., & Rothbart, M. (1985).
Associative storage and retrieval processes in person
memory. Journal of Experipentgl ngchology: Learning;
Mgpory. 9nd Cognition. ll, 316-345.

Stern, L.D., Marrs, 8., Millar, M.G., a Cole, E. (1984).
Processing time and the recall of inconsistent and
consistent behaviors of individuals and groups. Journal
of Personality and Socipl Ps cholo , 11, 253-262.

Williams, H.J., Cafferty, T.P., a DeNisi, A.S. (1990). The
effect of performance appraisal salience on recall and
ratings. Organizationgl Behgviorvpnd Human Decision
Processes, 33, 217-239.

Wyer, R.S., a Srull, T.K. (1986). Human cognition in its
social context. Psychological Review, 33, 322-359.

APPENDI CES

APPENDIX A

Instructions for Processing Goal Conditions

84

APPENDIX A

Instructions for Processing Goal Conditions

PERFORMANCE EVALUATION TASK

Instructions:

The purpose of this experiment is to examine how people form
impressions of others. We are also interested in how
different ways of presenting information affects the
impression that is formed. For instance, some ways of that
information can be presented are with print, pictures, or
video, etc.

You are going to be presented with a number of slides which
contain a series of statements. These statements concern
the performance of two police officers. You are to use
these statements to form an impression of the performance of
two police officers. You will be presented with four (4)
pieces of information about each officer. After
presentation of this information you will be asked to rate
the effectiveness of each officer’s performance.

You will use a scale like the one below to make your ratings
of the effectiveness of each officer’s performance.

1 2 3 4 5 6 7 8 9
: ----- ' ----- : ----- ' ----- : ----- : ----- : ----- : ----- :
very somewhat neither somewhat very
ineffective ineffective effective effective effective
or
ineffective

Before we begin the actual experiment let’s do an example.
Please attend to the following slides.

Now, use the rating scale above to rate the effectiveness of
Officer X’s performance. Circle the number on the scale
above which corresponds to your rating.

Again, your task is to form an impression of each officer so
that you can make a rating of his/her performance.

Are there any questions before we begin?

85

SENTENCE MEMORIZATION TASK

Instructions:

The purpose of this experiment is to examine how people
memorize information. We are also interested in how
different ways of presenting information affects what is
memorized. For instance, some ways of that information can
be presented are with print, pictures, or video, etc.

In this experiment, you are going to be performing in a
memorization task. You are going to be presented with a
number of slides which contain a series of statements.
There are eight (8) slides, each of which contains a
sentence that you are to memorize as best as you can. You
will later be asked to recall each sentence as it was

presented to you.

 

Before we begin the actual experiment let’s do an example.
Please attend to the following two slides.

Write in the space below the statements that you were just
presented. Try to recall the important aspects of each
sentence.

EXAMPLE:

 

 

 

 

Now, as the experimenter reads back the statements, check
what you have written.

Because memorizing 8 sentences may be difficult, all of the
sentences will concern two police officers. We hope that
grouping the sentences this way might help you to memorize
the information.

Again, at a later point in this experiment you will asked to
recall each sentence as closely as possible.
Are there any questions before we begin?

APPENDIX B

Stimulus Material: Performance Information

86

APPENDIX B

Stimulus Material: Performance Information

PERFORMANCE STATEMENTS FOR OFFICER RED (consistent/high
perf)

Initial Statements

Item Mean

Officer Red always gets to the station in
time to check the daily log 7.53

Officer Red turns in reports that are neat,
accurate, and well written 8.00

Officer Red does what is assigned and
occasionally initiates own activities 7.47

Officer Red maintains friendly relations
with the civilians in the patrol area 8.37

Subsequent Statements

Item Mean

Officer Red answers urgent calls from
other precincts when necessary 7.67

Officer Red remains cool under any circumstances 8.00

Officer Red has a highly developed vocabulary
and uses it appropriately 7.37

Officer Red is very polite and understanding
while in contact with the public 8.00

Officer Red takes the time necessary to

establish friendly relations with the public such

as showing youth in the patrol area the patrol

car, the equipment, and answering questions 7.77

Officer Red is considered a "friend to
everyone" during the watch or shift 7.10

1.45

1.71

87

PERFORMANCE STATEMENTS FOR OFFICER RED (consistent/low perf)

Initial Statements
Item

Officer Red disappears when a dangerous
situation occurs.

Officer Red uses choppy, incomplete
language in written reports

Officer Red gripes a great deal and considers
the work to be just a job

Officer Red always riles fellow officers
through actions and remarks

Subsequent Statements
Item

Officer Red panics upon receiving an
emergency call

Officer Red responds to calls, but often takes
the long way to get there, possibly hoping
other officers will arrive first

Officer Red has unsatisfactory grammar and is
unable to communicate well with others or
interact with the public

Officer Red only goes through the motions
of the job

Officer Red will "go out of the way" to
defy departmental regulations

Officer Red insults and bullies a father in
front of his family

Mean

1.40

2.77

Mean

1.57

1.67

1.41

1.43

1.17

1.35

1.88

1.34

88

PERFORMANCE STATEMENTS FOR OFFFICER BLUE (inconsistent/high

perf)

Initial Statements
Item

Officer Blue only uses a minimum of sick
days each year

Officer Blue speaks slowly and carefully
when testifying in court

Officer Blue shows interest in the job and
continually seeks ways of improving self

Officer Blue is very patient when dealing with
citizens who speak little English

Subsequent Statements (Consistent)
Item

Officer Blue reports for duty even if he
has a cold if trouble is expected in the city

Officer Blue attends special training schools
during off-duty hours to improve efficiency and
effectiveness as a law enforcement officer

Officer Blue has the rare ability to deal with
all people fairly and is well liked by officers,
supervisors, and the public

Subsequent Statements (Inconsistent)
Item

Officer Blue drinks alcoholic beverages on
duty causing performance to deteriorate

Officer Blue talks in an unintelligble
manner over the radio

Officer Blue gets along with other officers
on the shift but does not relate well with
the public

SD

1.99

1.23

1.78

1.36

89

PERFORMANCE STATEMENTS FOR OFFICER BLUE (inconsistent/low

perf)

Initial Statements
Item

Officer Blue drinks alcoholic beverages on
duty causing performance to deteriorate

Officer Blue talks in an unintelligble
manner over the radio

Officer Blue refuses training because
he sees himself as already an expert

Officer Blue embarrasses fellow officers
with horseplay while in public
Subsequent Statements (Consistent)

Item

Officer Blue cracks in dangerous situations
and tends to shout at other officers

Officer Blue fails to recognize and correct
own deficiencies without prompting by others

Officer Blue antagonizes superiors and fellow
officers by starting rumors about them

Subsequent Statements (Inconsistent)
Item

Officer Blue only uses a minimum of sick days
each year

Officer Blue speaks slowly and carefully
when testifying in court

Officer Blue shows interest in the job and
continually seeks ways of improving self

Mean

1.63

SD

1.51

1.05

APPENDIX C

Consent Form

90

APPENDIX C

Consent Form

For this research project you will be asked to examine a
number of pieces of information. This information will be
presented to you with a slide projector. Specific
instructions will be given on how to attend to this
information. After presentation of this information, you
will be asked to fill out a questionnaire relating to
different ways that information can be presented. You will
also be asked to fill out questionnaires regarding the
information that you were presented. Instructions will be
given on how to fill out these questionnaires.

This experiment requires one hour to complete and
participation in the experiment is voluntary. While you
participation will provide you with extra class credit in
your psychology course, a decision not to participate will
no affect your course grade. You also have the right to
discontinue your participation in the experiment at any time
for any reason without penalty.

All results from your participation will be treated with
strict confidence and all of your performance records will
remain anonymous. Within these restrictions, the final
results of the experiment will be availabe to you upon
written request.

You will also be fully debriefed at the conclusion of the
experiment. Any questions that you may have regarding the
research will be answered at that time.

I have read and understood the above statement. The tasks
involved in this research have been explained to me. I will
consent to participate in this experiment without waiving my
right to discontinue my participation in the experiment at
any time without recrimination.

 

Signature of Student

Experimenter: Lisa Wood
14 Baker Hall
355-2171

APPENDIX D
Interference Task: Questionnaire Dealing with Different

Modes of Processing Information

91

APPENDIX D

Interference Task: Questionnaire Dealing with Different
Modes of Processing Information

INSTRUCTIONS: Fill out the following questionnaire which
concerns different ways of presenting information. You will
have 15 minutes to complete this questionnaire. Record your
answers in the space provided. If you need additional
space, you may continue answers on the back.

On the average how many billboards do you think occur in one
highway mile?

 

On the average, how many TV commericals do you think occur
during one hour of prime time TV?

 

Name as many TV news reporters as you can.

 

 

 

Name as many Radio announcers as you can.

 

 

 

Name as many Photographers as you can.

 

 

 

Name as many Newspaper reporters as you can.

 

 

Name

92

as many Painters as you can.

 

 

 

What

books have you read in the past month?

 

 

 

What

magazines have you read in the past month?

 

 

 

What

movies have you seen in the past month?

 

 

 

APPENDIX E

Measures of Dependent Variables

93

APPENDIX E

Measures of Dependent Variables

RATING TASK

Instructions:

Now that you have been presented with two sets of statements
concerning Officers Red and Blue, we would like you to use
all of this information to make a rating of each officer.
You will be rating each officer on the effectiveness of
his/her performance as a police offficer. Use the 9-point
scales (1=very ineffective, 9=very effective) displayed
below to make your ratings.

Rate the effectiveness of Officer Blue’s performance.
Circle the number on the scale below which corresponds to

your rating.

1 2 3 4 5 6 7 8 9
I I I I I I I I I
I """"" I ’’’’ I """ I ““““ I """ I """ I """ I
very somewhat neither somewhat very
ineffective ineffective effective effective effective
or
ineffective

Rate the effectiveness of Officer Red’s performance. Circle
the number on the scale below which corresponds to your
rating.

1 2 3 4 5 6 7 8 9
I _____ I _____ I _____ I _____ I _____ I _____ I _____ I _____ I
I I I I I I I I
very somewhat neither somewhat very
ineffective ineffective effective effective effective
or

ineffective

94

RECALL TASK

Instructions:

You have been presented with two sets of information
concerning Officers Red and Blue (10 pieces of information
about each officer). Now you are to recall each piece as
best as you can. Try to recall each statement as it was
presented, however, it is important that you write down
whatever you can recall even if it is just a few words or a
phrase from each statement. If you can only recall the
general meaning of a sentence, then that is what you should
record.

Record what you are able to recall in the space provided
below. Begin each new statement on a new line. You will
have 15 minutes to complete this task.

 

 

 

 

 

 

 

 

 

 

 

 

APPENDIX F

Debriefing Statement

95

APPENDIX F

Debriefing Statement

The purpose of this experiment is to examine how
individuals process performance information. In this
experiment, performance information was presented for two
police officers. One-half of the subjects were told to
memorize this information so that they could later recall
it, and one-half were told to use this information to form
an impression of the officers so that they could later make
a rating of their performance. Later in the experiment, all
subjects (regardless of their initial instructions) were
asked to both recall the information and use it to make
performance ratings. This experiment was interested in
whether these different initial instructions caused
differences in (1) what was recalled and (2) the ratings
that were made.

In this experiment, the performance information was
presented in two parts. A questionnaire concerning
different ways of presenting information was filled out by
subjects after the first part and before the second part.
This questionnaire was only to act as a distractor between
the two presentations. It will not be used in any analyses.

If you like further information about this experiment,

please contact: Lisa Wood
14 Baker Hall 355-2171

APPENDIX G

Results of Analyses with Testing Order
as an Independent Variable

96

APPENDIX G

Results of Analyses with Testing Order

as an Independent Variable

Table G-l

ANOVA Summary for Recall Variables

 

Source
Testing Order (0)

O X Initial Performance
Level (P)

O X Processing Goal (G)

O X P X G

Within (3/OGP)

O X Level of
Inconsistency (I)

O X P X I
O X G X I

O X P X G X I

Within (lg/OPS)

165

165

E—values for Recall Variables

 

 

Subsequent Initial
.75 .00
2.17 .16
.08 19
3.99 1.40
(.85) (1.19)
.06 2.04
1.22 .97
.45 .00
4.01 .02
(.58) (.61)

 

Note: Values in parentheses represent mean

squares terms.

97
Table G-2

ANOVA Suppary for RplipgfiVariableg

 

fi-values for Rating Variables

 

Source 3: L0 Perf‘ HI Perf
Testing Order (0) 1 .20 .15
O X Processing Goal (P) 1 .64 1.64
Within (3/OP) 82l86]b (3.72) (2.44)

O X Level of

Inconsistency (I) 1 1.09 2.03
o x P x I 1 .00 .35
Within (I3/OP) 82186]b (1.95) (2.79)

 

Note: Values in parentheses represent mean squares terms.

aLO Perf = Analyses on ratings when initial performance
level indicated poor performance

HI Perf = Analyses on ratings when initial performance
level indicated good performance

bValues outside of brackets are the degrees of freedom for
analyses of L0 Perf - values in brackets are the degrees of
freedom for analyses of HI Perf.

 

 

[Heisman
33918 UBBIquW
AHVU 811

 

 

        

GRN STRTE UN V.

I LIBRRRIES
ii iilliil Iii Iii iiililliliililiil
6884

nICHI
"fl129300895