THE STANDARD OF OBJECTIVITY FOR INTERNAL
AUDITORS: A LABORATORY EXPERIMENT






BY






R. DAVID PLUMLEE


A DISSERTATION PRESENTED TO THE
GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY



UNIVERSITY OF FLORIDA

1982


























Copyright 1982

by

R. David Plumlee















ACKNOWLEDGEMENTS

I acknowledge the generous contribution of my

dissertation committee members: John Lynch, Bill Messier,

Dan Smith, and Doug Snowball. Particular thanks are

extended to Doug Snowball who demonstrated great wisdom and

patience in his role as chairman of the committee. I

appreciate David Shields for contributing his time and

effort in defining and refining the topic and the methods

employed here.

I appreciate the generous financial support of both the

School of Accounting and the Institute of Internal Auditors.

Thanks are extended to the internal auditors who gave their

time and effort as participants in the study, and to Linda

McDaniel and Ralph Fisher who dilegently coded the data.

Rob Muirhead and John Dattola, who arranged for the firms to

provide participants, made an invaluable contribution.

A stable and peaceful personal life is beyond value

when writing a dissertation. I thank my wife, Kathy, and

my parents for all their efforts to keep my domestic world

tranquil. Finally, credit must be given to Bart Ward for

convincing me that I was destined for an academic life and

helping me to reach that goal.















CONTENTS

ACKNOWLEDGMENTS . . . . . . . . iii

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

ABSTRACT . . . . . . . . ... . .viii



I. INTRODUCTION . . . . . . . . 1

Statement of the Problem . . . . . . 2
Scope and Methodology . . . ....... 4
Memory and Attributional Bias Effects .. .. 4
Experimental Method . . . . . . . 5
Organization of the Dissertation . . . . 7

II. RELATED LITERATURE . . . . . . . . 8

Memory and Problem-Solving Behavior . . . 9
Schemata and Their Role in Memory . . . 9
Schemata as E ing Mechanisms. . . ... 11
Schemata as Retrei al Mechanisms . . .. .13
Problem-Solving Beha ior and Memory . . .16
Causal Attribution . . . . . . . .. 21
Causal Schemata Defined . . . . . 22
Spontaneous Attributions . . . . .. 24
Empirical Studies of Causal Attribution . .. 27
Memory and Causal Attribution Research in
Accounting . . . . . . . . 30
Summary . . . . . . . . . . 32


III. EXPERIMENTAL METHODS . . . . . . . 34

Experimental Design . . . . . . . 35
Subjects . . . . . . . . .. ... . 36
Procedure . . . . . . . . . 37
Experimental Context and Task .. . .... 38
Experimental Context . . . . . ... 38
Experimental Task . . . . . . ... 39
Research Hypotheses . . . . . .... 42
Hypothesized Effects of Schemata . . ... 42
Hypothesized Causal Attributions . . ... 44









IV. DATA ANALYSIS AND RESULTS . . . . . ... 46

Tests for Schematic Structuring . . . .. 46
Operationalization of "Idea Unit" . . .. 47
Results of Memory Effect . . . . .. 49
Tests of Causal Attributions . . . . .. 55
Analysis of the Attribution Ratings . . .. 55
Results of Causal Attributiona . . . .. 57

V. CONCLUSIONS AND LIMITATIONS . . . . . . 70

Summary of the Experimental Study . . . . 70
Effects of Memory on the Review Process . .. 71
Causal Attribution Effects on the Review
Process . . . . . . . .. 72
Conclusions . . . . . . . . ... 73
Limitations . . . . . . . . . 75


Appendix

A. Design Materials

B. Design Materials

C. Review Materials

D. Review Materials

BIBLIOGRAPHY . .

BIOGRAPHICAL SKETCH


for the OWN

for the OTH

for the ERO

for the OWN


and ERO Conditions. ..

condition . . . .

and OTH Conditions. ..

Condition . . . .
















LIST OF TABLES


Table page

1. The Malfunctions Used in Causal Attribution Ratings 41

2. Number of Idea Units Identified in Each Group . 50

3. Total Number of Idea Units by Condition . . .. .51

4. Chi-Square Tests for HI, H2, and H3 . . . .. .53

5. Chi-Square Goodness-of-Fit Test for H4 . . .. .54

6. Pearson Correlations Between ICRAT, OTHFACT, and
EFFECT . . . . . . . . . . 57

7. Multivariate Analysis of Variance on ICRAT,
OTHFACT, and EFFECT . . . . . . ... .58

8. ANOVAs for Each of the Dependent Variables .... .59

9. Pair-wise Comparisons of Condition Means . . . 60

10. Univariate ANOVAs for Each Malfunction . . .. .62

11. Means by Malfunction for ICRAT, OTHFACT, and EFFECT 64

12. Correlations by Condition for "High" and "Low"
Effect. . . . . . . . .. .68














Abstract of Dissertation Presented to the Graduate
Council of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy


THE STANDARD OF OBJECTIVITY FOR INTERNAL
AUDITORS: A LABORATORY EXPERIMENT

By

R. David Plumlee


December 1982

Chairman: Douglas Snowball
Major Department: School of Accounting

A laboratory experiment was conducted which examined

the relative objectivity of internal auditors who reviewed

their own work. The goal of this study was to assess the

effects of two factors which were hypothesized to influence

the review process. One factor relates to the notion of

schemata as a memory structure which mediates a person's

access to previously stored knowledge. The other factor is

the potential biases which may arise from different

causal schemata employed by those who review their own work.

Thirty-nine practicing internal auditors participated as

subjects and were required to complete a two-stage task.

The first stage required each subject to flowchart and

document either a mail-order sales/receivable control system

or a purchasing system. Two-thirds of the subjects designed









a sales/receivables system and one-third designed a

purchasing system. The second stage of the task required

subjects to "review" a sales/receivables system by (1)

listing the strengths and weaknesses, and (2) providing

causal attribution ratings for ten system malfunctions

(these ostensibly occurred during the operation of systems

similar to that under review). One half of those subjects

who designed a sales/receivables system in the first stage

were chosen randomly to review their own system in the

second stage (OWN condition). The design developed by each

subject in the OWN condition was also reviewed by two other

subjects. One had designed a sales receivables system and

therefore was experienced with the system under review (ERO

condition). The other subject had designed a purchasing

system in the first stage and therefore was not experienced

with respect to the type of system under review (OTH

condition). The principal results were (1) subjects in the

ERO condition rated the controls as weaker than did subjects

in the OWN condition, as expected, (2) contrary to

expectations, subjects in the ERO condition rated the effect

of the malfunctions on the financial information produced by

the system lower than did subjects in the other two

conditions, (3) the OWN condition was better than either of

the other two conditions at identifying strengths in the

system they reviewed, and (4) the ERO condition was better

than either of the other two conditions at identifying

weaknesses.














CHAPTER I

INTRODUCTION

Internal auditors play an important role in the

internal control system established by corporate management.

This role requires them to participate in many diverse

activities including operational audits and the design of

internal control systems. As a consequence, an internal

auditor may be placed in situations where he is required to

audit an internal control system he designed (or assisted in

designing). Operational responsibilities such as

participation in systems design may have adverse effects

upon an internal auditor's ability to be independent and

objective when subsequently assessing the operations of the

system. The impact of the internal auditor's dual role on

the independence and objectivity of his audit judgment is

the subject of the research in this dissertation.

The Standards for Jth Professional Practice Di Internal

Auditing describe the standards of independence and

objectivity as follows: "Independence permits internal

auditors to render the impartial and unbiased judgments

essential to the proper conduct of audits. It is achieved

through organizational status and objectivity" (Section









100.01). At the same time the standard acknowledges that

operational responsibilities may impair the internal

auditor's objectivity and suggest the following remedy:

Internal auditors should not assume operating
responsibilities. But if on occasion management
directs internal auditors to perform nonaudit
work, it should be understood that they are not
functioning as internal auditors. Moreover,
objectivity is presumed to be impaired when
internal auditors audit any activity for which
they had authority or responsibility. This
impairment should be considered when reporting
results.

Persons transferred to or temporarily engaged by
the internal auditing department should not be
assigned to audit those activities they previously
performed until a reasonable time has elapsed.
Such assignments are presumed to impair
objectivity and should be considered when
supervising the audit work and reporting the
audit results. (Standards for the Professional
Practice Df Internal Auditing Section 120.02)

However, organizational factors (in particular, small

or highly specialized internal auditing staffs) may mean

that there are no practical alternatives to an internal

auditor reviewing his previous nonaudit work. This suggests

that research which examines the impact of operational

responsibilities upon audit assessments should be of

practical significance.


Statement of the Problem

Internal auditors are seldom likely to engage in

intentional misrepocting of the findings of an audit.

However, misreporting may occur as the result of the manner

in which they process information. Two effects of prior









operational responsibilities upon the judgments of internal

auditors are the primary focus of this research. These

effects are (1) the impact of what the internal auditor

remembers about the system from his previous experience and

(2) biases displayed in the review process that may result

from his perceptions of the system and its implementation.

The first of these effects relates to the role of

memory. The internal auditor's mental representation of the

control system may have either a facilitative or a

constraining effect on his audit work. On one hand, the

internal auditor's memory of the internal control system may

allow a better interpretation of the strengths and

weaknesses of his design. On the other hand, it may result

in a "fixation" on the system as it was originally designed.

The latter may inhibit his ability to recall general

knowledge about control systems and thus dysfunctionally

limit consideration of alternatives to his design.

The second effect concerns the potential biases

inherent in reviewing one's own work. The errors or

malfunctions observed by an internal auditor in reviewing a

system may be attributed to any number of potential causes.

An internal auditor observes only the result (for example, a

failure to bill for goods shipped to customers) of a

malfunctioning control system. From such observations, the

internal auditor must infer the causes of the malfunction. A

consequence of an internal auditor having designed the









system under review may be that he tends to attribute the

cause of malfunctions to sources other than the system's

design. The underlying question is whether an internal

auditor retains sufficient objectivity to critically analyze

his own product.

The purpose of this dissertation is to empirically

examine the effects of the internal auditor's memory of the

internal control system and the effects of possible

perceptual biases in situations where the internal auditor

is required to audit a system that he previously designed.

These effects were examined in a two-stage laboratory

experiment which involved participation by internal

auditors.


Scope and Methodoloav of the Dissertation

Though the phenomenon of reviewing one's own work may

be common to many accounting and auditing situations, the

present study is limited to the design and review by

internal auditors of a mail-order sales/receivables system.

The specific focus of the dissertation is on how the

internal auditor's review is affected by his relationship to

the internal control system being reviewed.

Memory and Attributional Bias Effects

The experimental methods employed in this study assume

that concepts stored in human memory are affected by some

form of schematic structuring. These structured concepts

are generic data structures known as schemataa" (Rumelhart










and Ortony, 1977). The proposition regarding memory

explored in this dissertation is that the mental

representation developed by an internal auditor while

designing an internal control system will be accessed as a

schema when reviewing the system.

The perceived cause-and-effect relationships among the

internal controls and malfunctions (whether prevented or

not) are assumed to be organized into causal schemata

(Kelley, 1971). Additionally, the causal schemata evoked by

an internal auditor who reviews his own work are assumed to

differ from the causal schema evoked when the work is not

his own. Three dimensions were chosen as constituents of the

causal schemata associated with internal control systems.

The effect of controls (or the absence of controls) in

preventing (or failing to prevent) malfunctions is one

dimension of these schemata. The other two dimensions

chosen were (1) the design of the internal control system,

and (2) any cause of malfunctions other than the design of

the system. The later two dimensions were considered to be

the causal factors.


Experimental Method

The effects of memory and attributional biases were

assessed in a laboratory experiment that used thirty-nine

practicing internal auditors as subjects. The experimental

task involved two parts. The first part required the










subjects to design an internal control system. Two-

thirds of the subjects designed a sales/receivables system

and the other third designed a purchasing system. In the

second part, one-half of the subjects that designed a

sales/receivables system were randomly chosen to review

their own system. Two other subjects also reviewed each of

these systems; one of whom had designed a sales/receivables

system in the first part and another that designed a

purchasing system.

The notion of "idea unit" (Bransford and Johnson, 1972)

was the basis for the experimental methods used to examine

the effects of schemata in a subsequent review of an

internal control system. The idea units in this research

were the controls listed by the subjects in the

documentation of the design and the listing of strengths and

weaknesses in the review. Inferences regarding the effects

of schematic structuring on information stored in memory are

based on the proportion of idea units the subjects

identified in the review relative to the set of possible

controls in the design they reviewed.

Also, as part of the review, the subjects were required

to rate each of the three dimensions of the causal schemata

for a set of malfunctions that were alleged to have occurred

in the system they reviewed. Differences along these

dimensions were examined between subjects who reviewed their

own work and those that did not.









Organization of the Dissertation

The remainder of the dissertation is organized into

four chapters. The relevant prior research is presented in

Chapter II. This discussion includes literature on how

schemata may operate in memory and the possible relationship

between schematic representation and problem-solving

behavior. Various types of causal schemata and the biases

that may result from an internal auditor reviewing his own

work are also discussed in Chapter II.

Chapter III describes the specific hypotheses drawn

from this body of research and describes the experimental

method used. Details of the experimental design, subjects,

procedure, and the task are also included. The methods of

analysis and the results of the study are presented in

Chapter IV. In addition, this chapter elaborates how the

notion of idea unit was operationalized in this study and

relates the results to the research hypotheses. The final

chapter provides a summary of the results, discusses the

conclusions of the study, and highlights the limitations

inherent in the methodology.

















CHAPTER II

RELATED LITERATURE


The cognitive aspects of reviewing one's own work may

be divided into two categories: (1) the effects of what the

individual remembers from his original involvement in the

system's design and (2) the possible biases that might occur

when the individual attempts to explain any operating

deficiencies that have been identified in that system.

Literature related to the aspects associated with reviewing

one's own work will be discussed in this chapter. The first

section contains a discussion of some relevant research from

the psychology literature which relates problem-solving

behavior and the roles of learning and memory. The second

section examines literature that addresses the process of

attributing causes from observed effects. This process,

know as causal attribution, has been extensively studied by

researchers in psychology. The limited number of studies

from the accounting literature which deal with topics

related to reviewing one's own work is presented in the last

section of this chapter.










Memory and Problem-Solvina Behavior

This section discusses a subset of the literature on

memory and its implications for the study of problem-solving

behavior. The purpose of the review is to provide a

conceptualization of how knowledge and memory might be

related. The review begins by defining the term "schema"

and then presents empirical evidence on how schemata may

function within the processes of storing and retrieving

information. The discussion of schemata will then be

related to some recent work in problem-solving. Finally, an

example will be constructed which integrates the concepts

introduced in the review with the experimental methods to be

used in the present study.


Schemata and Their Role in Memory

Recently, certain mental representations of knowledge

have come to be known as schemata. Rumelhart and Ortony

(1977) have defined schemata as "data structures for

representing the generic concepts stored in memory" (p.101).

They explain that "a schema contains, as part of its

specification, the network of interrelationships that is

believed to generally hold among the constituents of the

concept in question" (p.101). Thus, whether at a broad

general level or relative to a specific system, the concept

of "internal control system" is likely to reside in an

internal auditor's memory as a set of substructures and a










network of interrelationships among them. A schema is an

incomplete mental representation, that is, not all the

values of the attributes are included. In this

dissertation, the term schemata will refer to the imperfect

representations in memory which can exist on a continuum

from specific to generic concepts.

Hayes-Roth (1977) relates the structure of knowledge in

memory to the process of learning, which is assumed herein

to be analogous to gaining experience. Through the process

of learning, the elementary components of a knowledge

structure (such as control functions in the case of the

concept of internal control systems) become unitizedd"

(Hayes-Roth, 1977) so that they are activated in memory in

an all-or-nothing manner. Evidence of this type of

"unitization" as it relates to experience is found in

comparisons between the performance of masters and

nonmasters in tasks involving reconstruction of chess board

descriptions (Chase and Simon, 1973). For random

configurations of chess pieces on the board, no difference

in reconstruction between the masters and the nonmasters was

observed. However, when the reconstruction involved game-

like configurations the masters performed significantly

better. The difference was attributed to representations of

structured patterns which were stored in the memories of the

masters and accessed as a unit.










In a later work (Thorndyke and Hayes-Roth, 1979), the

relationship between learning and schemata was further

elaborated and extended to include the transfer of

knowledge. They postulated that one feature of the

development of a unitizedd" knowledge structure is that the

ability to transfer knowledge between tasks is a function of

the number of substructures that the tasks have in common.

The ability to transfer knowledge increases initially and

then decreases as the level of learning increases. The

implication for the present research is that internal

auditors are likely to transfer knowledge gained in a

previous task to subsequent related tasks because there will

be many shared substructures.

Schemata appear to facilitate the recall of new

information. However, there is some debate about whether

this should be attributed (1) to the role of schemata in the

encoding of inputs into memory or (2) to their function as

an organizing structure for the retrieval of information

stored in memory. Though determining the role of schemata in

the processing of information is not the purpose of the

present research, an elaboration of this debate will

facilitate understanding of the hypotheses presented in the

next section.

Schemata as Encodina Mechanisms. One of the first

attempts (Bransford and Johnson, 1972) to research the role

of schemata examined the first explanation (functioning in









encoding). Bransford and Johnson made use of paragraphs of

prose which were ambiguous and difficult to comprehend when

read without a context (an explanatory title). The subjects

were all instructed to read the passages for subsequent

recall. The subject's recall was measured by the "idea

units" which the experimenter designated a prior and which

corresponded to sentences, basic semantic propositions, or

phrases contained in the paragraphs. The titles for these

paragraphs were assumed to provide a cue for recalling

information from memory. This information functioned as the

context within which the prose could be organized and stored

as a unit in memory. The task was administered under three

conditions: one where the subjects were not given a title

for the paragraph ("no context"), another where the subjects

received the title prior to reading the paragraph ("context

before"), and a third condition that received the title

after reading the paragraph ("context after"). The results

of this series of experiments showed that the subjects in

the "context before" condition had higher comprehension than

either of the other two groups. Those receiving the context

prior to reading the paragraphs had significantly higher

recall than the other two groups, which did not differ in

their level of recall. These results were interpreted as

evidence of prior knowledge operating as a schema providing

a structure for the new information, and thus functioning in

the process of encoding.









Schemata as Retreival Mechanisms. Anderson and Pichert

(1978) examined the proposition that schemata function in

the retrieval of information stored in memory (the second

explanationn. In addition, they hypothesized that the

material would be assigned importance as a function of the

"perspective" of the reader. Their study also concerned the

recall of prose. More specifically, they used passages

which contained information which would be important to one

or the other of two perspectives. The information which was

important to one perspective was unimportant to the other

perspective. A different perspective was included in the

instructions for each group. The subjects were asked to

recall information twice. One-half of the subjects under

each perspective were told (after the first attempt to

recall the information) that they should change to the

previously unmentioned perspective. The subjects who

switched perspectives were able to recall more of the

information which was important to their new perspective but

unimportant to their old one. The subjects who were not

instructed to change perspectives actually recalled less of

the still unimportant information.

One possible explanation for these results may be that

the subjects who changed perspective failed to report (in

the first attempt to recall) material they considered

irrelevant prior to receiving the new perspective. Verbal

protocols taken post experimentally indicated that the









subjects wrote everything they could recall both times.

Thus, this supports the retrieval explanation. Specifically,

Anderson and Pichert (1978) suggested that many of the

subjects used a "top-down" schema-based retrieval strategy

based on the new perspective provided. That is, the schema

for the new perspective was activated in memory and the

previously unrecallable details of the prose were inferred.

Two other studies (Hasher and Griffin, 1978; Alba,

Alexander, Hasher and Caniglia, 1981) have provided further

evidence that information may be stored in memory when not

structured schematically at encoding. Hasher and Griffin

(1978) employed a manipulation which forced half of their

subjects to reinterpret an entire passage immediately prior

to recall. This was accomplished by telling the subjects

that they had mistakenly received the wrong title and they

were told the "correct" title. The subjects that had their

schematic theme invalidated by receiving a new title at

recall actually recalled more of the material than the

subjects whose title remained the same. They also were less

likely to recall information that was related to the assumed

theme of the story, but was not presented in the passage

they read. Items that are related to the theme but not

present in the material, are intrusions which result from

inferences of schema-consistent material. This again

implies a retrieval effect and also a dysfunctional aspect

of schemata-based retrieval strategies.











A possible reconciliation of the differences in recall

found in the studies cited may be that both reconstructive

and reproductive processes are available in memory. A

reproductive process would include a retrieval mechanism

where specific information could be accessed in memory as a

whole. The process of reconstruction would rely on

accessing general knowledge from which the specific

information would be inferred. The subjects in the

Bransford and Johnson (1972) study may have been faced with

such disjointed inputs that virtually no organization was

possible at encoding thus making recall of "chunks" of

material impossible. The subjects in the Anderson and

Pichert (1978) study may have had passages that were

sufficiently comprehensible that even the "other"

perspective material could be associated mildly at encoding.

The schema related studies cited up to this point have used

recall measures to explore the role of schemata in the

storage and retrieval of information. The processes of

recall and recognition are assumed to impose different

requirements on retrieval. Recall requires some strategy

for searching memory for stored information. Recognition

can be appropriately described as a matching process where

the item to be recognized is present and is matched with the

appropriate memory representation. Thus, differences found

between the ability to recall and recognize the same

material are due to retrieval problems in recall.











The only study found in the literature that employed a

recognition measure to examine the importance of a context-

inducing title for the acquisition of prose (Alba et al.

,1981) reported that the presence of the title had no effect

on recognition. Using the same materials as Bransford and

Johnson (1972), Alba et al. conducted a study in which the

subjects were asked to recognize parts of paragraphs they

read previously. They found that not only did the subjects

in both the "title" and the "no title" conditions recognize

the prose that they had seen, but there was no difference in

the pattern of misrecognition for items that had not been

present. This represents strong evidence that information

can be encoded without the structuring influence of

schemata.


Problem-Solvina Behavior and Memo~v

There is evidence that prior knowleCge also affects

performance in problem-solving tasks (Higgins and Chaires,

198F; Weisberg and Alba, 1981). The implications regarding

recognition and recall are not as well documented in this

research as they are in the prose based research. Einhorn

and Hogarth (1981) have suggested that the definition of the

task should not be limited to the task varibles but should

include the prior experience the person brings to the task

and the biological limitations on attention and memory.

These concerns also apply to the tasks undertaken outside









the laboratory, including the environment faced by

accountants and auditors (Joyce and Libby, 1981). During

the process of problem-solving (e.g. the modification of

an internal control system) the individual develops a mental

representation of the problem which serves as the basis for

the solution (Newell and Simon, 1972). A reasonable

extension regarding the influence of these mental

representations is that subsequent judgments involving the

task variables will be mediated by the representation formed

in the problem-solving event. When faced with a problem,

an individual will frame the problem in terms of the

Icesible decision alternatives (Joyce and Libby, 1981).

This constitutes the "problem space" (Newell, 1979) which is

the mental construct comprised of the states of the space

and the operators over the space.

Specifically for the problem of designing an internal

control system, the states of the problem space are assumed

to be the various configurations of potential control

elements (i.e., cashier, controller, mailroom clerk, and so

forth). The operators over the problem space may include

actions such as reconciliations, verifications, and

approvals. The goal of this problem-solving will be

specified as the achievement of an adequate, efficient

system of controls. The mental representation of the

configuration of control elements and operators which

constitutes the solution to the problem is assumed to bc










what the internal auditor stores in memory as "the internal

control system."

Higgins and Chaires (1978) conducted a study which is

relevant to both the problem-solving and the cognitive

representation aspects of the proposed research. Their

study was based on Duncker's (1945) "candle problem" where

the subject is given a small candle, a box of tacks, a book

of matches and a cardboard "wall." The subject is

instructed to affix the candle to the wall so that it does

not drip wax on the table or the floor. The key to solving

the problem is to perceive the box of tacks not only as a

container for the tacks but also as a platform for the

candle. Subjects, who were run uviCEr ':i c ccriticns in

three experiments, were shown slides of fifteen objects. Ten

of these objects would normally be designated by a phrase,

such as "jar of cherries" (for the slide depicting a jar

containing cherries). The "of" priming condition used

phrases such as the one above, while the "and" condition

used phrases like "jar and cherries." Initially, these

phrases were read by the experimenter as the image of the

object appeared on a screen. Then the subjects were

individually taken into a room where the elements of the

"candle problem" (none of these elements was included in the

slide display) were c. i. table. Fight of ten subjects ir

the "and" condition solved the problem while only two of the

ten subjects in the "of" condition solved the problem. These










results were construded as supporting the contention that

the manner in which a problem is presented can affect

retrieval from memory and, consequently, the possible

solutions. Support is also provided for the notion that

more complex knowledge structures can be accessed in memory

as a whole schema when there is a predisposition to do so

based on a learned relationship.

In a series of studies explicitly examining the role of

memory in providing cues in problem-solving, Weisberg and

Alba (1981) employed several traditional problem-solving

tasks, such as the "nine-dot" problem. The motivation for

this research was to demonstrate that the difficulty people

have in solving these problems is not the result of the

assumptions that are brought to the problem-solving

situation, but rather is a function of their task-specific

prior knowledge. They hypothesized that a "solution

domain", similar to Newell's (1979) "problem space", would

be generated by the problem-solver. The experiments used

various levels of "hints" as to the solution. Some of these

were intended to dispell assumptions; others provided

explicit hints as to the solution. Weisberg and Alba found

that merely providing hints to dispel assumptions failed to

faciit1:te .problen.-solving. This was interpreted as

discomfirming evidence for a "fixatior" hypothesis which

asserts that merely eliminating fallacious assumptions will

bring insight as to the solution. Regarding the









implications of their research for memory they state that

"People apply their knowledge to new problems, and if their

knowledge is not directly useful, they try to produce

something new through a straightforward extension of what

they know. No exotic processes, such as sudden insight, are

involved" (Weisberg and Alba, p.189).

Given the preceding research findings, two expectations

may be formed about the rudimentary cognitive processes that

may occur in the design and subsequent review of an internal

control system by an internal auditor. From exposure to the

problem, the internal auditor will develop a set of

alternatives which will be the problem space. The design of

the system which the internal auditor considers to be the

solution will be stored in memory as the representation of

that internal control system. The internal auditor who

subsequently reviews his own work is thus in an analogous

situation to that faced by the subjects who receive the

title before reading an otherwise ambiguous prose passage or

the subjects that maintained the same perspective. The

schema developed during the original work will mediate in

the process of accessing other information stored in memory.

During the review of an internal control system the

internal auditor will typically have documentation available

for reference. Thus, a cognitive process employed in the

evaluation of the controls present in the system is

recognition. Based on the findings of Alba et al. (1981), no










differences would be expected between the recognition of two

equally qualified internal auditors both of whom had

previous experience with the system being reviewed.

Determining whether the controls are adequate requires

recall of prior knowledge. The internal auditor who is

asked to determine the weaknesses in the system he designed

is likely to maintain his perspective and engage in a

"top-down" schema based search strategy. Therefore, he

might have difficulty recalling controls not contained in

his mental representation relative to the internal auditor

who did not design the system. The internal auditor who did

not design the system will also rely on the prior knowledge

stored in memory. This knowledge may be generic or

specific depending on the nature of his previous experience.

Even those internal auditors who have specific previous

experience may be less likely to rely on a schema-basee

strategy because their perspective has changed or the schema

has become "invalidated." This may enable the internal

auditor who did not design the system to be able to detect

weaknesses more readily.











Causal Attribution
This section contains some relevant literature dealing

with causal attributions. First, a definition of causal

schema is presented followed by a brief summary of a few of

the alternative formulations of attribution theory. A

taxonomy for different types of causal schemata is then

described, prior to presentation of four empirical studies.

The first two empirical studies will show that people engage

in spontaneous attributions; the second two studies

demonstrate the biasing effects which may result from use of

causal schemata associated with stereotypes. Finally, a

hypothetical example is used to relate the notion of causal

attribution to the experiment conducted within the present

research.


Causal Schemata Defined

While the schemata discussed in the previous subsection

dealt with the network of relationships among the

substructures of an entity, the notion of causal schema

extends to the relationships among the actions of entities.

A causal schema is a general conception that a person holds

regarding how certain causes interact to produce a given

kind of effect (Kelley, 1971). The person is assumed to

learn this cause-and-effect relationship through first-hand

observation or through learned, normative beliefs (Kassin,

1979). Thus, the causal schema becomes the basis for










inferences regarding the cause of observed effects or the

basis for predictions of effects given observations on

causes.

Tversky and Kahneman (1980) have considered causal

schemata in the area of judgments under uncertainty. They

defined causal inferences as those which lead from observed

causes to the inference of an effect. Conversely, the

inferences from observed effects to implied causes were

designated as diagnostic inferences. Tversky and Kahneman

found that people could make more accurate judgments of

conditional probabilities and had more confidence in those

judgments when they made causal as opposed to diagnostic

inferences. The implication for the present research is

that, given the nature of auditing, auditors are required to

perform the apparently more difficult task of inferring the

cause from observation of the effect (i.e. diagnostic

inferences).

Several theories of causal attribution that deal

principally with relationships among people (Mizerski,

Golden, and Kernan, 1979) have been developed. Jones and

Davis (1965), for example, focus on person-perception.

They state that the way a person perceives the behavior of

an actor is based on (1) the actor's choice among

alternative actions (2) the uniqueness of the effects given

the actions, and (3) the desirability of the action or

outcome. Jones and Davis (1965) view consensus (a











fundamental notion for all attribution theories) as a prior

probability based on the norms and values that dictate the

expected behavior of others. Their theory suggests

that actor's behaviors that are inconsistent with the

norm, such as seeking an undesirable effect, should be

causally attributed to the actor rather than the

environment. Weiner et al. (1971) applied a similar

attribution model (based on four causal elements) to explain

the causal attributions by an observer regarding failure in

the task performance of an actor.

Frequently, there are many potential causes for a given

effect, and these cause-and-effect relationships are not

limited to human behavior. This gives rise to the need for

causal schema in which multiple learned relationships

operate and interact. Kelley (1971) describes some general

criteria for identifying a number of these multiple cause

types of causal schemata including those with (1) multiple

sufficient causes, (2) multiple necessary causes, (3)

compensatory causes, and (4) graded effects (Figure 1).

Kelley indicates that for a schema based on multiple

sufficient causes the presence of any one of the causes will

produce the effect. If it is necessary for all the

potential causes to be present before the effect occurs, then

it is a multiple necessary schema. When one cause may be

substituted for another to produce a given effect, a
























Cause 2

weak strong


strong E E

Cause 1


weak E



(a) Multiple Sufficient
Causes


Cause 2
weakest strongest

strongest
E E E


E E
Cause 1

E


weakest

(c) Multiple Compensatory
Causes

E-Effect
EE-Strong Effect
EEE-Very Strong Effect


strong

Cause 1


weak


Cause 2

weak strong


(b) Multiple Necessary
Causes


Cause 2
weakest strongest

strongest
E E EE EEE


E EE
Cause 1

E


weakest

(d) Multiple Compensatory
Causes with Graded Effects


Figure 1: Patterns for Kelley's Types of Causal Schemata









compensatory schema is operating. Where there is a

distinction between degrees (or severity) of a given effect

this is referred to as a graded effect.

Spontaneous Attributions

The process of causal attribution has implications in

the context of reviewing one's own work if such attributions

normally occurs as part of an internal auditor's thought

processes. Research has provided support for the process

of spontaneous attribution (Smith and Hunt, 1978; Wong and

Weiner, 1980).

Wong and Weiner (1980) used a "self-probe" methodology

in which the subjects were presented with an effect and

asked to report what questions they would ask themselves.

They found that spontaneous attribution was more likely to

occur when the effect was contrary to expectations or

when the effect (outcome) was negative. In addition,

they found a significant interaction between the effect and

expectancy which resulted from low response rates for the

expected positive outcomes.

Smith and Hunt (1978) examined the spontaneous

attributions of consumers to advertisements. Again the

findings supported the notion of spontaneous attributions,

particularly when the advertisements indicated that the

products were not superior in all respects. This result









corroborates Wong and Weiner's (1980) finding that people

engage in attributions when an outcome is not consistent

with their expectation.

The research findings on spontaneous attribution

support the contention that an internal auditor would

ordinarily make causal attributions in reviewing his own

work for two reasons. One, the discovery of a malfunction

would be considered a negative outcome and, second, it would

be contrary to his expectations. Accordingly, the issue

addressed by the present research is not whether the

internal auditor engages in causal attributions but rather,

whether biases occur due to the causal attributions of an

internal auditor reviewing his own work.


Empirical Studies of Causal Attribution

The following two empirical studies obtained results

from tests of attribution theory that demonstrate the basis

for the methodology employed in the present study. Taylor

and Jaggi (1974) explored the impact of causal schemata

associated with stereotyping in a cross-cultural context.

The experiment was conducted in South India using thirty

Hindu adults as subjects. Each subject read a series of

paragraphs which depicted one of four situations involving

either a Hindu or a Muslim actor who was behaving in either

a socially desirable or undesirable manner. The subjects

were instructed to imagine that these behaviors were

directed toward themselves. The subjects were asked to make











a causal attribution for the behavior of the actor in each

of the paragraphs. For the Hindu actors, the desirable

behaviors were attributed to internal causes, whereas the

socially undesirable behaviors were attributed to external

causes. The reverse was found for out-group (Muslim)

actors.

Deaux (1976) investigated causal attributions about

task performances based on sexual stereotypes. Sexual

stereotypes were assumed by Deaux to be analagous to causal

schemata and thus allow inferences regarding the causes of

task performance (effects). Types of tasks were divided into

two groups, those that were considered masculine, and those

considered feminine. For each type of task it was assumed

that the actor was either male or female and the outcome of

the task performance was either success or failure. The

results imply sexually sterotypic causal attributions;

success or failure expectations appear to be based on

stereotypes. Outcomes consistent with stereotypes were

attributed to internal causes, such as, ability or effort.

Inconsistent outcomes were attributed to external causes,

like task difficulty or luck.

The following scenario illustrates these concepts in

the context of reviewing one's own work. Assume that an

internal auditor has a causal schema which involves failure

to ship the proper quantity and type of goods to customers.

This failure (the effect) may be graded from occasional











occurrences (considered a minimal effect) to a rate of

twenty percent (constituting a severe effect). The causal

schema for this relationship may include causes such as lax

performance by shipping clerks or lack of sufficient control

in the form of a second check on the quantity shipped. When

the shipping errors occur only occasionally, the presence of

either of these causes may be perceived as sufficient to

produce the mild effect, but a severe effect may result in

the internal auditor inferring that both causes must be

present. In addition, causes may be schematically

structured so that they are compensatory. For example, a

second check would compensate for the lax performance by the

shipping clerk.

The causal attribution tests in the present study

assume the development of distinct causal schemata by

internal auditors for the cause-and-effect relationships of

internal control systems. These cause-and-effect

relationships are assumed to be learned through experience.

The internal auditor is hypothesized to evoke a different

causal schema when he is reviewing a system that he designed

than the schema he evokes in reviewing a system he did not

design. As a result of the differences between these two

schemata, causal attributions toward causes other than the

design of the system are expected when an internal auditor

finds a malfunction in a system he designed.









Memory and Causal Attribution Research in Accounting

The topics of causal attribution and memory have

received little attention within accounting research. Two

works that have dealt with causal attributions are Birnberg,

Frieze and Shields (1977), and Shields, Birnberg and Frieze

(1981); only one study has examined memory in an

accounting context (Weber, 1980).

Birnberg et al. (1977) extended attribution theory into

the context of management control systems. A model was

developed which described a management control process where

both the superior and the subordinate made causal

attributions about the subordinate's reported performance.

An extended version of this model was the basis for the

latter study (Shields et al., 1981) which empirically

explored differences between managers and subordinates with

regard to the causal attributions they made about the

subordinate's reported performance. A within subjects

design was employed with MBA students as subjects. The

results indicated a tendency for the subjects in the role of

subordinate to attribute the reported performance to causes

which were external to the subordinate. The subjects in the

supervisor's role tended to attribute the causes of the

subordinate's performance to factors which were intrinsic to

the subordinate.

A moderately analogous situation is examined in the

present study. The internal auditor who reviews his own









work may employ cognitive processes which are similar to

those of the subjects who assumed the subordinate's role. If

so, the internal auditor reviewing his own work would tend

to attribute the causes of malfunctions to causes other than

his own work.

The possible role of memory in accounting judgments was
studied by Weber (1980). He used cluster analysis to search

for structure in the free recalls of EDP auditors. Using

this method he found little evidence of significant

structure in the auditor's free recalls.

Neither of these empirical studies dealt with the same

phenomenon that is the subject of the present study.

However, both provide clarification of the relevance of the

two distinct aspects of judgments about one's own work in

accounting contexts. The Shields et al. (1981) study

indicates that two people might arrive at different

explanations for the same event depending on their

relationship to the event. Though the results are not

supportive of the notion of cognitive organization, the

study on memory (Weber, 1980) recognizes the potential role

of information stored in memory for analysis of subsequent

information.


Summary
In summary, this chapter has provided a discussion of

the relevant literature on (1) the role of schemata in

problem-solving and (2) causal schemata and their role in










organizing one's experiences concerning causes and effects.

The evidence provided about the role of schemata indicated

that they can act in two nonexclusive ways. Schemata can

organize material which is to be encoded and stored in

memory or they can be used in retrieval to structure a

memory search. The relevant result from this literature is

that schemata, in some way, mediate in the retrieval

process. This allows some stored information to be more

readily accessible in memory than other information.

The discussion of causal attribution presented above

gives minimal attention to the substantial body of

literature dealing with the traditional aspects of causal

attribution, such as the location (internal versus external)

or the controllability of the causes. Instead, the notion

of causal schemata which allows combinations of multiple

causes and graded effects was emphasized. In memory, causal

schemata represent many learned relationships which can be

evoked or ignored as the context dictates. It is this

flexibility which allows an individual to employ one causal

schemata when reviewing his own work and another when he

reviews the work of someone else.

Extension of the results of this literature into the

situation where an internal auditor reviews his own work

yields some general expectations regarding the review

process. If memory mediates in the process then (1)

problems which were initially undetected will not be found in






33


a review, and (2) facets of the original work which were

carefully considered while designing the system should be

readily recognizable in the review. That is, the schemata-

based processes should result in schema-consistent

information being more readily available in the review of

previous work. Attributional biases regarding perceptions

of causes and effects would result in the relative tendency

for internal auditors reviewing their own work to perceive

(1) their work as better, (2) any problems found to be

caused by sources other than their own work, and (3) the

effects of those problems to be less severe.

















CHAPTER III

EXPERIMENTAL METHODS


The implications of the general hypotheses from Chapter

II are extended in this chapter to form seven specific

research hypotheses. The experimental design, subjects,

procedure, and tasks used in testing these hypotheses are

discussed in the first four sections. The hypotheses are

stated and discussed in the last section.

The research questions were addressed through

laboratory experimentation involving participation by

practicing internal auditors. The subjects engaged in two

related tasks which were administered at different times. In

the first task, the subjects designed a simple internal

control system. The second task was intended to represent a

subsequent review of the system designed in the first task.

The subjects were allocated to different treatment groups in

the second task so that comparisons could be made between

subjects who reviewed their own work and others who reviewed

the work of the first group. An analysis of the subject's

design and review materials provided evidence about the

effects of memory and causal attributions in reviewing one's

own work.










Experimental Design

As indicated, the actual task performed consisted of

two temporally separate parts. The first part of the task

required two-thirds of the thirty-nine subjects to design

the internal controls for the sales/receivables cycle of a

new mail-order center (for the hypothetical firm described

below). The other one-third of the subjects designed the

controls for the purchasing cycle of the hypothetical firm.

The tasks were considered to be comparable in complexity and

difficulty. The second part of the task required each

subject to "review" a sales/receivables system designed in

the first part.

Based on this task structure three experimental

conditions were formed (see Figure 2). The OWN condition

consisted of subjects who reviewed the system they

originally designed. The ERO condition also designed a

sales/receivables system for a mail-order center, then

reviewed a system designed by a subject from the OWN

condition. The OTH condition designed a purchasing system

then reviewed the sales/receivables system of a subject from

the OWN condition, as in the ERO condition. The important

distinction between the ERO and the OTH conditions is that,

though neither reviewed their own system, the ERO subjects

had experience with the type of system they reviewed while

the OTH subjects did not.











TASK 1 TASK 2

Experimental Design the internal
Group controls for; Review

OWN sales/receivables own design
system
ERO sales/receivables "another internal
system auditor's design"

OTH purchasing system "another internal
auditor's design"

Figure 2: Experimental Tasks by Condition



Subjects

Thirty-nine practicing internal auditors participated

as subjects in this study. Their experience as internal

auditors ranged from three months to 12 years, with an

average of 4.89 years. Each of the subjects held a

bachelors degree and one-third also had earned graduate

credit. Fifty-four percent of the subjects had at least one

professional certification, such as, Certified Internal

Auditor or Certified Public Accountant. Though specific

knowledge of the retailing industry was not required,

twenty-six of the subjects indicated that they had some

experience in the retailing industry.










Procedure
The internal audit directors of four major corporations

were contacted by the Institute of Internal Auditors on

behalf of the researcher. After they consented, these audit

directors were sent envelopes which contained the

appropriate number of first-stage experimental packets (see

Appendices A and B). Arrangements where made with these

audit directors to distribute the packets to the

participants in their firm. The participating staff

auditors completed the flowchart along with the

"documentation" and returned them to their audit director at

least five days prior to completing the second part.

The researcher went to the firms and collected the

completed design packets from the audit director. From the

set of sales/receivables designs, one-half were randomly

chosen as the OWN group. These flowchartE were phctocopied

and included in the packets given to the subjects in the

review phase. The subjects in each firm gathered as a group

with the experimenter. They were told that certain

sales/receivables systems had been chosen to be reviewed.

These flowcharts were ostensibily "representative" of

systems previously evaluated by the experimenter and the

choice did not necessarily reflect on the quality of the

design. In addition to the flowchart, the subjects received

a packet of materials on which they were to respond (see

Appendices C and D).










The OWN condition received packets which referred to

the system they were reviewing as "your system." The

subjects in the ERO and the OTH conditions received packets

which indicated that the system was designed by "another

internal auditor." Otherwise, the review-phase materials

for all three groups were identical.


Experimental Context and Task

This section describes the scenario the OWN and ERO

subjects received as part of the experimental materials.

Within the constraints of a laboratory experiment, the tasks

were intended to replicate the real-world activities of

designing an internal control system and then conducting a

compliance audit.


Experimental Context

The subjects were told to assume that they were members

of the internal auditing staff for a national discount

retailer. During the time prior to this assignment, the

firm has dealt exclusively in over-the-counter sales of a

diverse consumer goods inventory. As a result of a high-

level marketing decision, the firm had decided to expand

into the mail-order market usirgj e invertcrr lEy ctrrieC.

The firm has many retail stores dispersecs across the natcrn.

The role of these warehouses was to be expanded to include a

mail-order center which was to be a profit center. Each

mail-order center was to handle all custorrr orders,









accounts receivable, shipping and collection. Purchasing

and promotion were to be done at the corporate headquarters

with merchandise shipped directly from the manufacturers to

the mail-order centers.


Experimental Tasks

The materials each subject received in the design task

included: (1) a narrative which described the hypothetical

firm, (2) a pre-labeled flowchart along with a set of "peel-

off" labels with symbols indicating documents and books, and

(3) a form for "documenting" the controls they employed.

The subjects were instructed to design an internal control

system using those elements from the labels and flowchart

that they considered necessary to provide good, efficient

internal controls. Items not included on the labels were

allowed and unlabeled columns were provided on the

flowchart. The subjects were required to produce flowcharts

which included the elements that they incorporated in their

designs and arrows indicating the flow of documents through

the system.

Also, as part of the system design, the subjects were

required to provide "documentation" for their design. They

were asked to write a brief description of the purpose of

the controls that they included in their design. The stated

purpose should have included the specific types of errors

and system failures each control was intended to prevent or








reduce. Thus, in the first part of the task the subjects

produced a simple flowchart and a narrative which documented

the purpose of the controls included in their design.

In the second task, every subject was asked to answer

two sets of questions. The first set required open-ended

answers to two questions. The questions asked: (1) "Which

controls in this system are strong controls which are very

likely to detect and correct errors, and what errors or

malfunctions do they control?" and (2) "What areas of this

system are weak and therefore provide little detection or

correction of errors, and what errors or malfunctions do

they fail to control?" The subjects were provided forms

which required them to give information about the personnel,

functions, and documents involved.

The second set contained three questions (presented

below) and required the subjects to respond on eleven-point

scales. These responses were with regard to ten

"malfunctions" which ostensibilly occurred in the system

under review. The malfunctions included in the materials

given to all the subjects are shown in Table 1. The three

types of ratings were in answer to the following questions

(see Appendices C and D):

1. How strong are the controls which are intended to
prevent or detect this malfuntion?

2. How important are factors other than the design of
the system in causing this malfunction?

3. How severe is the effect of this malfunction on
the financial information provided by this system?















0 Shipping documents have been found which indicate
that merchandise had been shipped be no customer
order can be located.

1 Customers have been complaining about being billed for
prepaid merchandise. The merchandise was shipped and
included a "packing slip."

2 Customer orders for items which were temporarily
out of stock and backordered remained unfilled up to
two weeks after the backordered items had been received
at the warehouse.

3 Merchandise has been returned but the customer's
account was not credited. Receiving reports were
properly prepared.

4 The quantity of merchandise shipped was not what
the customer ordered. The invoice and the customer
order do not agree.

5 Payments were received from customers whose accounts
had been written-off.

6 Items shipped are not deducted from the inventory
records.

7 Merchandise continues to be shipped to customers
who are no longer approved credit customers or whose
current account balance exceeds their authorized credit
limit.

8 The type of merchandise shipped is not what the
customer ordered. The customer order, shipping
document, and invoice are properly prepared.

9 Customers have returned merchandise indicating that
a duplicate shipment had been made. No duplicate
documents were found.


TABLE 1: The Malfunctions Used in Causal Attribution Ratings









Each question was rated on a separate page and the

responses were elicited in the same order that the questions

are listed above. The ten malfunctions were presented in a

different order for each of the three questions. The second

part of the task also included a debriefing questionnaire in

which the subject was asked to provide personal demographic

informatirr a. well as complete a motivational questionnaire.



Research Hypotheses

This section presents the research hypotheses for both

the mental representation and the causal attribution aspects

of an internal auditor's review of his own work. These

hypotheses involve comparisons alcng the E.perirental

conditions previously described.


Hypothesized Effects of Schemata

There are four hypotheses regarding the influence of

what an internal auditor remembers from an initial design

problem on the review process which follows. The first two

hypotheses are based on the results of Alba et al. (1981)

which indicate that recognition is not affected by the

presence of schematic structuring at encoding. For the

subjects in the OWN and the ERO conditions, each internal

auditor has a set of controls which they believe to be an

adequate solution to the problem. The two sets of controls

will not necessarily be identical. Thus, the controls from








the two designs can be classified as follows: (1) controls

which are common to both designs and (2) those controls

contained in one design but not the other (see Figure 3).


D I
A B C









PROBLEM SPACE

A = the set of controls found only in the OWN
subject's design

B = the set of controls found in both the OWN and the
ERO subject's designs

C = the set of controls four orly ir ti ErC
subject's ttic;i.

D = the set of controls found in neither the OWN nor
the ERO subject's designs



Figure 3: The Hypothesized Sets of Controls in the Problem
Space









HI: The OWN subjects and the ERO subjects will not
differ in their ability to identify strengths (for the
set of controls contained in the designs of both
subjects).

H2: The OWN subjects will be able to identify strengths
better than the ERO subjects (for the set of controls
that were not in the design of the ERO subjects).


The two hypotheses below refer to the weaknesses which

are present in the system under review. The internal

auditor would be forced to remember controls either from his

general knowledge of internal control systems or from

specific experience with the system under review. The

results from the problem-solving literature (Weisberg and

Alba, 1981; Higgins and Chaires, 1980) indicate that the

representation of the problem in memory may inhibit

formulation of alternative solutions. For subjects in the

ERO condition, his schematic representation of the internal

control system may facilitate recall for the set of controls

present in his design but not in the system being reviewed.


H3: The ERO subjects will be able to identify
weaknesses better than the OWN subjects (for the set of
controls not in the system).

H4: The OTH subjects will be able to identify
weaknesses better than either the OWN or the ERO
subjects(for the set of controls not contained in
either of the original sales/receivables systems).


Hypothesized Causal Attributions

The next three hypotheses concern the differences in

causal attributions between the OWN condition and either the









ERO or the OTH conditions. These hypotheses involve three

elements which could be expected to be part of the causal

schemata associated with internal control systems. Those

elements are the internal auditor's perceptions of: (1) the

severity of the effect of the observed malfunctions on the

financial information, (2) the strength of the controls

associated with those malfunctions, and (3) the importance

of causes of the malfunctions other than the system's

design. These elements are assumed to be organized into

causal schemata. Within these causal schemata, the effect

(the malfunction) is caused by either the design of the

system or some other source, such as failure by firm

employees to perform specified procedures.

H5: The OWN subject will rate the effect of the
malfunctions found in the system as less severe than
either the ERO or the OTH subject.

H6: The OWN subject will rate the controls associated
with a malfunction as better than either the ERO or the
OTH subject.

H7: The OWN subject will rate the importance of some
source other than the design of the system causing the
malfunction as higher than either the ERO or the OTH
subject.















CHAPTER IV

DATA ANALYSIS AND RESULTS


This chapter discusses the methods of analysis and the

results of the study. The first section describes the

process of analyzing the data obtained from the listings of

strong controls and weaknesses. The results of the tests

concerning the effect of memory in reviewing one's own work

are included. The second section discusses the tests

employed on the ratings associated with the causal

attribution aspect of reviewing one's own work. The results

of these tests are also presented in the second section.


Tests for Schematic Structuring

The measure upon which the memory tests were based is

the notion of "idea unit" (Bransford and Johnson, 1972).

The idea units in the subject's responses for both parts of

the experimental task were utilized in the tests reported in

this section.










Overationalization of "Idea Unit"

The rules used to determine what constitutes an idea

unit in this research were specified a priori. An idea unit

must include four elements: (1) the personnel, (2) the

functions performed, (3) the documents involved, and (4) the

errors or malfunctions controlled. For example, an idea

unit which meets this standard is "The cash listing and the

daily deposit slips are reconciled by a clerk who does not

have access to the cash or the accounting records to avoid

theft of cash." The personnel is a clerk who does not have

access to cash or the accounting records, the documents are

the cash listings and the daily deposit slips, the function

is reconciliation, and this process controls for the theft

of cash.

Prior to any statistical analysis, the idea units

were coded by two independent raters who were blind to the

nature of the experiment and the conditions the subjects

were under. Both raters were graduate students with at

least two years experience as public accountants. The

raters made judgments about the idea units contained in

both parts of the experiment. Initially, the idea units in

the documentation of each subject's design (for the subjects

who designed a sales/receivables cycle) were identified by

the raters independently. Then the raters' lists were

compared and differences reconciled.








Based on the designs of the pair of subjects in the OWN

and the ERO conditions, the idea units in the documentation

of the systems were grouped into three categories. The

three categories were: (1) the set of idea units which were

common to both of the original designs for the OWN and the

ERO conditions, (2) the set of idea units found only in the

OWN subject's design, and (3) the idea units which were only

in the ERO subject's design. Thus, the codings of the idea

units in the subject's designs produced three sets of

mutually exclusive idea units for each triple of subjects.

In the second experimental task, the subjects were

required to provide listings of the strong controls and the

weaknesses for the system they reviewed. Each subject had

the flowchart (but not the explanatory "documentation")

available in the second task. Thus, regardless of the

condition the subject was under, they were required to

recall (from either general or specific prior knowledge) the

malfunction controlled by any strength or weakness they

listed. The unavailability of the documentation helped

assure that if the idea unit in these lists matched the

original idea unit the matching was due to schematic

structuring of the information.

The idea units contained in the lists of strengths and

weaknesses were also identified by the raters. The raters

made two additional judgments for these idea units (1) they

determined whether the idea unit from the listings matched








the idea unit in the subject's original documentation, and

(2) they determined whether an idea unit that did not match

the original idea units, should be characterized as a

"strength" or "weakness." An idea unit was classified as a

"strength" (or "weakness") according to whether it provided

a high (or low) probability of detection and correction of

the stated malfunction.

The two additional evaluations made by the independent

raters strengthened the tests of the effects of memory. The

first evaluation was necessary to determine the effects of

memory as opposed to mere identification of the controls in

the flowchart. The second evaluation controlled for the

subjects who simply listed all the controls as strengths or

weaknesses regardless of their probability of detecting or

correcting errors.


The Results of the Memovy Effect

The hypotheses concerning memory were tested by chi-

square tests (Conover, 1980, p. 144) on the idea units from

both parts of the experimental task. The idea units listed

as strengths or weaknesses were classified into four

categories, the three described earlier (both, OWN only, and

ERO only) and a fourth category for those that were not

listed in the documentation of either the OWN or the ERO

subject's documentation in each triple (see Tables 2 and 3).













TABLE 2

Number of Idea Units Identified in Each Groupa



1 2 3 4 5 6 7 8 9 10 11 12 13


0 3 1
10 14 10
1 4 5


4 2 2 4
5 4 21 29
10 3 11 5


Controls
found in
document-
ation of:

BOTH
OWN only
ERO only

Strengths
identified
in:

BOTH
OWN only
ERO only
NEITHER

Weaknesses
identified
in:

BOTH
OWN only
ERO only
NEITHER


0 0
0 0
0 6
6 10


a Each group was the triple of subjects that reviewed
the same sale/receivables system.


5 3 0
2 3 17
5 1 2
8 13 3





1 1 0
0 0 0
2 1 2
6 7 3









TABLE 3

Total Number of Idea Units by Condition

Condition:

PMI ERO OTH Total
Controls
found in BOTH 25a
document- OWN only 133 133
of: ERO only 86 86


Strengths BOTH 17 13 30
identified OWN only 72 72
in: ERO only 13 13
NEITHER 29 30 8 67

Weaknesses BOTH 0 4 4
identified OWN only 5 5
in: ERO only 17 17
NEITHER 16 34 15 65
a Each of these controls were found in the
documentation of both the OWN subject and the ERO
subject.

Each memory hypothesis specified an area of the problem

space and predicted a condition which should have a

comparative advantage in identifying controls in that area.

For the set of controls contained in the documentation of

both the OWN and the ERO subjects, Hi predicted that no

difference would be found in the ability to identify

controls in this area. Twenty-five controls were listed in

this area. High proportions of the possible idea units were

identified (68 percent for the OWN condition and 52 for the

ERO condition) and, as predicted, no difference in ability

to identify controls which were common to both designs was

found (see Table 4). A prediction of finding no difference








is a test of both the adequacy of the experimental method

and the hypothesized effect. Therefore, the result for the

test of El described above is more meaningful when it is

considered in light of the following two tests which

employed the same experimental method and statistical test.

The second hypothesis (E2) predicted that (for the set

of controls not found in the design of the ERO subject) the

OWN subjects would be able to identify strengths better than

the ERO subjects. As shown in Table 4, the predicted result

was found. For controls not found in the OWN subjects'

designs, H3 predicted that the ERO subject would be able to

identify weaknesses better than the OWN subjects. The

results for the test of weaknesses found in the area of the

problem space outside the OWN subject's design were

consistent with H3 (see Table 4).

The proportions of the controls identified were lower

for the areas of the problem space where the controls were

unique for one condition in the group than the area which

contained the set of controls originally listed by both

subjects (54.1 percent and 3.8 percent for the OWN

condition, and 15.1 and 19.8 percent for the ERO condition

for the strengths and weaknesses respectively). The

controls that were in both designs tended to be very common

controls; accordingly, the observed differences may reflect

that they are readily identifiable or come to mind more

easily.









TABLE 4

Chi-Square Tests for Hl, H2, and H3

number Signifi-
condi- number not chance
Hypothesis tion listed listed levels

El: The OWN subjects and the
ERO subjects will not differ OWN 17 8 .250
in their ability to identify
strengths (for the controls ERO 13 12
contained in the designs of
both subjects).

H2: The OWN subjects will be
better able to identify OWN 99 101 .001
strengths than the ERO subjects
(for the set of controls that ERO 29 171
were not in the design of the
ERO subject).

H3: The ERO subjects will be
able to identify weaknesses OWN 16 135 .001
better than the OWN subjects
(for the set of controls that ERO 40 111
were not in the system).


The last hypothesis regarding the effect of memory (H4)

predicted that the subjects in the OTH condition would be

able to identify weaknesses in the system (for the set of

controls not found in either sales/receivable system) better

than either of the other two conditions. This hypothesis

was tested using a chi-square goodness-of-fit test (Conover,

1980, p. 190). The test result indicates that the

proportion of controls identified in the area where neither

design had specified controls were significantly different

(see Table 5). The subjects in the ERO condition identified

more than twice as many weaknesses as either of the other

two conditions.









TABLE 5

Chi-Square Goodness-of-Fit Test for H4

Signif-
Condi- Number Expected icance
tion Listed Number Level

H4: The OTH subjects will be
able to identify weaknesses OWN 16 21.67 .005a
better than either the OWN
or the ERO subjects (for the ERO 34 21.67
set of controls not
contained in either of the OTH 15 21.67
original sales/receivables
systems).

a This test indicates that a hypothesis of equal
frequencies for all condition may be rejected.
Inspection shows that the ERO condition identified
significantly more weaknesses than the either of the
other two conditions in the area of the problem space
specified in H4.

Differences in motivation may have contributed to this

result. The motivation of the ERO subjects to list more

weaknesses (or the OWN subjects to list more strengths)

cannot be assessed directly. The subject's motivation to

perform the experimental task was assessed via the post-

experimental questionnaire. No differences in motivation to

perform the experimental task were found among the

experimental conditions (F=.09, p<.917). An indication of

the subject's willingness to participate in the experiment

is the fact that, in spite of the rigor of the task, 23 of

the thirty-nine subjects were willing to participate in a

similar task at least one more time.










Tests of Causal Attributions

Analysis of the Attribution Ratings

The subject's responses to the three attribution

questions were made on eleven-point scales. The end points

of the three scales were as follows: (1) the severity of the

effect of the malfunction on the financial information was

rated from "no effect" to "severe effect," (2) the

importance for causes of the malfunction other than the

design of the system was rated from "not important" to "very

important," and (3) the strength of the controls for

preventing or detecting the malfunction was rated from "very

weak" to "very strong."

Using gach of the three types of attribution responses

as dependent variables, a multivariate analysis of

variance (MANOVA) was conducted. The model was



jijk ai + Yj + Pk + (Ta)ij + (P)ij + (T)jk + (QTp)ijk
where,

Y-ijk is the vector of responses in the ith condition
for the jth triple to the kth malfunction for the three

dependent variables.

ai is the main effect of condition (COND) at the ith
level.

8 is the main effect of the triples (GROUP) at the jth

level.










Sk is the main effect of the malfunctions (MALF) at the
kth level

and, (aB)ij, (ay)ik, (BY) jk, and (aBT)ijk are the

interaction effects.


The analysis of variance (ANOVA) model for eafh of the

causal attribution measures separately analyzed was


Yijk = ai + + (+ ~+) + (aP)ik (Yj + (ayP)ijk

where,

Yijk is the response to one of the three attribution
ratings in the ith condition (COND) for the jth triple

(GROUP) to the kth malfunction.

These models are treated as a repeated measures design

where COND is a fixed factor for the three conditions, GROUP

is a randomized blocking factor for those subjects who

reviewed the same system, and MALF represents the ten

malfunctions to which the subjects responded. Each subject

received a unique treatment because (1) the flowcharts

received were common only to those subjects within a triple

(GROUP level), and (2) the COND levels (OWN, ERO, and OTH)

occurred only once in a triple. Thus, GROUP and MALF are

repeated measures within each condition (COND level). The

effect of GROUP cannot be tested because it has no error

term in this design. The effects for COND, MALF, and COND X

MALF were tested using the Hotelling-Lawley Trace (Harris,










1980) for the MANOVA and the usual omnibus F test for the

ANOVAs. The error terms for these tests were the

appropriate mean squares based on the design (Keppel, 1982)

rather than the residual error.


Results of Causal Attributions

Significant correlations existed among the three

dependent variables, ICRAT (the rating of the internal

controls), OTHFACT (the importance of other factors), and

EFFECT (the severity of the effect) as shown in Table 6.



TABLE 6

Pearson Correlations Between ICRAT, OTHFACT, and
EFFECT


OTHFACT EFFECT

.0732 .0950
ICRAT p=.1669 p=.07050
n=358 n=363

.1309
EFFECT p=.0112
n=375






The overall MANOVA exploits these correlations and

provides a test for differences among conditions that

controls for Type I error at the nominal level (Harris,

1982, p. 18). The test excluded all observations where a










subject failed to respond to one of the three measures.

This resulted in the exclusion of 8.7 percent of the

observations. No systematic pattern of missing data was

found. The MANOVA produced approximations for the F

statistic which were significant for all three estimatable

effects: COND (F=3.31, p<.0001), MALF (F=3.33,p<.0001), and

COND X MALF (F=5.33,p<.0001). These results are detailed in

Table 7.


TABLE 7


ultivariate Analysis of Variance on
and EFFECT


Error
Matrix

COND X
GROUP

MALF X
GROUP

COND X
MALF X
GROUP


Approximate F
Based on the
Hotelling-
Lawley Trace
(d.f.)

3.31 (6,42)


3.33 (27,314)


ICRAT, OTHFACT




Probability
of F (assum-
ing the null
hypothesis)

.0092


.0001


5.33 (54,536)


The separate ANOVA's for each dependent variable

provided tests which were more directly related to the first

three causal attribution hypotheses (see Table 8). The


M









I


Independent
Variable

COND


MALF


COND X
MALF


I
























































































0 0

r:


a
D
o





N


"
o

= 5



6
o
o

c
N

e N


" R
$ c

% "
S



c
1
o

3


N


P
Z a

n


" 6

~1N












X~s


W6"


VICN















"
nlN
" 6


8









dependent variable resulted in marginal significance for the

COND effect (F=2.98, p<.070) and a significant COND X MALF

interaction effect (F=4.17, p<.0001). The effect of MALF

was not significant. For tests regarding the importance of

causes other than the design of the system (OTHFACT), there

was no main effect of COND (F=.87, p<.43). However,the main

effect of MALF was significant (F=2.71, p<.007) as was the

COND X MALF interaction (F=2.54, p<.0008). The ANOVA on the

ratings of EFFECT for the ten malfunctions yielded

significance for both main effects and the interaction.

Thus, for all three dependent variables at least two of

the omnibus F tests resulted in significant effects. More

importantly, the interaction term was highly significant for

all three dependent variables. In order to clarify the

implications of these results, analyses of the factor level

means were undertaken. Pair-wise tests among the condition

means for the three dependent variables were planned

comparisons. All other comparisons were post hoc tests

whose overall significance for the families of related tests

was controlled by the Scheffe' method (Neter and Wasserman,

1974) at an alpha level of .05. The pair-wise comparisons

between the mean responses of the three conditions on the

three dependent variables are presented in Table 9.

The differences in the means for ICRAT are consistent

with H6 which predicted that the OWN condition would rate

the strength of the controls for their system as higher than









either of the other two conditions. The ERO condition rated

the controls as weakest on average. This suggests that the

effect of specific experience may be to heighten awareness

of the potential impact of shortcomings in the design.

However, due to the significant interaction between COND and

MALF, no conclusions regarding the main effects should be

drawn at this point.


TABLE 9

Pair-wise Comparisons of Condition Means

MEANS OWN OWN ERO
Dependent vs vs vs
Variable OWN ERO OTH ERO OTH OTH

ICRAT 5.545 4.055 4.756 *
OTHFACT 8.192 7.346 7.992
EFFECT 6.575 5.015 6.805 ** **

p<.05.
** p<.01.



The mean ratings for the importance of other factors

(OTHFACT) were not significantly different, but the direction

of these differences was consistent with H7; i.e. that the

OWN condition would perceive the other factors as more

important in causing the malfunctions than either of the

other two conditions.

The expectation in H5, regarding the effect of

malfunctions (EFFECT), was that the OWN subjects would rate

the effect as less severe than the other two conditions.

Contrary to that expectation, the ratings on EFFECT were










significantly lower for the ERO condition than in either of

the other two conditions. The other two conditions did not

differ from each other. Thus, the OWN condition apparently

did not perceive the effect as expected when compared to the

ERO condition.

Given the significance of the interaction between

conditions and malfunctions in the omnibus tests, univariate

ANOVAs were conducted for each malfunction individually

in order to determine whether the COND effect was present for

each malfunction. This analysis was performed for each of

the dependent variables and the results are reported in

Table 10. The only malfunction that did not have at least

one significant effect was number 6 (which stated that items

shipped had not been deducted from inventory records).


TABLE 10

Univariate ANOVAs for Each Malfunction


Malfunction


0
1
2
3
4
5
6
7
8
9

ns = not significant


Prob Value for F

ICRAT OTHFACT EFFECT

.0091 .0184 .0791
ns ns .0001
.0601 ns ns
.0148 ns .0004
ns ns .0143
ns .0026 .0052
ns ns ns
.0001 ns .0100
.0173 ns .0198
.0085 ns .0001

at .1 or less











The pair-wise comparisons between conditions for each

malfunction which had a significant main effect in th

univariate ANOVAs are shown in Table 11. These comparisons

further explain the COND X KALF interaction effects for each

of the dependent variables.

The significant differences in the strength of controls

ratings (ICRAT) were in the expected direction except for

malfunctions 2 and 3. These two malfunctions dealt with

back-orders and returned goods, respectively. The reason

the OWN condition rated these two malfunctions lower may be

that very detailed aspects of the system (such as these)

were not incorporated into most of the systems. Those

subjects in the OWN condition who had not included these

aspects in their design realized that no provision had been

made for these items in their system. Consequently, they

rated the controls over these areas as weak. The other two

conditions (ERO and OTH) seemed to rate the controls as

adequate when the controls in question were not apparent

This result may have occurred because different levels of

factual knowledge existed among the conditions (i.e. the OWN

subjects knew the controls were absent).

In the importance ratings for other factors (OTHFACT),

all of the significant differences between the conditions

were in the predicted direction except for malfunction 0.

This is consistent with the assertion that the OWN condition











TABLE 11

Means by Malfunction for ICRAT, OTHFACT, and

dependent variable: ICRAT
OWN OWN E
vs vs
# OWN ERO OTH ERO OTH C


6.769
6.000
3.308
1.917
6.385
5.600
7.917
5.769
6.385




6.000
8.583
8.462
7.846
8.000
8.538
8.692
9.000
8.923
7.909



5.923
8.333
5.077
4.083
7.154
7.462
9.154
7.846
4.692


2.667
4.231
5.692
5.077
4.308
3.000
3.667
5.385
2.769


5.333
5.615
5.462
4.182
4.000
5.231
2.545
2.500
6.000


*
* *

* *


OTHFACT


8.077
5.846
6.846
7.231
8.308
4.000
7.462
8.159
9.154
8.385


9.000
7.083
7.769
9.385
7.500
6.583
8.154
8.077
8.692
7.462


EFFECT


5.000
1.615
2.846
5.231
5.846
8.692
6.000
7.462
2.769


8.333
7.538
7.833
7.231
4.077
9.231
7.000
4.308
9.000


*

*
*
*


*significantly different means in pair-wise
comparisons when the family of tests for a
malfunction across conditions is controlled by
Scheffe' method at alpha level =.05.


EFFECT


RO
vs
)TH

the






* I










*I







*
*
*
*
*


*




' the I


----










would attribute the causes of malfunctions found in their

system to causes other than the system's design. The

exception involved shipment of goods when no customer order

could be found. The OWN subjects may have reasoned that

controls over this type of malfunction would control

apparent "other factors." The controls over this

malfunction were rated as good by the OWN subjects (the

second highest rating) therefore other factors could not

have been the cause of this malfunction. In other words,

the OWN condition subjects may have employed a necessary

causal schema which required both causes (weak controls and

some other cause) to be present for a severe effect to

occur. They rated the controls as significantly stronger,

the effect as lower, and may have inferred that other causes

could not have been present.

The omnibus F test for the ratings on effect of the

malfunction on the financial information produced by the

system indicated that the OWN subjects tended to perceive

the effect as more severe. The comparisons between the OWN

condition and the other two conditions show malfunction to

have been rated by the OWN subjects as having a

significantly lower EFFECT. An additional three

malfunctions were rated as significantly lower by the OWN

condition when compared only to the OTH subjects

(malfunctions 0, 3, and 9). These appear generally to be

malfunctions that result from clerical errors in contrast to










possible premature write-offs (malfunction 5) or shipping

goods to unauthorized customers (malfunction 7). If this

dichotomization represents the way the OWN subjects

perceived the effects of the malfunctions then it may be

that H5 (which relates the expectation of lower ratings by

the OWN subjects with the effect of the malfunctions on the

information provided) is applicable only to items the OWN

subjects perceive to be outside their control (or the

control of their system).

Based on inspection of Table 11 it appears that when

malfunctions were perceived by the OWN subjects to have a

severe effect they inferred that both weak controls and some

other factor must be present. The subjects in the other two

conditions appear to consider the presence of either cause

as sufficient to produce a severe effect. To test for this

relationship, the data were partitioned into two categories.

The "high effect" category contained all the malfunctions

which were rated by the OWN subjects as having a

significantly higher effect (malfunctions 2, 5, and 7). The

"low effect" category contains the remaining seven

malfunctions (which were by default considered to be those

rated as having a more severe effect by the other two

conditions). Pearson product-moment correlations were

calculated for both high and low effect categories among the

three dependent variables. A negative correlation between

ICRAT and OTHFACT would suggest that both factors were










perceived as necessary to produce a malfunction with a given

effect; a positive correlation would indicate a perception

that either cause would be sufficient to produce such a

malfunction. These correlations are shown in Table 12.

The correlations for the OWN condition suggest that

both weak controls and some other causes were perceived as

necessary to produce the malfunctions in the high effect

set. The other two conditions tended to produce positive

correlations which indicates that either causal factor was

perceived as sufficient to produce these malfunctions.

Further relationships are suggested by these

correlations. For the "high effect" set, the OWN condition

apparently perceived OTHFACT as the principle factor in

producing these malfunctions. The ERO and the OTH subjects

produced correlations for the "low effect" malfunctions

(their "high effect" malfunctions) which suggest that they

perceived different causes for these malfunctions.

The ERO subjects tended to indicate that higher levels

of OTHFACT were present when these malfunctions produced

severe effects. The OTH subjects, rather than vary their

attibutions toward other factors with their attibutions

regarding the effect of the malfunction, found the strength

of controls and the severity of effect to be positively

related. This result may have occurred because the OTH

subjects did not have previous experience with this type of















TABLE 12

Correlations by Condition for "High" and "Low


"lish Effect"

OTHFACT EFFECT

-.2382 -.0222
ICRAT p=.1750 p=.9010
n=34 n=34

.4000
EFFECT p=.0142
n=37


CONDITION:







OWN








ERO








OTH

I


!LJa Elff

OTHFACT

.0642
p=.5665
n=82

.0756
p-.4942
n=84


-.0441 .0665
p=.7912 p=.5380
n=34 n=88

.2678
p=.0107
n=90


-.0233 .0372
p=.8912 p=.7432
n=37 n=80

-.0222
p=.8383
n=87


SEffect





EFFECT

-.0267
p=.8087
n=85






.1268
p=.2362
n=85






.2117
p=.0594
n=80


.2146
ICRAT p=.1957
n=38

-.0990
EFFECT p=.5486
n=39


.1300
ICRAT p=.4498
n=36

-.0643
EFFECT p=.7015
n=38









system. Therefore, they relied on more general causal

schemata. Reliance on more general knowledge may have led

the OTH subjects to rate the strength of controls as

stronger (as long as some control was found) for those

malfunctions they believed ought to be controlled The

reason that they may have believed these malfunctions ought

to be controlled could have been that this set of

malfunctions had the potential to produce a severe effect.

The reliance on a previously formed causal schemata may have

resulted in the positive correlation between ICRAT and

EFFECT for the malfunctions the OTH subjects considered to

be more severe.

The results of this chapter can be summarized as

supporting the assertion that the OWN subjects differ from

subjects who were not reviewing their own work for both the

effects of memory and their attributions regarding the

causes and effects of malfunctions. Significant differences

exist among the conditions for the dependent variables,

ICRAT and EFFECT, but any interpretation of these main

effects must be conditioned by the effect of the interaction

effects. All of the results are summarized in the following

chapter.
















CHAPTER V

CONCLUSIONS AND LIMITATIONS


The primary purpose of this chapter is to present the

conclusions (and some limitations) of the study. The first

section briefly reviews the experiment that was conducted

and summarizes the results obtained. The conclusions that

may be drawn from these results are presented in the second

section. The final section identifies those characteristics

of the experimental methods that limit the generalizability

of the results obtained.


Summary of the Experimental Study

The design of the experiment allowed comparisons

between internal auditors who reviewed their own work (OWN

condition) and two other groups. One of the other groups was

experienced with the type of system under review (ERO),

whereas the other was not (OTH). Two factors assumed to

affect the review process were incorporated within the

experiment. These factors were (1) the effect of previously

stored system-specific knowledge on the auditor's ability to

access general knowledge about control systems, and (2) the

effect of biases due to causal attributions that differed








with the relationship of the internal auditor to the system

being reviewed.


Effects of Memory on the Review Process

The experimental task required the subjects to design

and review an accounting system and later review either

their own or another's system. As part of the design, the

subjects provided documentation of the controls. The

subsequent review required the subjects to list the

strengths and weaknesses in the system they reviewed. Two

independent raters deduced from the subjects written

responses what the raters considered to be the "idea units"

in the responses. An idea unit in this study included the

personnel, documents, function performed, and the

malfunction prevented. The proportions of idea units

identified in the listings of strengths and weaknesses

relative to the possible idea units (as determined by those

idea units contained in the original documentation) were the

bases of the tests for memory effects.

The tests on these proportions indicated that the OWN

subjects did not differ from the ERO subjects in

identification of strengths that were common to the original

designs of both. When the test focused only on the

strengths identified by the raters as part of the OWN

subjects' systems, the OWN subjects were significantly better

than the ERO subjects at identifying the strengths of the

system. Weaknesses (which were controls not contained in










the system reviewed) were identified significantly better by

the ERO subjects than either the OWN or the OTH subjects.

For the controls that were not included in the documentation

of either the OWN or the ERO condition, all three conditions

were compared with regard to their ability to identify

weaknesses. The ERO condition identified significantly more

of these weaknesses than either of the other two conditions.

The OWN and the ERO conditions did not differ from each

other.


Causal Attribution Effects on the Review Process

The cause-and-effect relationships associated with

malfunctions in an internal control system were examined

across three dimensions. These dimensions are (1) the

severity of the effect on the information produced by the

system (EFFECT), (2) the strength of the controls intended

to prevent the malfunction (ICRAT), and (3) the importance

of causes other than the design of the system (OTHFACT).

Comparisons among the conditions on EFFECT showed that

the ERO condition rated the effect as significantly lower

than either of the other two conditions (the other two

conditions did not differ from each other). The OWN

condition was significantly higher than the ERO condition in

the ratings given ICRAT, but no other pair-wise differences

were significant. No difference was found among conditions

for their ratings of OTHFACT. These results for the main









effects must be conditionally interpreted due to the

significant interactions that existed between the conditions

and the malfunctions.

Post hoc tests revealed some possible explanations for

the interactions. Regarding EFFECT, it appears that the OWN

subjects differentiated between those malfunctions that an

internal control system could be expected to control and

those that a system could not be expected to control. The

effect of malfunctions which, according to the subjects,

could not be expected to be controlled were rated as

significantly weaker by the OWN condition than the other two

condition. The ratings of ICRAT for two malfunctions were

rated as significantly weaker by the OWN condition. These

two malfunctions typically were not controlled by the system

being reviewed. The OWN subjects recognized the

deficiencies of their system and rated the controls as weak,

while the other two conditions rated these controls as

adequate though no control were apparent.


Conclusions

The objective of this study was to determine whether

memory or attributional biases affected an internal auditor

when he reviews his own work. Conditioned by the

limitations discussed in the next section,the answer is yes

for both aspects.

With regard to the effect of memory, three general

conclusions can be reached. First, the internal auditor who










reviews his own work appears better able to identify the

strengths of the system. Second, internal auditors who are

familiar with the type of system under review (but are not

reviewing their own system) appear to be better able to

identify weaknesses in the system. The third conclusion

regarding the effects of memory is that, when an internal

auditor is not sufficiently familiar with the type of system

being reviewed, there is apparently no ability to organize

general knowledge to interpret the specific context. This

last conclusion is similar to that reached by Weisberg and

Alba (1981) who stated that without task-specific experience

the problem-solver would be unable to find the solution.

The causal attributions concerning the system being

reviewed varied with the reviewer's relationship to the

system. From the results found in this study, the general

conclusion is that those with experience who review the work

of another tend to be more critical than either an internal

auditor who reviews his own work or an internal auditor who

is not experienced with the specific type of system under

review. Exceptions to this conclusion (e.g. the weaker

ratings for the strength of absent controls by the internal

auditor who reviews his own work) appear to result from

differences in factual knowledge.










Limitations
The limitations on the operationalization of the

theoretical constructs and the methodology will be discussed

in this section. The operationalization of mental

representation assumed that the design the subjects give as

their solution is their cognitive representation of the

control system. By using the documentation the subjects

provided as a surrogate for their their mental

representation, some items the subjects may have implicitly

considered as part of their design may not have been

considered as part of the design in the analysis. This

could have resulted when the subjects failed to "document"

all the controls in their system.

Also, the results of Higgins and Chaires (1980) suggest

that even subtle manipulations can influence the availability

of certain substructures in memory. Problem-solving

associated with the original designing is assumed to

increase the availability of those substructures. However,

if the increased availability did not occur due to some

cause such as lack of involvement, then the memory for the

specific internal control system would not mediate in the

cognitive processes during the review.

The methodology has two obvious limitations for which

attempts were made to control their effects. They are (1)

the potential bias in reporting the strengths and weaknesses

in the OWN group, and (2) the subjectivity of the coding of










idea units. The steps taken to minimize the effects of

these items were discussed in the methodology and data

analysis sections. Any residual effects can only be

acknowledged as unavoidable limitations.

The results of this research should contribute to an

understanding of the effects on an internal auditor's

evaluations when the work being reviewed is his own. The

sources of those effects appear to include (1) results of

the causal attributions for malfunctions identified, (2)

biological effects due to the functioning of memory in the

judgment process, or (3) some combination of these two

sources. Knowledge about these factors will provide

direction for attempts to mitigate any detrimental effects

of an internal auditor reviewing his own work.

As in any laboratory experiment, the findings cannot be

readily generalized due to factors such as the limited,

nonrandom sample and the artificiality of the task. The

subjects were volunteers from large internal auditing staffs,

thus they may not be representative of internal auditors as

a whole. The level of experience, as measured in years did

not differ among the conditions, but the relevance of the

subject's prior experience may have enabled some subjects to

perform the task better. The effect of prior experience was

assumed to be minimized by random assignment of the subjects

to conditions. Though the experimental task was structured

to be as representative of a real-world task as possible,





77



the subjects did not have all the information (such as

personal knowledge of the personnel and the firm) or the

same type of incentives that they would have had in a real-

world environment. Consequently, the results should only be

considered as part of the evidence in the area of the

effects of memory and motivational biases in reviewing one's

own work.
















Appendix A

Design Materials for the OWN and ERO Conditions




















Thank you for volunteering your time to participate in this study.
Its purpose is to gather information about the role internal auditors play
in the design and evaluation of internal control systems. You are being
asked to complete part of a hypothetical feasibility study for a modifica-
tion of a Sales/Receivables internal control subsystem. Later, when we
meet at your company, you will be asked to evaluate a system as though it
had been implemented. There will be no need for you to consult reference
books or your colleagues, there is not right answer to this problem. I
simply want a preliminary type of solution. I realize that those of you
who have volunteered to assist in this study have very diverse experience
and backgrounds. That diversity is good because it is also true of internal
auditors in general.

You will be acting as representatives of your profession. The Institute
of Internal Auditors (which is financially supporting this study) is inter-
ested in the implications of what I may find for the profession. Thus, it
is important that you take these tasks seriously and perform as though this
assignment was part of your regular job. I .assure you that your individual
responses will be kept strictly confidential and will not be used for any
purpose other than my study.

The remainder of this packet of material includes the following items:

1. Some background material about the hypothetical firm.

2. A description of the present Sales/Receivables subsysten.

3. The requirements for modifying the subsystem.

4. Two forms for you to provide your (internal auditing's) design
and documentation.

The Burke-Layne Case

Assume that you are on the internal auditing staff of Burke-Layne and
Company (hereafter, B-L). B-L is a national, retail chain which sells a
broad range of general merchandise (i.e., clothes, appliances, housewares,
jewelry, etc.). B-L presently has 237 retail stores organized into seven
regional sales divisions. Each region is a profit center with regional
offices located with the regional warehouse. B-L's vendors ship merchandise
directly to the warehouses which in turn supply the retail stores in the
region.

All Purchasing and General Administration (including the internal
auditing staff) is centralized in Chicago at the corporate headquarters.
Personnel and all accounting functions (except corporate reporting) are
done at the regional level.

B-L sells merchandise exclusively in the over-the-counter market. These
sales have been both cash and credit sales. The credit sales are made via
8-L's own credit card. The credit approval and receivables/collection








80










activities are conducted at the regional level (but based on corporate
policies). Sales made using other credit cards (i.e., VISA and Master Card)
and layaway sales are considered as cash sales.

The Present Sales/Receivables Subsystem

The present system is designed only for over-the-counter sales. A
flowchart of the present system is attached (see Figure 1).
Sales originate at the individual retail stores. The automated cash
registers extend and foot sales tickets based on an internally stored
price list. The prices are updated and verified when necessary (two or
three times a week). The sales information for each store is automatically
tabulated to form the Daily Transaction Listing. Two copies of the Daily
Transaction Listing are sent to the regional headquarters daily along with
the second copy of the Sales Tickets. Deposit slips are prepared and for-
warded with the cash receipts to the bank daily. One copy of the validated
deposit slip is given to the manager and another Is sent directly to the'
regional office. The Daily Transaction Listing and the deposit slips are
reconciled daily at the regional office.

Sales returns for cash are approved by the store manager in accordance
with corporate policy. Other returns are handled by the cashier supervisor.

As part of the Daily Transaction Listing, information on credit sales
is transmitted to the regional office daily. This data is posted to the
Accounts Receivable Subsidiary ledger by the accounts receivable clerk.
All posting is done by preparing a batch of transactions which is sent to
Data Processing. Checking of batch control totals and making corrections
are the responsibility of the person who provides the original input.
Monthly statements are sent directly to the customer by the accounts re-
ceivable clerk. The accounts are aged monthly and the listing is reviewed
by the regional credit manager.
The regional office's incoming mail (including payments on account for
the region) is received in the mailroom. In the mailroom all checks are
restrictively endorsed and a cash receipts listing is prepared. The cash
receipts listing is forwarded to the Accounts Receivable Department for
posting to the Accountants Receivable Subsidiary ledger. The checks and
remittance advices are forwarded to the Cashier's Department for posting
to the Cash Receipts Journal. At the end of the day the Account Receivable
Department and the Cashier's Department agree on the amounts posted. A
bank deposit slip is prepared ll the Cashier's Department and forwarded
with the checks to the bank for deposit in B-L's checking account. B-L's
checking accounts are reconciled monthly by the Controller.

The Reason for the Feasibility Study
The Vice President for Sales and Marketing has suggested that 8-L
move into nail-order sales. The course of action being considered is to
expand the role of the regional offices to include these mail-order centers.



















Customer orders will be received and the merchandise will be shipped di-
rectly to the customer from the new regional mail-order centers. There are
no planned changes in the product lines being considered at this time.

This project is presently undergoing a feasibility study. The Sales
Department is forecasting sales and demands on inventory. Given these
estimates and the inputs from Internal Auditing the Personnel Department
will estimate additional personnel requirements. Other details are being
handled by a person in the President's office.

The portion of the feasibility study which involves you (as a member
of the Internal Auditing staff) is the designing of the Sales/Receivable
internal control subsystem for one of the individual mail-order centers
which provides efficient and effective internal controls. Though you are
not asked to estimate the actual manpower requirementsyou should consider
the eventual cost of the controls when you make your recommendations.

Your Task

This task is limited to the Sales/Receivables cycle of the internal
control system. You can assume that:

1. The internal controls in the other accounting areas are strong.

2. The procedures for handling the cash receipts and making the de-
posits will remain as they are presently.

3. As a matter of policy all documents are sequentially numbered.

4. The EDP system is a batch system whose controls are strong and
it is outside the scope of this study.

5. All merchandise shipped to customers will be made via a reliable
independent carrier.
Two forms have been provided for your inputs to the feasibility study.
The first is a blank form for a flowchart of your system. The second is
a form for you to "document" your system.

To simplify the task of flowcharting the system, the flowchart form
has been labeled with a set of possible departments. You may exclude any
department that you consider to be unnecessary. If any additional depart-
ment is desired, use the column labeled "Other Department."

A set of prelabeled "peel-off" symbols has been included. These repre-
sent the documents, journals and books you may wish to include in your
design. The set includes various forms with many copies. You may exclude
any of these you find unnecessary or add to the set of documents by using
blank labels to create the documents or journals you want. This simple
way of flowcharting can be summarized in the following steps:








82










1. Remove the appropriate symbol and place it on the flowchart
where you want it (the labels can be moved if you change your
mind).
2. Draw an arrow and indicate a) where the document would be sent
(another department or to be filed) or b) the operation to be
performed.
3. Repeat these steps until the copies of all the necessary documents
are flowcharted.
After you have completed your flowchart, identify all the accounting
controls in your flowchart using the format provided on the forms labeled
"Documentation of Controls." In order to properly document your system
each control should be identified by listing: (1) the person involved,
(2) the documentss, (3) the operation performed, and most importantly
(4) a-brief description of what type of errors) is prevented or detected
by each control.
Please indicate the date you completed this work and initial the forms
so I can identify you when we meet at your company. Finally, return the
entire packet to the appropriate person at your company by the date they
Indicated.


Abbreviations used on the "peel-off" labels

A/R Sub Accounts Receivable Subsidiary Ledger
B of L Bill of Ladding
C & Chks Cash and Checks
C/R Jour Cash Receipts Journal
Csh Lst Cash Listing
Cr Memo Credit Memo
CO Customer Order
Gen Led General Ledger
M Stmt Monthly Statement of Account
Process any processes) or activity performed
RA Remittence Advice
Sales Jrn = Sales Journal
SI Sales Invoice
SO Sales Order
SR&A Jrn = Sales Returns and Allowances Journal
UAAF Uncollectable Account Authorization Forms

O are connectors indicating document flow.

A indicate file storage (0 = by date; N numerically;
A = alphabetically).

E indicate documents, journals, and books.


















- 4 -
t \, ',
SL __


/'\a \

"F " '^ "
i- s1 __ J 1- ^ ~

'"'"" T r x iiiii.; -
-S Fi2 1.


!,1-
SIS

It






ip il1
lii 5
E-.i Z*=
'iifc :|2
351f I"il" -




;.ai~s
^-~ i-sjl


: 0 ila?^.












I
---------------------------------------


-
f.

i

~I


-------------------------------------







85





Documentation of Controls
Initials:
Date:
Personnel Document(s) Operation Perfored Error Detected or Controlled
















Appendix B

Design Materials for the OTH Condition








87










Thank you for volunteering your time to participate in this study.
Its purpose is to gather information about the role internal auditors play
in the design and evaluation of internal control systems. You are being
asked to provide a preliminary design for the purchasing internal con-
trol subsystem for a hypothetical firm. Later when we meet at your company,
you will be asked to evaluate a system as though it had been implemented.
There will be no need for you to consult reference books or your colleagues,
there is no right answer to this problem. I simply want a preliminary
type of solution. I realize that those of you who have volunteered to
assist in this study have very diverse experience and backgrounds. That
diversity is good because it is also true of internal auditors in general.

You will be acting as representatives of your profession. The Insti-
tute of Internal Auditors (which is financially supporting this study) is
interested in the implications of what I may find for the profession.
Thus, it is important that you take these tasks seriously and perform as
though this assignment was part of your regular job. I assure you that
your individual responses will be kept strictly confidential and will not
be used for any purpose other than my study.

The remainder of this packet of material includes the following items:

1. Some background material about the hypothetical firm.

2. The requirements for the Purchasing subsysten.

3. Two forms for you to provide your (internal auditing's) design
and documentation.

The Burke-Lvne Case

Assume that you are on the internal auditing staff of Burke-Layne
and Company (hereafter, B-L). B-L is a national, retail chain which sells
a broad range of general merchandise (i.e., clothes, appliances, housewares,
jewelry, etc.). B-L presently has 237 retail stores organized into seven
regional sales divisions. Each region is a profit center with regional
offices located with the regional warehouse. B-L's vendors ship merchan-
dise directly to the warehouses which in turn supply the retail stores in
the region.

All Purchasing and General Administration (including the internal aud-
iting staff) is centralized in Chicago at the corporate headquarters.
Personnel and all accounting functions (except corporate reporting) are
done at the regional level.

B-L sells merchandise exclusively in the over-the-counter market.
These sales have been both cash and credit sales. The credit sales are
made via B-L's own credit card. The credit approval and receivables/
collection activities are conducted at the regional level (but based on
corporate policies). Sales made using other credit cards (i.e., Visa and
Master Card) and layaway sales are considered as cash sales.







88











Your Task
You are asked to provide a preliminary design for 8-L's Purchasing
cycle. All goods are purchased from outside vendors. You tas a member of
the Internal Auditing staff) are asked to design an internal control sub-
system which will provide efficient and effective internal controls over
inventory purchases (not fixed asset acquisitions). Though you are not
asked to estimate personnel requirements, you should consider the eventual
cost of the controls when you make your recommendations.
This task is limited to the Purchasing cycle of the internal control
system. You can assume that:

1. The internal controls in the other accounting areas are strong.
2. As a matter of policy all documents are sequentially numbered.
3. The EDP system is a batch system whose controls are strong and it
is outside the scope of this study.
4. All goods are shipped via a reliable independent carrier.
5. The inventory management controls are outside the scope of this
study. The Purchase Requisitions are properly authorized and
are consistent with good inventory management.
Two forms have been provided for your inputs. The first is a blank
form for a flowchart of your system. The second is a form for you to
"document" your system. A sample flowchart of part of a Payroll cycle is
attached as Figure I to provide an example of what is expected for the
Purchasing cycle.
To simplify the task of flowcharting the system, the flowchart form
has been labeled with a set of possible departments. You may exclude any
department that you consider to be unnecessary. If any additional depart-
ment is desired, use the column labeled "Other Department."

A set of prelabeled "peel-off" symbols has been included. These
represent the documents, journals and books you may wish to include in your
design. The set includes various forms with many copies. You may exclude
any of these you find unnecessary or add to the set of documents by using
blank labels to create the documents or journals you want. This simple
way of flowcharting can be summarized in the following steps:

1. Remove the appropriate symbol and place it on the flowchart where
you want it (the labels can be moved if you change your mind).
2. Draw an arrow and indicate: a) where the document would be sent
(another department or tobe filed) or b) the operation to be
performed.
3. Repeat these steps until the copies of all the necessary documents
are flowcharted.


















After you have completed your flowchart, identify all the accounting
controls in your flowchart using the format provided on the forms labeled
"Documentation of Controls." In order to properly document your system
each control should be identified by listing: (1) the person involved,
(2) the documentss, (3) the operation performed, and most importantly
(4) a brief description of what type of errors) is prevented or detected
by each control.
Please indicate the date you completed this work and initial the forms
so I can identify you when we meet at your company. Finally, return the
entire packet to the appropriate person at your company by the date they
indicated.


Abbreviations used on the "peel-off" labels

A/P Sub = Accounts Payable Subsidiary Journal
C/D Jrn Cash Disbursements Journal
Gen Led = General Ledger
P Req = Purchase Requisition
PO = Purchase Order
Process = any processes) or activity performed
R Rep Receiving Report
V Inv Vendor's Invoice
V Stat Vendor's Statement



O are connectors indicating document flow.

A indicate file storage (0 = by date; N = numerically; A alphabetically).



E indicate documents, journals, and books.




















Timekeeping Payroll Dept


Abbreviations

EERecords = Employee Earning Records
Pay register Payroll Register
R&DA Rate and Deduction Authorization


File Ledgend

A alphabetically
D by date
N numerically


A flowcharting example
from the Payroll cycle.

FIGURE 1B


Personnel Dept





--------------------------------------










-------------------------------------


- - - - - - - - - - - - - - - - - - -
















Appendix C

Review Materials for the ERO and OTH Conditions