Mitigating work conditions that can inhibit learning from errors: Benefits of error management climate perceptions

Abstract

Introduction
Professionals do not always learn from their errors; rather, the way in which professionals experience errors and their work environment may not foster, but can rather inhibit error learning. In the wake of a series of accounting scandals, including Royal Ahold in Netherlands, Lehman Brothers in the United States, and Wirecard in Germany, within the context of financial auditing, we explore four audit-specific conditions at the workplace that could be negatively associated with learning: small error consequences, routine-type errors, negative emotions, and high time pressure. Then, we examine how perceptions of an open or blame error management climate (EMC) moderate the negative relationship between the four work conditions and learning from errors.

Methods
Using an experiential questionnaire approach, we analyze data provided by 141 Dutch auditors across all hierarchical ranks from two audit firms.

Results
Our results show that open EMC perceptions mitigate the negative relationship between negative emotions and error learning, as well as the negative relationship between time pressure and error learning. While we expected that blame EMC perceptions would exacerbate the negative relationship between negative emotions and error learning, we find a mitigating effect of low blame EMC perceptions. Further, and contrary to our expectations, we find that blame EMC perceptions mitigate the negative relationship between small error consequences and error learning, so that overall, more error learning takes place regardless of consequences when participants experience a blame EMC. Post-hoc analyses reveal that there is in fact an inverted- U-shaped relationship between time pressure and error learning.

Discussion
We derive several recommendations for future research, and our findings generate specific implications on how (audit) organizations can foster learning from errors.

Full text

Frontiers in Psychology 01 frontiersin.org
Mitigating work conditions that can
inhibit learning from errors: Benefits
of error management climate
perceptions
Oscarvan Mourik
1†, ThereseGrohnert
2†* and AnnaGold
1
1 Department of Accounting, School of Business and Economics, Vrije Universiteit Amsterdam, Amsterdam,
Netherlands, 2 Department of Educational Research and Development, School of Business and Economics,
Maastricht University, Maastricht, Netherlands
Introduction: Professionals do not always learn from their errors; rather, the way in
which professionals experience errors and their work environment may not foster,
but can rather inhibit error learning. In the wake of a series of accounting scandals,
including Royal Ahold in Netherlands, Lehman Brothers in the United States, and
Wirecard in Germany, within the context of financial auditing, we explore four
audit-specific conditions at the workplace that could be negatively associated with
learning: small error consequences, routine-type errors, negative emotions, and
high time pressure. Then, we examine how perceptions of an open or blame error
management climate (EMC) moderate the negative relationship between the four
work conditions and learning from errors.
Methods: Using an experiential questionnaire approach, we analyze data provided by
141 Dutch auditors across all hierarchical ranks from two audit firms.
Results: Our results show that open EMC perceptions mitigate the negative relationship
between negative emotions and error learning, as well as the negative relationship
between time pressure and error learning. While we expected that blame EMC
perceptions would exacerbate the negative relationship between negative emotions
and error learning, we find a mitigating eect of low blame EMC perceptions. Further,
and contrary to our expectations, we find that blame EMC perceptions mitigate the
negative relationship between small error consequences and error learning, so that
overall, more error learning takes place regardless of consequences when participants
experience a blame EMC. Post-hoc analyses reveal that there is in fact an inverted-
U-shaped relationship between time pressure and error learning.
Discussion: We derive several recommendations for future research, and our findings
generate specific implications on how (audit) organizations can foster learning from
errors.
KEYWORDS
learning from errors, error management climate, error consequences, error type, emotions,
time pressure
1. Introduction
Human errors at work are unavoidable, even when organizations develop sophisticated systems
for their prevention (van Dyck etal., 2005; Frese and Keith, 2015; Metcalfe, 2017). Eorts to prevent
planes from crashing and patients from dying have had a signicant impact, yet are unable to fully
guard against errors at work (Ely etal., 2011; Hagen, 2013; Seckler etal., 2017). Consequently, eective
error management in organizations requires both prevention and subsequent learning from errors that
do occur (van Dyck etal., 2005), dened as “the process through which individuals (a) reect on errors
that they have made, (b) locate the root causes of the errors, (c) develop knowledge about
TYPE Original Research
PUBLISHED 20 January 2023
DOI 10.3389/fpsyg.2023.1033470
OPEN ACCESS
EDITED BY
Thorsten Semrau,
University of Trier,
Germany
REVIEWED BY
Hendrik Wilhelm,
Witten/Herdecke University,
Germany
Stefania Bumbuc,
Nicolae Bălcescu Land Forces Academy,
Romania
*CORRESPONDENCE
Therese Grohnert
t.grohnert@maastrichtuniversity.nl
†These authors have contributed equally to this
work and share first authorship
SPECIALTY SECTION
This article was submitted to
Organizational Psychology,
a section of the journal
Frontiers in Psychology
RECEIVED 31 August 2022
ACCEPTED 04 January 2023
PUBLISHED 20 January 2023
CITATION
van Mourik O, Grohnert T and Gold A (2023)
Mitigating work conditions that can inhibit
learning from errors: Benefits of error
management climate perceptions.
Front. Psychol. 14:1033470.
doi: 10.3389/fpsyg.2023.1033470
COPYRIGHT
© 2023 van Mourik, Grohnert and Gold. This is
an open-access article distributed under the
terms of the Creative Commons Attribution
License (CC BY). The use, distribution or
reproduction in other forums is permitted,
provided the original author(s) and the
copyright owner(s) are credited and that the
original publication in this journal is cited, in
accordance with accepted academic practice.
No use, distribution or reproduction is
permitted which does not comply with these
terms.

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 02 frontiersin.org
action–outcome relationships and the eects of these relationships on the
work environment, and (d) use this knowledge to modify or improve their
behavior or decision making” (Zhao, 2011, p. 436). In this study,
weexplore error learning in the context of nancial statement auditing.
Auditors assess whether organizations have reported their nancial
statements fairly and in line with international reporting standards, a
complex task that is carried out in hierarchical teams. Auditors’ judgments
of a client’s nancial statements are communicated to the wider public,
such as governments, investors, and other stakeholders. Over the last
decades, the domain of auditing has experienced a series of scandals, such
as Royal Ahold in Netherlands, Lehman Brothers in the UnitedStates, and
Wirecard in Germany, resulting in tighter regulation, public oversight, and
public shaming of key actors (i.e., Storbeck, 2020; Rotteveel, 2022).
Consequently, audit rms are investing signicantly in procedures that
foster both error prevention and subsequent learning from errors (e.g.,
ICAEW, 2016; KPMG, 2016; FRC, 2021). Meanwhile, research on error
learning in auditing is still scarce (i.e., Gold etal., 2022; Smeets etal., 2022).
Learning from errors does not occur spontaneously; rather, it is an
eortful process that requires time, resources, and vulnerability, which
may not always beavailable or desirable in the workplace (Zhao, 2011;
Lei etal., 2016; Metcalfe, 2017). Prior error learning research indicates
that individual learning from errors depends on whether conditions at
work are perceived by individuals to positively or negatively aect their
self-worth and well-being (Kanfer and Ackerman, 1989; Edmondson,
2004; Tulis etal., 2016). Building on insights from an interview study
with auditors on error management by Gold etal. (2022), weexplore
four work conditions that could inhibit auditors’ error learning in daily
practice: small consequences for errors, routine-type errors, strong
negative emotions, and high time pressure. First, weexpect that auditors
will beless likely to engage in error learning when an error has smaller
(rather than larger) error consequences (in line with, i.e., Levitt and
March, 1988; Cannon and Edmondson, 2005; Homsma etal., 2009).
at is, errors with smaller consequences can beconsidered less ‘learn-
worthy’; however, ignoring their learning potential may lead to
repetition and escalation in the future (Cannon and Edmondson, 2005).
Second, wehypothesize that auditors will report less error learning from
routine (compared to non-routine) errors (in line with Embrey, 2005;
Zhang etal., 2019), as these errors are easily attributable to inattention
or coincidence, rather than a lack of knowledge (Sutclie and Rugg,
1998). ird, wehypothesize that auditors will learn less from errors
when they experience strong (rather than weak) negative emotions, as
these emotions take up crucial cognitive resources needed for learning,
and may inhibit error learning (in line with, i.e., Heimbeck etal., 2003;
Zhao etal., 2019; Smeets etal., 2022). Finally, weexpect that auditors
will report less error learning with high (rather than low) time pressure,
as they are likely to prioritize urgent tasks over learning, and because the
error’s cause can beexternalized (in line with, i.e., Zhao and Olivera,
2006; Homsma etal., 2009; Putz etal., 2012).
To date, most research focuses on conditions that foster error
learning (Rybowiak etal., 1999; van Dyck etal., 2005; Bligh etal., 2018).
By exploring work conditions that may inhibit error learning,
wecontribute to extant research by identifying where organizations can
intervene to create necessary conditions for enabling error learning. To
this end, weexamine how auditors’ error management climate (EMC)
perceptions interact with the four work conditions with regard to error
learning. EMC describes the beliefs, norms, and practices related to
how errors are dealt with that are shared within an organization (van
Dyck etal., 2005). An open EMC promotes opportunities for learning
from errors. Weexpect that the negative eect of the aforementioned
conditions on error learning will bemitigated (i.e., weakened) when
auditors perceive that they work in an open EMC. On the other hand,
the negative eect of the work conditions on learning will
beexacerbated (i.e., strengthened) when auditors work in a perceived
blame EMC, where errors are typically seen as personal failures and are
therefore punished (van Dyck etal., 2005). By studying the interaction
between EMC perceptions and the four conditions with regard to error
learning, wecontribute to extant research on the direct relationship
between EMC and error learning (for reviews, see Lei etal., 2016;
Metcalfe, 2017). Wealso provide novel insights into how organizations
can mitigate the negative impact of the four work conditions on
error learning.
2. Hypothesis development
In their interview study on error management in auditing, Gold
et al. (2022) reveal that in the wake of signicant pressure on the
profession, audit rms predominantly focus on error prevention, rather
than fully embracing the learning potential of errors to improve their
performance. Practitioners describe a series of work conditions that
appear to inuence error learning, such as the potential consequences
of errors for the client, the types of errors dealt with, and experiencing
emotions in connection with errors, along with experienced time
pressure. In this study, weexplore how these four work conditions relate
to error learning by building hypotheses that follow prior research.
2.1. Work conditions inhibiting error learning
2.1.1. Small error consequences
First, auditors, like other professionals, make errors that vary in
their consequences (Gold etal., 2022). While all errors, regardless of their
consequences, carry learning potential (van Dyck, 2009), prior research
has shown that learning is more likely to occur when an error has
relatively larger consequences, usually aecting the person committing
the error or others, with regard to individuals’ health, nances, or social
standing (Cannon and Edmondson, 2005; Homsma etal., 2009). ese
errors challenge the existing state of aairs and stimulate individuals to
engage in learning to prevent these signicant consequences (Levitt and
March, 1988). Meanwhile, professionals typically fail to learn from
errors with relatively minor consequences (Baumard and Starbuck,
2005). Despite their theoretical learning potential, errors with smaller
consequences are more easily discounted as irrelevant because, as shown
by Baumard and Starbuck (2005), professionals focus on achieving
expected outcomes; errors that do not signicantly aect these outcomes
are not made a priority for learning (see also Levitt and March, 1988;
Cannon and Edmondson, 2005). Hence, failing to learn from errors with
minor consequences may have serious long-term implications.
Following these insights from prior research, we formulate our
rst hypothesis:
Hypothesis 1a: Professionals will report less error learning when they
experience small (rather than large) error consequences.
2.1.2. Routine errors
Second, professionals may not deem all error types to beequally
‘learn-worthy’. Within organizational psychology, the inuential

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 03 frontiersin.org
classication of error types by Rasmussen (1986) distinguishes between
errors that are routine and non-routine. On the routine side, skills-based
errors occur during recurring and predictable situations and typically
result from lack of attention/memory, while knowledge to solve the
problem is technically present and available (in line with Sutclie and
Rugg, 1998; Zhang etal., 2019). For example, a professional may forget
to attach a document to an email when in a hurry. Non-routine errors
occur either when the necessary knowledge to solve a problem is present
but not used correctly, or when a task is so complex and/or unfamiliar
that new knowledge is needed to address the situation (respectively
known as rules-based and knowledge-based errors; Sutclie and Rugg,
1998; Zhang etal., 2019), for example, applying a checklist or decision
rule in the wrong context. e wider error learning literature shows that
with non-routine errors, individuals are typically challenged to create
understanding of complex contextual information related to the error,
forcing them to reconsider the applicability and limitations of their
knowledge, activities naturally linked to learning (Rasmussen, 1986;
Embrey, 2005). In contrast, routine errors are oen ignored, as they are
not perceived to berelated to knowledge, and hence, the need to learn
(Embrey, 2005). In this study, wetherefore hypothesize the following:
Hypothesis 1b: Professionals will report less error learning when they
experience routine (rather than non-routine) errors.
2.1.3. Strong negative emotions
ird, emotions that individuals experience when discovering they
made an error may also inhibit error learning (Zhao, 2011; Rausch etal.,
2017). Past research has established that experiencing strong negative
emotions such as shame, fear, or guilt can limit learning for two principal
reasons. First, experiencing these emotions may cause professionals to
withdraw from the situation, missing out on important learning
opportunities (Zhao etal., 2019). Second, these emotions occupy a
person’s working memory, limiting the information processing that is
essential for leaning (Zhao, 2011; Tulis etal., 2016; Smeets etal., 2022).
Literature on emotional regulation has shown that individuals prioritize
relieving strong emotions over problem-solving and learning, with the
consequence that strong negative emotions may inhibit error learning
(Kanfer and Ackerman, 1989; Ilgen and Davis, 2000; Tulis etal., 2016;
Hökkä etal., 2020). Yet, studies show that when individuals succeed in
tempering these negative emotions, they free up mental space to adopt
a more accepting view of errors, resulting in a stronger motivation to
learn (van Dyck etal., 2005; Frese and Keith, 2015). Indeed, experimental
research shows that individuals are more eective in learning from
errors when they do not experience strong negative emotions (Heimbeck
etal., 2003), leading us to the following:
Hypothesis 1c: Professionals will report less error learning when they
experience stronger (rather than weaker) negative emotions related
to the error.
2.1.4. High time pressure
Fourth, auditors frequently experience time pressure (Pierce and
Sweeney, 2004; Gold etal., 2022), rooted in the cyclical and commercial
nature of auditing. Auditors are faced with strict ling deadlines, as an
audit becomes more protable when auditors use fewer hours than paid
for upfront by the client (i.e., Kelley and Margheim, 1990; DeZoort and
Lord, 1997). Several organizational behavior studies suggest that
contextual factors causing time pressure can impair learning from errors
for two main reasons (Zhao and Olivera, 2006; Homsma etal., 2009;
Putz etal., 2012). First, when individuals attribute errors to external
causes such as time pressure, it is less likely that they will recognize the
opportunity to learn, as they rst seek closure (Ellis and Davidi, 2005;
Putz etal., 2012). While such a strategy enables them to meet important
deadlines, it may be problematic for learning because it leads to
supercial error analysis, at best (Kruglanski etal., 1993; Putz etal.,
2012). Second, high time constraints may cause individuals to use
information-processing strategies that limit their cognitive capability
(Zhao and Olivera, 2006), which is an essential element of learning from
errors (Rasmussen, 1986; Metcalfe, 2017). Both Ford etal. (1989) and
Zhao and Olivera (2006) found that professionals ignore competing
hypotheses and lter out contradicting information under time pressure,
limiting decision making in the short term and learning in the long
term, as was also found in two related audit-specic studies (Choo, 1995;
Glover, 1997). Consequently, wehypothesize the following:
Hypothesis 1d: Professionals will report less error learning when
they experience more (rather than less) time pressure.
2.2. The moderating role of EMC for error
learning
We explore whether the negative relationships between the four
work conditions and error learning is moderated by auditors’
perceptions of the error management climate (EMC). In their landmark
paper, van Dyck etal. (2005) distinguish between two facets of EMC; an
open EMC is characterized by organizations or management showing a
high tolerance for making errors, as long as learning occurs and errors
are not repeated. At the other extreme, in a blame EMC, making errors
is considered unacceptable and therefore should bepenalized to prevent
reoccurrence. Weaim to understand whether perceptions of an open
EMC can mitigate the negative consequences of the four work conditions
for error learning, while perceptions of a blame EMC may exacerbate
the negative relationships predicted in Hypothesis 1a–d. is argument
builds on Salancik and Pfeer's (1978) social information-processing
theory, which posits that individuals interpret their work environment
through their own personal lens based on past experiences, socialization,
and personal values, and that this personal interpretation, rather than
the work environment itself, drives an individual’s attitudes and
behaviors within the work context. For non-error-specic learning, this
notion is empirically shown by, for example, Eldor and Harpaz (2016),
Park and Rothwell (2009), and Mikkelsen etal. (1998). Consequently,
wefocus on individual perceptions of error management values, beliefs,
and behaviors within audit rms in relation to the four work conditions
and individual-level error learning.
2.2.1. Open EMC perceptions as moderator
An open EMC is driven by management that actively encourages
organizational members to learn from their errors by emphasizing the
omnipresence of errors, encouraging them to analyze error causes to
develop strategies for preventing the same errors from re-occurring in
the future, and shielding individuals who make errors from being
punished (van Dyck etal., 2005; Keith and Frese, 2011). Prior audit
research shows that an open EMC increases auditors’ willingness to
share errors aer discovering them (Gronewold etal., 2013; Gold etal.,

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 04 frontiersin.org
2014), and that it aects the degree to which auditors feel responsible for
acting on errors (Gronewold and Donle, 2011). To date, audit research
on EMC focuses on error reporting (e.g., Stefaniak and Robertson, 2010;
Gronewold etal., 2013); research on actual error learning in this context
is rare (Gold etal., 2022; Smeets etal., 2022), as are studies that explore
the interaction of perceived EMC with other factors (van Dyck etal.,
2005; Zhao and Olivera, 2006; Gronewold and Donle, 2011).
Consequently, wealso build on extant ndings in other contexts and on
related concepts to develop our next hypotheses.
First, regarding small error consequences, Stefaniak and Robertson
(2010) showed that lower-ranking auditors are more likely to report
errors when their direct supervisors did not punish errors in the past.
ey conclude that supervisor behavior communicates acceptable
behavior within a group, so that learning from errors—even when the
error consequences are small—is increased by open EMC perceptions.
As a result, wepredict that the negative eect of small error consequences
on error learning is mitigated by open EMC perceptions. Second, while
prior research acknowledges that error type matters for the eectiveness
of error management (e.g., Zhao and Olivera, 2006), wewere unable to
identify prior research on its interactions with EMC. However, an open
EMC emphasizes learning potential across all error types (e.g., van Dyck
etal., 2005), which should mitigate the dierence between routine and
non-routine errors for learning. ird, concerning strong negative
emotions, research suggests that subordinate emotions are closely
related to leader expectations and attitudes (i.e., Edmondson, 2004;
Gronewold and Donle, 2011; Zhao, 2011). Both Smeets etal. (2022) and
Grohnert etal. (2019) found that beginning auditors engage in more
error learning when they perceive that they work in a supportive climate
for learning from errors, while also experiencing less strain.
Consequently, weexpect open EMC perceptions to mitigate the negative
relationship between negative emotions and error learning. Finally,
studies on error learning frequently mention time pressure as a
characteristic of the environment in which errors are made, but do not
explicitly study its impact on error learning or how it interacts with
climate (i.e., Zhao and Olivera, 2006). In one recent study, Zhao etal.
(2022) found that professionals working in an open EMC reported more
learning from errors, even when faced with work stressors that include
time pressure. Wetherefore expect auditors’ open EMC perceptions will
alleviate the negative relationship between high time pressure and error
learning. is leads to our rst interaction hypothesis:
Hypothesis 2: Professionals’ perceptions of working in an open EMC
mitigate the negative relationship between error learning and (a)
small rather than large error consequences, (b) routine rather than
non-routine errors, (c) stronger rather than weaker negative
emotions, and (d) more rather than less time pressure.
2.2.2. Blame EMC perceptions as moderator
In contrast to an open EMC, in a blame EMC, upper management
shares an attitude of ‘getting things right the rst time’, without
acknowledging that eradicating all errors is impossible. Initiatives to
learn from errors are, at best, lip service, and organizational members
are led to believe that making errors leads to formal and informal
sanctions, such as lower performance evaluations (van Dyck etal.,
2005; Frese and Keith, 2015). We expect that such a blame EMC
perception will exacerbate the negative relationships between the four
work conditions and error learning predicted in Hypothesis 1.
Overall, wefound fewer studies focusing specically on a blame EMC
(and related concepts), compared to research on an open EMC. e
studies by Fruhen and Keith (2014) and Matthews etal. (2022) are
two exceptions, showing that in a blame EMC, errors are actually
more likely to occur, although professionals are not more likely to
manage these errors, for example, by learning. Further, working in a
blame EMC is associated with negative emotions of fear and stress
resulting from blame and punishment (Gorini etal., 2012; Matthews
et al., 2022), suggesting a greater likelihood for strong negative
emotions to adversely aect error learning. Wewere unable to identify
studies that explicitly link perceptions of a blame EMC to error type
and time pressure with regard to (error) learning; however, the
theoretical logic maintains that the negative relationships predicted
in Hypotheses 1b and 1d are likely to be exacerbated when
professionals perceive the EMC to beone of blame. Based on the
established outcomes associated with a blame EMC and related
concepts in prior research, we formulate our second and nal
moderation hypothesis. All hypotheses are illustrated in Figure1, our
conceptual model.
Hypothesis 3: Professionals’ perceptions of working in a blame EMC
exacerbate the negative relationship between error learning and (a)
small rather than large error consequences, (b) routine rather than
non-routine errors, (c) stronger rather than weaker negative
emotions, and (d) more rather than less time pressure.
3. Materials and methods
3.1. Setting and sample
We contacted 247 practicing auditors across hierarchical ranks
working for two dierent audit rms in Netherlands. In one rm,
participants completed an experiential survey as part of rm-internal
in-person training sessions, in the presence of at least one of the authors.
Authors present during the study were unknown to the participants and
provided support in case of technical diculties, along with a brieng
and debrieng. ese trainings were endorsed by top management, so
that mandatory attendance of participants limited self-selection bias.
e other rm recruited participants as part of internal training sessions
for qualifying auditors, meaning in the rst 3 years of practice, and
provided dedicated time for completing our instrument. Screening of
the responses revealed that 98 responses were incomplete (91
participants abandoned the study aer reading the instructions; seven
participants indicated that they did not have an error to report before
abandoning the instrument), a further eight observations were excluded
as the entire instrument was completed in an unrealistically short
timeframe, less than 5 minutes. Our nal sample therefore includes 141
usable records (57% of the total response). Auditors in this sample are
between 21 and 56 years of age (M = 29.87, SD = 7.46), and 30% are
publicly registered accountants in Netherlands. Our sample includes 44
sta auditors, the lowest rank, followed by 34 senior sta auditors, 27
managers, 16 senior managers, and eight directors/partners, the last
being the highest rank (12 missing values), and includes 67% male
participants, in line with the pyramid structure and common gender
balance in audit rms (EY, 2018; KPMG, 2018; PwC, 2018). Moreover,
93% work in an audit function, 6% work in audit-adjacent roles, such as
advisory (two missing values); 92% work for a large audit rm, while 7%
work for a smaller rm (one missing value).

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 05 frontiersin.org
3.2. Materials and measures
e starting point of our experiential questionnaire was developed
in close collaboration with an experienced audit partner, asking
participants to recall and describe a specic error, which was the focus
for the remainder of the questionnaire. Aer assuring participants of the
anonymity and condentiality of all information provided, the
experiential questionnaire started with the following denition of an
error: “Any unintentional action or omission by you that caused
consequences for audit planning, procedures or other goals during your
or your peers’ audit work.” Weacknowledge that this denition departs
somewhat from common denitions in the error literature, where errors
are explicitly separated from their consequences (e.g., van Dyck, 2009;
Klamar etal., 2022). Rather than presenting an academic denition of
errors, the description weprovided to survey respondents was somewhat
narrower than denitions commonly used in the literature (Rasmussen,
1986; Reason, 1990; Sitkin, 1992; Zhao and Olivera, 2006), as welimited
the requested respondent recall to errors that were in fact followed by
consequences. ere were two reasons for this instruction, which was
developed together with a practicing auditor. First, when auditors are
confronted with the term ‘error’ they immediately associate it with
errors made by their client (i.e., nancial statement errors).
We intentionally aimed to direct their attention to errors made by
themselves that would potentially aect not only the client, but also their
own or their peers’ work. As a result, we emphasized the notion of
consequences in the survey denition. Second, since one of our key
interests lies in the eect of error consequences on error learning,
weaimed to direct respondents’ attention to errors that would carry
variability in consequences. While some of the errors wetarget may
indeed beclassied as failure in the literature, werefrained from using
this value-laden label in our survey. Finally, we recognize that the
denition might bias responses away from errors with zero
consequences; on the other hand, weexpected that recall of such errors
might berelatively minor. As will beshown, wend substantial variance
in the extent to which reported errors carry consequences (including
zero-consequence errors), as intended.
We then asked participants to describe an error that t the
provided error denition. Following the general prompt, participants
answered a range of closed-ended, Likert-type, and open-ended
questions that measured the relation between the described error and
the inhibiting factors, perceived EMC, and learning from the error. In
developing the open-ended questions (see Appendix), weavoided
leading questions to increase accuracy and the participant’s recall and
reporting (i.e., Downey and Bedard, 2019). e questionnaire was
presented in English using original scale items. In a nal step,
participants provided demographic information.
3.2.1. Error learning
In line with the denition of learning from errors by Zhao (2011),
wemeasured error learning using Gronewold and Donle’s (2011) audit-
specic adaptation of Rybowiak etal.’s (1999) scales for error reection,
error analysis, and error knowledge. To suit the setting of the error-
focused experiential survey, weadapted the instrument in two ways.
First, the original statements (e.g., “I oen think about how an error
might have been avoided”) were reformulated into past tense statements,
and participants were instructed to provide answers based on their
chosen error, rather than on general practices within their rm.
Participants responded to 11 items (measured on a scale ranging from
1 “strongly disagree” to 5 “strongly agree”). e rst sub-scale, error
reection, captures participants’ behavior in purposefully reecting aer
discovering errors (example item: “Aer the error, Ithoroughly thought
thoroughly about how to correct it”). e second sub-scale, error
analysis, reects whether participants analyzed the cause of their errors
(example item: “Because Imade a mistake, Ianalyzed it thoroughly.”).
Finally, the sub-scale for error knowledge indicates whether participants
were able to create insights that help to improve future behavior or
decision making (example item: “I learned a lot from my error for
mastering my work.”). We performed principal component analysis
FIGURE1
Conceptual model.

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 06 frontiersin.org
(PCA) with direct oblimin rotation to explore our adapted scale.
Weaccepted a one-component solution aer the removal of two items,
resulting in a reliable overall scale (Cronbach’s alpha = 0.799)
1
.
Wetherefore calculate the mean value of the remaining nine items as
our measure for the dependent variable error learning.
3.2.2. Work conditions that may inhibit error
learning
Following the interview study by Gold etal. (2022), wemeasure four
factors that may inhibit learning from errors: error consequence, error
type, error emotions, and time pressure. First, to measure error
consequence, weinductively coded participants’ descriptions of the error
made and the consequences reported by distinguishing whether errors
resulted in adjustments to procedures, additional work for colleagues,
or an additional information request from the client. ese criteria were
developed with an experienced audit partner and applied by the second
author and an independent coder with audit experience as 0 = no error
consequences (n = 41), 1 = small error consequences (n = 57), and
2 = large error consequences (n =25; 20 error consequences could not
becategorized with the information given). Interrater reliability was
high (Cohen’s kappa = 0.80, p < 0.01).
Second, error type was coded as either routine or non-routine by the
third author and an independent coder with audit experience, in line
with the conceptualizations by Sutclie and Rugg (1998) and Zhang
et al. (2019), both based on Rasmussen (1986). During the coding
process, both raters initially distinguished between skills-, rule-, and
knowledge-based errors; 17 errors could not becoded. At this stage,
interrater reliability was low (Cohen’s kappa = 0.38, p < 0.01), due to
dierent perceptions about the nature of a skills-based error as resulting
either from the person’s perception or from the nature of the task
performed. Aer calibration between coders, skills-based errors were
coded from the perspective of the auditor making the error (1 = routine
error, n = 42), and rule-and knowledge-based error were grouped
together (2 = non-routine error, n = 84), which resulted in high interrater
reliability (Cohen’s kappa = 0.85, p <0.01).
ird, negative emotions are captured using the circumplex model
in line with Rausch etal. (2017), with eight emotional categories, four
positive and four negative. Weasked participants to select at least three
of the eight emotional categories and to indicate how intensely they
experienced their chosen emotional states on a 3-point scale (“1” a little,
“2” somewhat, “3” intensely). Most (79.7%) of participants indicated
that they felt unhappy/gloomy/sad (mean intensity = 2.04, SD = 0.693,
n = 115), followed by 72% who were irritated/annoyed/angry (mean
1 Using all 11 items, wefind acceptable sampling adequacy (KMO = 0.794), and
variables are not overly correlated to aect the outcome of the PCA (Bartlett’s
test of sphericity: X
2
(55) = 0.794, p = 0.000). Initial analyses revealed three
components with eigenvalues above 1, in line with Kaiser’s criterion, explaining
a cumulative 57.3% of the variance. The scree plot showed inflections that would
justify both a one and a three-component solution. The component plot indicates
that nine of the 11 items are closely clustered together, with two items of the
error reflection scale separated (“I did not know how to proceed after my error,
so Irelied on my colleagues” and “I was unable to correct the error by myself,
so Iturned to my colleagues”), which is likely due to the kind of errors reported
by participants. Rerunning the PCA without these two items revealed a
one-component solution (KMO = 0.861; Bartlett’s test p = 0.000; eigenvalue = 3.661,
variance explained = 40.676%).
intensity = 1.97, SD = 0.760, n = 103), and 67.8% who experienced
nervousness/worry/fear (mean intensity = 1.96). e negative emotions
of boredom/dullness/disinterest were only selected by 4.9%. At the same
time, positive emotions were only reported by 11.9–30.1% of the
auditors, with a mean intensity of 1.96. ese positive emotions include
(1) motivated/delighted/curious (selected by 30.1%), (2) condent/
happy/glad (selected by 11.9%), (3) contented/accepted/proud (selected
by 12.6%), and (4) calm/even-tempered/daydreaming (selected by
21.7%). Prior studies including multiple emotions at the same time oen
include them as separate variables (i.e., Reio and Callahan, 2004), or use
a mean score across dierent emotions to indicate the presence of
certain emotions (i.e., Scott and Sutton, 2009). Following Hypothesis 1c
and taking a quantitative approach, weoperationalize error emotions as
the sum score of the emotional intensity of the dominant three negative
emotions selected by participants, focusing on intensity of experience of
negative emotions.
Finally, wecaptured time pressure in a manner as closely as possible
related to auditors’ actual time perceptions. Auditors in practice are
given a budget of time paid for by the client to complete a certain task,
and auditors refer to these budgets when discussing time pressure
(Kelley and Margheim, 1990; Gold etal., 2022). Wetherefore asked what
percentage of additional time participants would have liked to use for
managing the error compared to the available time at that moment
(measured on a 5-point scale and ranging from “between 0 and 20%” to
“between 80 and 100%). is measure anchors all participants on a xed
entity, eliciting a relative measure that is more easily compared across
participants, given individual dierences in perception of time pressure
(e.g., Margheim etal., 2005). Time pressure is measured as participants’
indication of their perceived additional time needed (69 reported 0–20%
extra time needed, 19 reported 21–40%, and 9 wanted more than 41%
extra time).
3.2.3. Error management climate
We captured two facets of the participants’ rm’s perceived error
management climate using van Dyck etal.’s (2005) validated instrument.
Perceived open EMC is measured through 17 items on a ve-point scale,
including “Aer an error, people think through how to correct it” and
“Our errors point us at what wecan improve.” Perceived blame EMC is
measured by 11 items, including “In this organization, people feel
stressed when making mistakes” and “ere are advantages in covering
up one’s errors.” Due to a formatting error that occurred when creating
the online survey, one item from each scale was accidentally omitted
(“Aer making a mistake, people try to analyze what caused it” and
“During their work, people are oen concerned that errors might
occur”), resulting in 16 and 10 items, respectively. PCA analysis with
direct oblimin rotation reveals a two-component solution in line with
van Dyck etal.’s original scales, which is also reected in the reliability
scores for both scales, open and blame EMC (Cronbach’s alpha = 0.848
and 0.802, respectively). Error management climate is therefore
measured through participants’ mean perceived values for both the open
and the blame EMC items at the individual level.
3.2.4. Control variables
Given prior ndings that professionals can experience error-
related strain and vulnerability dierently depending on a range of
characteristics (Metcalfe, 2017; Grohnert etal., 2019), weinclude ve
covariates in our analyses to represent ndings from prior research
as well as to account for the current research setting. In line with the
conceptual model of individual error learning by Tulis etal. (2016),

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 07 frontiersin.org
weinclude three covariates that represent our sample characteristics:
gender, certication status, and rank, as described in the sample
setting above. Additionally, women tend to bein the minority within
auditing, especially at higher ranks, which has been associated with
increased vulnerability in prior studies (Hardies etal., 2011). In line
with this notion, Grohnert etal. (2017) found that female auditors
were more likely to cover up errors, an action that inhibits learning
(e.g., van Dyck etal., 2005). Regarding experience and rank, prior
studies found that professionals with more experience tend to report
more learning from errors as well, which may berelated to their
status in the organization and/or to the ability to make better sense
of errors with more prior knowledge (Carmeli and Gittell, 2009;
Smeets etal., 2022). Methodologically, weaccount for memory eects
by including the number of months that have passed between the
error occurrence and completing the experiential questionnaire
(timing), as reported by participants (Mahajan, 2010; Rausch etal.,
2017). Finally, weinclude rm type as a covariate to take into account
variance in EMC across employers (van Dyck etal., 2005; Vera and
Crossan, 2016).
3.3. Analysis strategy
Aer reporting descriptive statistics and correlations, weexplore
our hypotheses through conditional process modelling using Hayes’
(2018) PROCESS macro for SPSS, and specically model 2, the
moderation model, given our sample size. Conditional process models
estimate the conditional interaction eects and generate bias-
corrected 95% condence intervals (CI) for the interaction eects at
various values of the moderator variable (M−1SD, M, M + 1SD,
labeled as low, medium and high, respectively; Aiken and West, 1991;
Hayes, 2018). Due to the fact that many participants in our sample
work for the same organization, wemade use of robust standard errors
to account for a potential violation of the independence assumption
of OLS regression (Hayes and Cai, 2007). In a preliminary step,
weused Harman’s single factor test to check for common method bias;
with 15.5% of variance explained, common method bias is unlikely to
aect our results.
4. Results
4.1. Descriptive statistics and correlations
Table1 reports the means, standard deviations, and correlations for
all study variables. In line with the categorical descriptive statistics
reported under 3.2.2 wend that overall, auditors reported errors with
mostly small consequences that are mostly non-routine, they
experienced medium-strength negative emotions, and required around
20% additional time to adequately manage the error. At the same time,
auditors reported working in an EMC that is more open (M = 3.91 out
of 5, SD = 0.46) than blame-oriented (M = 2.87 out of 5, SD = 0.64), along
with medium to high error learning (M =3.51 out of 5, SD = 0.58). ese
distributions inform our ndings. Examining the correlations, wend
that error learning correlates positively with error consequences
(r = 0.22, p < 0.05) and error emotions (r = 0.38, p < 0.001), but does not
correlate signicantly with error type and time pressure. In turn, an
open EMC correlates negatively with a blame EMC (r = −0.32, p < 0.001)
and with time pressure (r = −0.30, p < 0.01), and a blame EMC correlates
positively with error emotions (r = 0.33, p < 0.05) and time pressure
(r = 0.19, p < 0.10). Regarding our covariates, female relative to male
auditors reected on errors that took place further in the past (r =0.23,
p <0.01), with larger consequences (r =0.25, p <0.01) and more time
pressure (r =0.26, p <0.05). Auditors of higher relative to lower ranks
also reported errors that took place further in the past (r =0.26, p <0.01),
were less likely to work at smaller audit rms (r = −0.38, p < 0.001), and
experienced a more open EMC (r =0.17, p < 0.05) and more error
learning (r = 0.33, p < 0.001).
TABLE1 Means, standard deviations, and correlations among variables.
Variable MSD (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
(1) Gender 1.33 0.47 1
(2) Certication 1.70 0.46 0.02 1
(3) Rank 2.29 1.21 0.02 0.79*** 1
(4) Timing 9.66 14.37 0.23** 0.27*** 0.26** 1
(5) Firm Type 1.15 0.36 0.02 −0.27** −0.38*** −0.13 1
(6) Consequences 1.13 0.71 0.25** 0.04 0.02 0.17+0.07 1
(7) Type 1.71 0.43 0.06 0.20*0.11 −0.10 0.03 −0.07 1
(8) Emotions 5.93 1.25 0.13 0.07 0.12 −0.03 −0.13 0.15 0.09 1
(9) Time Pressure 1.40 0.72 0.26*0.02 0.06 0.03 0.11 −0.01 −0.18+0.113 1
(10) Open EMC 3.91 0.46 −0.13 0.15 0.17* −0.07 0.03 0.13 0.05 −0.12 −0.30** 1
(11) Blame EMC 2.87 0.64 0.11 0.09 0.11 0.08 −0.17*0.04 −0.11 0.33*0.19+−0.32*** 1
(12) Error
Learning
3.51 0.58 0.08 0.30*** 0.33*** 0.09 −0.03 0.22*0.02 0.38*** 0.08 0.32*** 0.06
N = 141 individuals. For gender, 1 = male, 2 = female, 3 = prefer not to say. For role, 1 = audit, 2 = other. For certication, 0 = no, 1 = yes. For rank, 1 = sta, 2 = senior, 3 = manager, 4 = senior,
5 = director/partner. For timing, scores represent months between error commission and completing the instrument. For rm type, 1 = large rm, 2 = small rm. For consequence, score is between 1
and 3, no to large error consequence. For type, 1 = routine error, 2 = non-routine error. For emotions, score is how intensely participants felt nervous/worried/afraid on a scale from 0 = not at all to
3 = intensely (Rausch etal., 2017). For time pressure, score is between 1 and 5, from 0 to 100% extra time required to address error. For open and blame EMC, score is between 1–5, based on van
Dyck etal.’s EMC and EAC instruments (2005). For error learning, score is between 1–5, based on Gronewold and Donle’s instrument (2011). Signicance of correlations is indicated as + = p < 0.10,
*p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed).

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 08 frontiersin.org
4.2. Hypothesis testing
4.2.1. Error consequences, open/blame EMC, and
error learning (H1a, H2a, H3a)
Table2 reports moderation analyses relating our covariates, the
four work conditions along with the measures for open/blame EMC,
and their interactions to error learning. Model 1in Table2 reports the
ndings for Hypotheses 1a, 2a and 3a, relating error consequences,
open and blame EMC to error learning. Overall, this model is
signicant (F(10,109) = 5.21, p < 0.001) explaining 32% of the variance
in error learning. In line with Hypothesis 1a, we nd that error
consequences relate positively to error learning, so that less learning
takes place from errors with smaller consequences (B = 1.80, p < 0.05;
SE = 0.81). Contrary to Hypothesis 2a, wedo not nd a signicant
interaction for an open EMC (B = −0.20, p > 0.10; SE = 0.15). At the
same time, with respect to Hypothesis 3a, we nd a signicant
interaction for a blame EMC (B = −0.30, p < 0.05; SE = 0.14, ΔR
2
Blame = 0.05, p < 0.05). Figure2 Panel A illustrates this interaction
eect through three conditional eects: more error learning takes
place from errors with larger compared to smaller consequences when
auditors perceive themselves to work in a low (M − 1SD; Blow= 0.349,
p < 0.001) and medium (M; B
medium
 = 0.172, p < 0.01) blame
EMC. However, in a high blame EMC, auditors reported equal and
high error learning regardless of error consequences (M + 1SD;
Bhigh=−0.005, p> 0.10; see Figure2, Panel A). is conditional eect
is contrary to Hypothesis 3a: while weexpected that perceptions of a
blame EMC would further exacerbate the negative eect of small error
consequences on learning, wend instead that blame EMC perceptions
mitigate this relationship.
4.2.2. Error type, open/blame EMC, and error
learning (H1b, H2b, H3b)
Considering Hypotheses 1b, 2b, and 3b focusing on error type in
interaction with open/blame EMC for error learning, wereport our
ndings in Table2, Model 2. is model is signicant (F(10,112) = 3.71,
p < 0.001), explaining 22% of the variance. Yet, contrary to Hypotheses
1a, 2b, and 3b, wend neither signicant direct relationships nor
interactions. Wetherefore do not nd support for the hypothesis that
routine errors can impede error learning, nor for the expectation that
this relationship is moderated by an open or a blame EMC.
TABLE2 OLS Moderation Analysis.
Model 1 Model 2 Model 3 Model 4
BSE BSE BSE BSE
Variable
Gender 0.01 0.12 0.08 0.11 0.08 0.21 0.05 0.14
Certication 0.06 0.14 0.16 0.16 0.21 0.30 0.22 0.18
Rank 0.122** 0.05 0.09 0.05 0.13 0.12 0.04 0.08
Timing 0.00 0.01 0.00 0.01 0.00 0.01 0.00 0.00
Firm Type 0.12 0.17 0.13 0.16 0.00 0.01 0.16 0.16
Consequence 1.80*0.81
Typ e −1.83 2.17
Emotions −0.63 1.20
Time Pressure −1.68 1.20
Open EMC 0.50** 0.18 −0.15 0.39 1.14 1.42 −0.16 0.33
Blame EMC 0.58*** 0.19 −0.27 0.22 0.64 1.15 −0.01 0.25
Consequences × open −0.20 0.15
Consequences × blame −0.30*0.14
Type × open 0.32 0.22
Type × blame 0.19 0.18
Emotions × open 0.25 0.23
Emotions × blame −0.07 0.19
Time × open 0.41+0.24
Time × blame 0.11 0.17
Model
F-statistic 5.21*** 3.71*** 4.04*** 2.81***
R20.32 0.22 0.48 0.23
ΔR2 open 0.01 0.01 0.05*0.05
ΔR2 blame 0.05*0.01 0.01 0.01
N = 141. Table reports unstandardized regression coecients and robust standard errors (Davidson-MacKinnon; Hayes and Cai, 2007). Signicance is indicated as +p < 0.10, *p < 0.05, **p < 0.01, and
***p < 0.001 (2-tailed).

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 09 frontiersin.org
4.2.3. Negative emotions, open/blame EMC, and
error learning (H1c, H2c, H3c)
Table2, Model 3 reports the ndings for Hypotheses 1c, 2c, and 3c,
exploring the links between negative emotions, open/blame EMC, and
error learning. is model is signicant (F(10,44) = 4.04, p < 0.001),
explaining 48% of the variance. Contrary to Hypothesis 1c, wedo not
nd a signicant direct eect of negative emotions, and wend no
signicant interactions with perceptions of an open or a blame EMC in
line with Hypothesis 2c and 3c. Yet, the model explains a high percentage
of variance, and adding the interaction term between negative emotions
and an open EMC signicantly increased the variance explained of the
model (ΔR
2
Open = 0.05, p < 0.05). Wetherefore performed additional
post-hoc analyses to explore Hypotheses 1c, 2c, and 2c further. Please
refer to section 4.3.1 below.
4.2.4. Time pressure, open/blame EMC, and error
learning (H1d, H2d, H3d)
Finally, Table2, Model 4 reports our ndings on Hypotheses 1d, 2d,
and 3d with respect to time pressure, open/blame EMC and error learning.
Overall, the model is signicant (F(10,83) = 2.81, p < 0.001), explaining
23% of the variance in error learning. Contrary to Hypothesis 1d, wedo
not nd a signicant direct relationship between time pressure and error
learning. Regarding Hypothesis 2d, wend a signicant interaction
between time pressure and an open EMC (B = 0.41, p < 0.10; SE = 0.24).
Figure2, Panel B illustrates this interaction through three conditional
eects, showing that more error learning was reported with more time
pressure when participants experienced high (M +1SD; B
high
 = 0.380,
p < 0.01) and medium open EMC (M; B
medium
 = 0.234, p < 0.05). At the
same time, reported error learning under high and medium open EMC is
higher for auditors experiencing higher time pressure compared to lower
time pressure. Open EMC perceptions therefore did not mitigate the
negative relationship between time pressure and error learning, but rather
fostered learning from errors specically under conditions of high time
pressure. Finally, we do not nd a signicant interaction with blame
EMC. Hence, wend no support for Hypothesis 3d.
4.3. Additional analyses
Following up on our limited ndings regarding negative emotions
and time pressure, weperformed additional post-hoc analyses for the
relationships posited in Hypotheses 1c/2c/3c and 1d.
4.3.1. Post-hoc analysis of the relationship
between negative emotions, open/blame EMC,
and error learning
We found that Model 3in Table2 explained a high percentage of the
variance in error learning (48%) in the absence of signicant coecients.
At the same time, the coecient representing the interaction between
negative emotions and open EMC was insignicant, while it contributed
signicantly to the model’s variance explained. To explore these ndings
further, weperformed two post-hoc analyses. First, to test whether our
non-signicant results are driven by the correlations between error
learning, open EMC, and negative emotions (see Table 1),
weorthogonalized these three variables and reran Model 3 (Little etal.,
2006). Orthogonalizing variables means setting the correlations between
variables to zero, so that in an interaction model, entering the interaction
term does not aect the partial regression coecients of the main
eects. is approach has been associated with more stable regression
ndings with correlated variables, especially in interaction models
(Little etal., 2006). Table3 reports the original OLS model (Model 1),
alongside the orthogonalized one (Model 2). Model 2 is just as signicant
as Model 1 (F(10,44) = 4.61, p < 0.001), explaining 27% of the variance.
We nd signicant coecients for negative emotions (B = −0.13,
p < 0.05, SE = 0.05). In line with Hypothesis 1c, we nd participants
reported less error learning with more intense negative emotions. At the
same time, the interaction terms are insignicant, just as in Model 1.
Yet, given the signicant amount of variance explained that remains
even aer orthogonalizing our variables, weran the analysis separately for
an open and a blame EMC as moderators, respectively. By splitting the
analysis, weidentify distinct eects of both EMC facets that appear to
overlap when included at the same time (see Table3, Models 3 and 4).
First, weconsider the interaction of negative emotions with open EMC
with regard to error learning. is model is signicant (F(8,85) = 2.29,
p < 0.05, R
2
= 0.43) and wend a signicant interaction eect (B = 0.34,
p < 0.10; SE = 0.20, ΔR
2
Open = 0.12, p < 0.10; s ee Table 3, Model 3).
Figure3, Panel A graphically represents this interaction through three
A
B
FIGURE2
Moderating role of open and blame error management climate (EMC)
in the relationship between error consequences, time pressure, and
error learning. (A) Consequences × blame EMC. (B) Time pressure × open
EMC caption: N = 141; conditional eects are illustrated using the pick-
a-point method, where continuous moderators are binned at M − 1SD,
Mean, M + 1SD (Aiken and West, 1991; Hayes, 2018). Interactions are
calculated separately for open EMC and blame EMC. Wereport
unstandardized coecients and p-values. Significance is indicated as
+ = p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed).

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 10 frontiersin.org
TABLE3 Additional analyses of the emotions—error learning relationship.
Model 1 Model 2 Model 3 Model 4
BSE BSE BSE BSE
Variable
Gender 0.08 0.21 0.08 0.10 0.10 0.19 0.06 0.22
Certication 0.21 0.30 −0.14 0.13 −0.25 0.28 −0.30 0.35
Rank 0.13 0.12 0.10+0.06 0.14 0.10 0.12 0.12
Timing 0.00 0.01 0.00 0.00 0.00 0.01 0.00 0.01
Firm type 0.00 0.01 0.20 0.16 0.03 0.27 0.00 0.26
Emotions −0.63 1.20 −0.13** 0.05 −1.11 0.79 0.76+0.39
Open EMC −1.14 1.42 0.16*** 0.05 −1.81 1.18
Blame EMC 0.64 1.15 0.06 0.05 1.33+0.73
Emotions × open 0.25 0.23 0.05 0.04 0.34+0.20
Emotions × blame −0.07 0.19 0.00 0.04 −0.20+0.12
Model
F-statistic 4.04*** 4.61*** 2.29*1.68
R20.48 0.27 0.43 0.34
ΔR2 open 0.05*0.01 0.12+
ΔR2 blame 0.01 0.00 0.08+
N = 141. Table reports unstandardized regression coecients and robust standard errors (Davidson-MacKinnon; Hayes and Cai, 2007) for Model 1, 3, and 4. Model 2 reports the results for
orthogonalized error emotions, open/blame EMC, and error learning [in line with Little etal., 2006]. Signicance is indicated as +p < 0.10, *p < 0.05, **p < 0.01, and ***p < 0.001 (2-tailed).
conditional eects: more error learning occurred with more intense
negative emotions the more open the EMC was perceived to be. Regarding
Hypothesis 2c, wend signicant slopes for high (M+ 1SD, Bhigh=0.356,
p < 0.001) and medium (M; Bmedium=0.187, p < 0.001) open EMC scores,
suggesting that a high or medium open EMC increases reported learning
from errors even given strong negative emotions. e same analyses
performed for blame EMC perceptions reveal a signicant interaction
(B = −0.20, p < 0.10; SE = 0.12, ΔR
2
Blame = 0.08, p < 0.10; s ee Table 3,
Model 4). is model explains signicant variance, but t is suboptimal,
with a signicant percentage of variance explained (F(8,85) = 1.68,
p > 0.10; R2 = 0.34). Exploring this interaction visually (Figure3, Panel B),
wend a signicant slope for low blame EMC (M–1SD; B
low
 = 0.279,
p < 0.05), showing more learning takes place with more intense negative
emotions; at the same time, auditors report the same amount of error
learning from weaker and stronger negative emotions with medium and
high blame EMC, which is contrary to Hypothesis 3c.
4.3.2. Post-hoc analysis of the relationship
between time pressure and error learning
Following up on the insignificant main effect of time pressure
on error learning (Hypothesis 1d), weconducted a post-hoc analysis
to explore a potential non-linear relationship. A series of extant
studies on time pressure have found that medium levels of time
pressure can bepositively associated with a range of outcomes,
such as creativity, learning, or performance, for example, by
creating a sense of urgency or motivating action (e.g., Spilker, 1995;
Baer and Oldham, 2006; Prem etal., 2017; Ryari etal., 2021). At the
same time, higher levels of time pressure can impede these
outcomes, for example, by shifting priorities, or by occupying
working memory through creating stress. Wetherefore explore
whether the relationship between time pressure and error learning
in our sample is also characterized by an inverted-U shaped
function. Lind and Mehlum (2010) propose a methodology for
testing non-linear relationships not through regression, but by
comparing slopes on either side of the apex. With this approach,
two points are selected on the curve, representing a 90% Fieller
interval of the apex, comparing whether the slope of the lower
point is positive and significantly different from the negative slope
of the upper point. Using the ‘utest’ command in Stata yields a
t-value of 1.51, with a one-sided value of p of 0.03. Figure4 shows
that the relationship between time pressure and error learning is
indeed characterized by an inverted-U shape, suggesting that a low
level of time pressure can bebeneficial for reported learning from
errors while higher levels of time pressure have an adverse effect,
as predicted in H1d.
5. Discussion
In this study, weexplored four work conditions that may impede
learning from errors during professionals’ work. In the context of
nancial auditing, weexplored whether professionals’ perceived EMC
moderates the negative relationship of small error consequences,
routine-type errors, strong negative emotions, and high time pressure
with error learning. Our analyses of experiential questionnaire data
from 141 practicing auditors resulted in three key ndings. First,
perceptions of an open EMC positively moderate the relationship
between strong negative emotions, high time pressure, and error
learning. ese results contribute to extant research showing a positive
direct relationship between open EMC and error learning (i.e., Grohnert
etal., 2019; Smeets etal., 2022), and to studies relating emotions (Ilgen
and Davis, 2000; Heimbeck etal., 2003; Zhao etal., 2019) and time
pressure (Kruglanski etal., 1993; Zhao and Olivera, 2006; Homsma
et al., 2009) to error learning. Second, we also found signicant

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 11 frontiersin.org
moderation by blame EMC perceptions—albeit not always in line with
our predictions. Regarding negative emotions, auditors engaged in more
learning from errors accompanied by strong negative emotions when
they perceived themselves to work in an EMC with low blame
characteristics. In the absence of prior research on this relationship,
wend that an absence of blame values and beliefs can contribute to
error learning from negative emotions; a strong open EMC is not
required (van Dyck etal., 2005; Rausch etal., 2017). Regarding small
error consequences, wefound that more error learning takes place from
smaller errors when auditors perceive themselves to work in a blame
EMC. is result is in contrast to our prediction and the nding by
Stefaniak and Robertson (2010). Further analyses into this unexpected
benet of a blame EMC showed that participants in our sample reported
blame EMC perceptions around the mid-point of the scale. Weargue
that some blame-oriented values and beliefs may create a sense of
urgency to learn from smaller errors to prevent their recurrence in the
future. For example, Smeets et al. (2022) found that auditors who
experienced more error strain (which correlated negatively with a
supportive climate for learning from error) were more likely to reect
on their errors in order to learn, in line with arguments by Zhao etal.
(2019). Weconclude that some blame-oriented values and beliefs within
an organization may not be as problematic as previously assumed
(Heimbeck etal., 2003; Lei etal., 2016; Metcalfe, 2017; Smeets etal.,
2022) and may even carry some potential benets for error learning.
Finally, in post-hoc analyses, wefound that time pressure and error
learning are related through an inverted-U-shaped function, showing
that medium levels of time pressure are associated with more error
learning than lower or higher time pressure. is nding is in line with
prior research linking time pressure to performance, creativity, and
learning in general (e.g., Spilker, 1995; Baer and Oldham, 2006; Prem
etal., 2017; Ryari etal., 2021), but is novel in the context of learning
from errors, extending earlier ndings and conceptualizations.
5.1. Theoretical contribution
e current paper contributes to extant research on learning from
errors and error management in four principal ways. First, weadd to the
scarce research on factors that inhibit learning (e.g., Bligh etal., 2018;
Grohnert etal., 2019; Zhao etal., 2022) by adding insights on conditions
that relate to the error (such as small consequences, routine-type errors,
strong negative emotions, and high time pressure) and that may inhibit
error learning, as well as to perceptions of the work environment in
which error learning takes place (perceived blame EMC). Exploring
factors that hinder as well as foster learning from errors can aord a
more complete understanding of the learning process (e.g., Lei etal.,
2016; Tulis et al., 2016). Second, in this study weexplored open and
blame EMC perceptions simultaneously as separate concepts, in line
with van Dyck etal. (2005). Prior research has predominantly studied
the open facet, implicitly inferring information about the blame facet
(van Dyck etal., 2005; Klamar etal., 2022; Matthews etal., 2022). Our
results show that there is a limited correlation between perceptions of
an open and blame EMC—professionals can perceive themselves to
work in an open and a blame EMC at the same time. However, the two
facets produce distinct, rather than mirrored ndings. Consequently,
wesuggest that both concepts should bemeasured separately within the
same study, and that inferring eects of a blame EMC based on the open
EMC scale may not beappropriate. ird, this study is one of the rst to
explore the interaction between EMC and other conditions for error
learning. Our ndings add to prior research that has established a direct
positive relationship of an open EMC and (error) learning (i.e., Grohnert
etal., 2019; Smeets etal., 2022), showing that values and beliefs around
learning from error can have a direct as well as a moderating eect in
the face of conditions that make learning from errors challenging.
Moreover, weadd to the limited prior research on these interactions (i.e.,
Stefaniak and Robertson, 2010; Matthews etal., 2022; Zhao etal., 2022).
We therefore propose taking into account perceptions of the work
environment as a moderator in studies on factors that drive and/or
inhibit error learning. Finally, following prior research that
conceptualizes the relationship between time pressure and performance
as non-linear (e.g., Spilker, 1995; Baer and Oldham, 2006; Prem etal.,
2017; Ryari etal., 2021), wepropose that conceptualizations of error
learning processes take into account potential non-linear relationships
in which both inhibiting conditions and fostering factors are not related
to error learning in a straightforward manner. On the one hand, this will
add nuance to our insights on error learning; on the other hand, it can
inform practice on the degree to which specic mechanisms need to
beabsent/present for learning to take place.
A
B
FIGURE3
Additional analyses negative emotions and error learning.
(A) Emotions × open EMC. (B) Emotions × blame EMC. N = 141;
conditional eects are illustrated using the pick-a-point method,
where continuous moderators are binned at M − 1SD, Mean, M + 1SD
(Aiken and West, 1991; Hayes, 2018). Interactions are calculated
separately for open EMC and blame EMC. Wereport unstandardized
coecients and p-values. Significance is indicated as + = p < 0.10,
*p < 0.05, **p < 0.01, ***p < 0.001 (2-tailed).

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 12 frontiersin.org
FIGURE4
Non-linear relationship between time pressure and error learning. N = 141; non-linear relationship tested using Lind and Mehlum's (2010) approach of
calculating a 90% Fieller interval around the apex of the function, comparing sign and strength of the slopes at the lower and upper boundaries with a
t-test.
5.2. Practical implications
Following our insights on how professionals’ perceptions of their
rm’s EMC interact with the four inhibitors in relation to error learning,
wederive implications for organizations, leaders and professionals at all
levels. First, organizations may benet from more explicitly
communicating the value of all errors, including those with small
consequences to their members, by emphasizing their learning potential,
especially given that errors with no or non-severe consequences are
fairly common in the workplace. Interestingly, it appears that some
blame-based values and beliefs diminish this particular barrier to
learning, possibly due to heightened alertness on the part of the error-
maker. While weare cautious in recommending the maintenance of a
blame climate based on this nding, it does imply that there is some
merit in communicating error repercussions to organizational members.
Indeed, we also nd that auditors reported more error learning
accompanying stronger negative emotions when they perceived
themselves to work in an open EMC, implying that organizations should
seek to strike a balance between components of both open and blame
EMCs. Finally, the auditing profession suers from intense levels of time
pressure due to the cyclical nature of the work and tight deadline
pressures—as is also the case in other elds. As a result, our nding that
excessive time pressure results in lower levels of error learning is
particularly concerning. Since time pressure is an inherent feature of
many workplaces, organizations must nd ways to cope with the
resulting threats, such as reduced learning from error. Our results point
to the benets of an open EMC in mitigating this problem.
More generally, following Salancik and Pfeer’s (1978) social
information-processing theory, organizations can only indirectly inuence
professionals’ individual perceptions. Prior research has shown that
auditors’ perceptions of a learning from error climate are driven
predominantly by leaders’ behaviors, rather than by ocial
communications and formal structures (Grohnert etal., 2019; Smeets etal.,
2021). According to Schein and Schein (2017), leaders inuence what is
valued and rewarded at work by the way they allocate attention, time,
monetary resources, and praise, how they choose to select, mentor, and
promote professionals, and how they react to critical incidents (Schein and
Schein, 2017). Creating an open EMC then centers on being a role model
for learning from one’s own errors, providing opportunities for reporting
errors rather than punishing subordinates, listening and assisting in the
analysis and mitigation of future errors, and sharing knowledge derived
from errors with others (van Dyck etal., 2005; Putz etal., 2012; Grohnert
etal., 2019; Gold etal., 2022; Smeets etal., 2022). is requires that leaders
have clear expectations for managing errors in their teams, as well as
holding team members accountable for creating an open EMC (e.g.,
Edmondson, 2004; Lupton and Warren, 2018). According to Schein and
Schein (2017), the eectiveness of these leader behaviors will depend on
whether or not their underlying values and beliefs are also anchored in an
organization’s structures, systems, and routines, for example, in selection
and promotion decisions (Schein and Schein, 2017). Both mechanisms
need to bealigned for the creation and maintenance of a coherent and
eective EMC. Additionally, organizations need not strive for a ‘perfect’
EMC—the presence of some blame values in a rm may not beproblematic
for eective error learning—implementation need not beperfect, but ‘good
enough’ to foster learning from smaller errors, with stronger negative
emotions and when dealing with time pressure.
5.3. Limitations and future research
e results and conclusions presented in this paper should
beinterpreted in light of the following limitations. First, by using an
experiential questionnaire, we relied on auditors’ self-reported
perceptions. While wedid not nd evidence of common method bias in
participants’ responses to our instrument, future research can further
address this limitation by combining multiple information sources, such
as reports by several team members on the same error, or combining
perceptions by leaders and team members around a specic critical
incident (Flanagan, 1954). Second, our data were collected in a single
professional setting, namely, Dutch audit rms. Auditing is a highly
relevant setting for research on error learning, and the focused approach
limited noise in the data due to standardized certication and

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 13 frontiersin.org
professional development, detailed national and international
regulation, and high levels of proceduralization. However, this focus also
leads to limitations in terms of generalizability. Consequently, future
research is needed to establish whether the relationships found in this
study translate to other contexts. Specically, wepropose future research
to explore both open and blame EMC in the same study. In selecting
other contexts, our literature review suggests focusing both on
knowledge-intensive elds with team-based work, and on settings with
dierent levels of complexity and time pressure (Matthews etal., 2022;
Smeets etal., 2022; Zhao etal., 2022). ird, our sample does not enable
us to conduct a multilevel analysis in which perceptions of their rm’s
climate are aggregated across participants, in line with common
measurements of EMC (van Dyck etal., 2005). In the current study,
weconsequently relied on individual perceptions in relation to their
behaviors, following prior studies on learning climate (i.e., Park and
Rothwell, 2009; Eldor and Harpaz, 2016; Grohnert et al., 2019).
We suggest that future research designs, where possible, include
members across a signicant number of organizations and/or
organizational units, allowing the exploration of the nested nature of
EMC at the group level versus behaviors at the individual level, such as
error learning. Finally, weasked participants to recall errors from their
own experience. Recall based on the experiential questionnaire method
may not becomplete or fully representative of the event (Mahajan, 2010;
Rausch et al., 2017). Weadded time between event and recall as a
covariate, which was only signicantly related to other covariates, not to
model variables. Wenote, however, that more experienced and female
auditors reported errors that were further in the past; this might bea
starting point for future research to explore values, beliefs, and
assumptions as a driver behind expectations about who is ‘permitted’ to
make mistakes, providing a more nuanced picture of the role an EMC
plays in fostering learning from errors for dierent groups of employees.
Data availability statement
e datasets presented in this study can be found in online
repositories. e names of the repository/repositories and accession
number(s) can befound at: https://doi.org/10.34894/UZOLBE, via e
Dataverse Project.
Ethics statement
e studies involving human participants were reviewed and
approved by SBE Ethical Review Board, Vrije Universiteit Amsterdam.
e patients/participants provided their written informed consent to
participate in this study.
Author contributions
OM, TG, and AG contributed to the conception and design of
this study, all authors were involved in the data collection process.
OM and TG organized the dataset and performed the statistical
analyses. OM wrote the rst dra of the manuscript, TG and AG
wrote the current version of the manuscript, and engaged in
supervision of OM. All authors contributed to the article and
approved the submitted version.
Acknowledgments
We would like to thank the participating firms and all
participating auditors for their time and input to this study. Wealso
thank Prof. Philip Wallage for his expertise and inputs in the data
coding process, as well as Arnie Wright, Tjibbe Bosman, Tom Groot,
Ann Vanstraelen, Kris Hardies, Justin Leiby, and Ulfert Gronewold
for comments on earlier versions of this paper. Wewould also like
to thank the organizers, reviewers and participants of the 2021
European Accounting Association Annual Congress, and the 2022
European Association for Research on Learning and Instruction
Special Interest Group14 Learning and Professional Development
Bi-Annual Conference.
Conflict of interest
e authors declare that the research was conducted in the absence
of any commercial or nancial relationships that could beconstrued as
a potential conict of interest.
Publisher’s note
All claims expressed in this article are solely those of the authors
and do not necessarily represent those of their affiliated
organizations, or those of the publisher, the editors and the
reviewers. Any product that may be evaluated in this article, or
claim that may be made by its manufacturer, is not guaranteed or
endorsed by the publisher.
References
Aiken, L. S., and West, S. G. (1991). Multiple Regression: Testing and Interpreting
Interactions. Newbury Park, CA: Sage Publications Inc.
Baer, M., and Oldham, G. R. (2006). e curvilinear relation between experienced
creative time pressure and creativity: moderating eects of openness to experience and
support for creativity. J. Appl. Psychol. 91, 963–970. doi: 10.1037/0021-9010.91.4.963
Baumard, P., and Starbuck, W. H. (2005). Learning from failures: why it may not happen.
Long Range Plan. 38, 281–298. doi: 10.1016/j.lrp.2005.03.004
Bligh, M. C., Kohles, J. C., and Yan, Q. (2018). Leading and learning to change: the role
of leadership style and mindset in error learning and organizational change. J. Chang.
Manag. 18, 116–141. doi: 10.1080/14697017.2018.1446693
Cannon, M. D., and Edmondson, A. C. (2005). Failing to learn and learning to fail
(intelligently): how great organizations put failure to work to innovate and improve. Long
Range Plan. 38, 299–319. doi: 10.1016/j.lrp.2005.04.005
Carmeli, A., and Gittell, J. H. (2009). High-quality relationships, psychological safety,
and learning from failures in work organizations. J. Organ. Behav. 30, 709–729. doi:
10.1002/job
Choo, F. (1995). Auditors’ judgment performance under stress: a test of the predicted
relationship by three theoretical models. J. Account. Audit. Financ. 10, 611–641. doi:
10.1177/0148558X9501000311
DeZoort, F. T., and Lord, A. T. (1997). A review and synthesis of pressure eects research
in accounting. J. Account. Lit. 16, 28–85.
Downey, D. H., and Bedard, J. C. (2019). Coordination and communication challenges
in global group audits. Audit. A J. Pract. eory 38, 123–147. doi: 10.2308/ajpt-52016
Edmondson, A. C. (2004). Learning from mistakes is easier said than done–group and
organizational inuences on the detection and correction of human error. J. Appl. Behav.
Sci. 40, 66–90. doi: 10.1177/0021886304263849

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 14 frontiersin.org
Eldor, L., and Harpaz, I. (2016). A process model of employee engagement: the learning
climate and its relationship with extra-role performance behaviors. J. Organ. Behav. 37,
213–235. doi: 10.1002/job
Ellis, S., and Davidi, I. (2005). Aer-event reviews: drawing lessons from successful and
failed experience. J. Appl. Psychol. 90, 857–871. doi: 10.1037/0021-9010.90.5.857
Ely, J. W., Graber, M. L., and Croskerry, P. (2011). Checklists to reduce diagnostic errors.
Acad. Med. 86, 307–313. doi: 10.1097/ACM.0b013e31820824cd
Embrey, D. (2005). Understanding human behaviour and error. e skill, rule and
knowledge based classication. Available at: http://zonecours.hec.ca/documents/
A2007-1-1399575.UnderstandingHumanBehaviourandError.pdf (Accessed August 15, 2022).
EY (2018). Global Review 2018: How Do We Create Value and Build Trust in is
Transformative Age? London, UK: EY.
Flanagan, J. C. (1954). e critical incident technique. Psychol. Bull. 51, 327–358. doi:
10.1037/h0061470 (Accessed August 21, 2022).
Ford, J. K., Schmitt, N., Schechtman, S. L., Hults, B. M., and Doherty, M. L. (1989).
Process tracing methods: contributions, problems, and neglected research questions.
Organ. Behav. Hum. Decis. Process. 43, 75–117. doi: 10.1016/0749-5978(89)90059-9
FRC (2021). Audit Firm Culture: Challenge Trust Transformation. Available at: https://
www.frc.org.uk/getattachment/dba4544f-6d28-4185-b987-14e4a1122354/Culture-
Conference-Event-Summary_August-2021.pdf (Accessed August 15, 2022).
Frese, M., and Keith, N. (2015). Action errors, error management, and learning in
organizations. Annu. Rev. Psychol. 66, 661–687. doi: 10.1146/annurev-psych-010814-
015205
Fruhen, L. S., and Keith, N. (2014). Team cohesion and error culture in risky work
environments. Saf. Sci. 65, 20–27. doi: 10.1016/j.ssci.2013.12.011
Glover, S. M. (1997). e inuence of time pressure and accountability on auditors’
processing of nondiagnostic information. J. Account. Res. 35:213. doi: 10.2307/2491361
Gold, A., Detzen, D., Mourik, O., Wallage, P., and Wright, A. (2022). Walking the talk?
Managing errors in the audit profession. Contemp. Account. Res. 39, 2696–2729. doi:
10.1111/1911-3846.12794
Gold, A., Gronewold, U., and Salterio, S. E. (2014). Error management in audit rms:
error climate, type, and originator. Account. Rev. 89, 303–330. doi: 10.2308/accr-50592
Gorini, A., Miglioretti, M., and Pravettoni, G. (2012). A new perspective on blame culture:
an experimental study. J. Eval. Clin. Pract. 18, 671–675. doi: 10.1111/j.1365-2753.2012.01831.x
Grohnert, T., Meuwissen, R. H. G. G., and Gijselaers, W. H. (2017). Valuing errors for
learning: espouse or enact? J. Work. Learn. 29, 394–408. doi: 10.1108/JWL-11-2016-0102
Grohnert, T., Meuwissen, R. H. G., and Gijselaers, W. H. (2019). Enabling young
professionals to learn from errors–the role of a supportive learning climate in crossing help
network boundaries. Vocat. Learn. 12, 217–243. doi: 10.1007/s12186-018-9206-2
Gronewold, U., and Donle, M. (2011). Organizational error climate and auditors’
predispositions toward handling errors. Behav. Res. Account. 23, 69–92. doi: 10.2308/
bria-10061
Gronewold, U., Gold, A., and Salterio, S. E. (2013). Reporting self-made errors: the
impact of organizational error-management climate and error type. J. Bus. Ethics 117,
189–208. doi: 10.1007/s10551-012-1500-6
Hagen, J. (2013). Confronting Mistakes-lessons From the Aviation Industry When Dealing
With Error. London: Palgrave Macmillan UK.
Hardies, K., Breesch, D., and Branson, J. (2011). Male and female auditors’
overcondence. Manag. Audit . J. 27, 105–118. doi: 10.1108/02686901211186126
Hayes, A. F. (2018). Introduction to Mediation, Moderation, and Conditional Process
Analysi s. 2nd. New York, NY: e Guilford Press.
Hayes, A. F., and Cai, L. (2007). Using heteroskedasticity-consistent standard error
estimators in OLS regression: an introduction and soware implementation. Behav. Res.
Methods 39, 709–722. doi: 10.3758/BF03192961
Heimbeck, D., Frese, M., Sonnentag, S., and Keith, N. (2003). Integrating errors into the
training process: the function of error management instructions and the role of goal
orientation. Pers. Psychol. 56, 333–361. doi: 10.1111/j.1744-6570.2003.tb00153.x
Hökkä, P., Vähäsantanen, K., and Paloniemi, S. (2020). Emotions in learning at work: a
literature review. Vocat. Learn. 13, 1–25. doi: 10.1007/s12186-019-09226-z
Homsma, G. J., van Dyck, C., De Gilder, D., Koopman, P. L., and Elfring, T. (2009).
Learning from error: the inuence of error incident characteristics. J. Bus. Res. 62, 115–122.
doi: 10.1016/j.jbusres.2007.12.003
ICAEW (2016). Improving Audit Quality Using Root Cause Analysis-What, WHY, How,
Who and When?London: ICAEW.
Ilgen, D. R., and Davis, C. A. (2000). Bearing bad news: reactions to negative
performance feedback. Appl. Psychol. 49, 550–565. doi: 10.1111/1464-0597.00031
Kanfer, R., and Ackerman, P. L. (1989). Motivation and cognitive abilities: an integrative/
aptitude-treatment interaction approach to skill acquisition. J. Appl. Psychol. 74, 657–690.
doi: 10.1037/0021-9010.74.4.657
Keith, N., and Frese, M. (2011). “Enhancing rm performance and innovativeness through
error management culture” in e Handbook of Organizational Culture and Climate. eds. N.
M. Ashkanasy, C. P. M. Wilderom and M. F. Peterson (Los Angeles, CA: SAGE), 137–157.
Kelley, T., and Margheim, L. (1990). e impact of time budget pressure, personality, and
leadership variables on dysfunctional auditor behavior. Audit. A J. Pract. eory 9, 21–42.
Klamar, A., Horvath, D., Keith, N., and Frese, M. (2022). Inducing error management
culture—evidence from experimental team studies. Front. Psychol. 12, 1–18. doi: 10.3389/
fpsyg.2021.716915
KPMG (2016). e Why of the Why-root Cause Analysis Nourishes the Learning
Organisation. Amsterdam: KPMG.
KPMG (2018). Transparency Report. KPMG: Geneva.
Kruglanski, A. W., Webster, D. M., and Klem, A. (1993). Motivated resistance and
openness to persuasion in the presence or absence of prior information. J. Pers. Soc.
Psychol. 65, 861–876. doi: 10.1037/0022-3514.65.5.861
Lei, Z., Naveh, E., and Novikov, Z. (2016). Errors in organizations: an integrative review
via level of analysis, temporal dynamism, and priority lenses. J. Manage. 42, 1315–1343.
doi: 10.1177/0149206316633745
Levitt, B., and March, J. G. (1988). Organizational learning. Annu. Rev. Sociol. 14,
319–338. doi: 10.1146/annurev.so.14.080188.001535
Lind, J. T., and Mehlum, H. (2010). With or without U? e appropriate test for a U-shaped
relationship. Oxf. Bull. Econ. Stat. 72, 109–118. doi: 10.1111/j.1468-0084.2009.00569.x
Little, T. D., Bovaird, J. A., and Widaman, K. F. (2006). On the merits of orthogonalizing
powered and product terms: implications for modeling interactions among latent variables.
Struct. Equ. Model. 13, 497–519. doi: 10.1207/s15328007sem1304_1
Lupton, B., and Warren, R. (2018). Managing without blame? Insights from the
philosophy of blame. J. Bus. Ethics 152, 41–52. doi: 10.1007/s10551-016-3276-6
Mahajan, R. P. (2010). Critical incident reporting and learning. Br. J. Anaesth. 105, 69–75.
doi: 10.1093/bja/aeq133
Margheim, L., Tim, K., and Pattison, D. (2005). An analysis of auditor time deadline
pressure versus time budget. J. Appl. Bus. Res. 21, 23–36. doi: 10.19030/jabr.v21i1.1497
Matthews, J., Love, P. E. D., Ika, L. A., and Fang, W. (2022). Error aversion or
management? Exploring the impact of culture at the sharp-end of production in a mega-
project. Dev. Built Environ. 10, 100074–100010. doi: 10.1016/j.dibe.2022.100074
Metcalfe, J. (2017). Learning from errors. Annu. Rev. Psychol. 68, 6.1–6.26. doi: 10.2307/20493670
Mikkelsen, A., Saksvik, P. Ø., and Ursin, H. (1998). Job stress and organizational learning
climate. Int. J. Stress. Manag. 5, 197–209. doi: 10.1023/A:1022965727976
Park, Y., and Rothwell, W. J. (2009). e eects of organizational learning climate, career-
enhancing strategy, and work orientation on the protean career. Hum. Resour. Dev. Int. 12,
387–405. doi: 10.1080/13678860903135771
Pierce, B., and Sweeney, B. (2004). Cost-quality conict in audit rms: an empirical
investigation. Eur. Account. Rev. 13, 415–441. doi: 10.1080/0963818042000216794
Prem, R., Ohly, S., Kubicek, B., and Korunka, C. (2017). riving on challenge stressors?
Exploring time pressure and learning demands as antecedents of thriving at work. J. O rgan.
Behav. 38, 108–123. doi: 10.1002/job.2115
Putz, D., Schilling, J., Kluge, A., and Stangenberg, C. (2012). Measuring organizational
learning from errors: development and validation of an integrated model and questionnaire.
Manag. Learn. 44, 511–536. doi: 10.1177/1350507612444391
PwC (2018). Transparency Report. PwC: New York, NY.
Rasmussen, J. (1986). Information Processing and Human Computer Interaction: An
Approach to Cognitive Engineering. New York, NY: North-Holland.
Rausch, A., Seifried, J., and Harteis, C. (2017). Emotions, coping and learning in error
situations in the workplace. J. Work. Learn. 29, 374–393. doi: 10.1108/JWL-01-2017-0004
Reason, J. (1990). Human Error. New York, NY: Cambridge University Press.
Reio, T. G., and Cal lahan, J. L. (2004). Aect, curiosity, and socialization-related learning:
a path analysis of antecedents to job performance. J. Bus. Psychol. 19, 3–22. doi: 10.1023/B:
JOBU.0000040269.72795.ce
Rotteveel, M. (2022). AFM Verscherpt Controle op Accountants, er is te Weinig Oog
voor Fraude. Financ. Dagbl. Available at: https://fd.nl/bedrijfsleven/1440921/afm-
accountants-zien-fraude-te-vaak-over-het-hoofd (Accessed December 2, 2022).
Ryari, H., Alavi, S., and Wieseke, J. (2021). Drown or blossom? e impact of perceived
chronic time pressure on retail salespeople’s performance and customer–salesperson
relationships. J. Retail. 97, 217–237. doi: 10.1016/j.jretai.2020.05.005
Rybowiak, V., Garst, H., Frese, M., and Batinic, B. (1999). Error orientation
questionnaire (EOQ): reliability, validity, and different language equivalents. J. Organ.
Behav. 20, 527–547.
Salancik, G. R., and Pfeer, J. (1978). A social information processing approach to job
attitudes and task design. Adm. Sci. Q. 23, 224–253. doi: 10.2307/2392563
Schein, E. H., and Schein, P. (2017). Organizational Culture and Leadership. 5th Edn.
Hoboken, NJ: John Wiley & Sons.
Scott, C., and Sutton, R. E. (2009). Emotions and change during professional
development for teachers: a mixed methods study. J. Mix. Methods Res. 3, 151–171. doi:
10.1177/1558689808325770
Seckler, C., Gronewold, U., and Reihlen, M. (2017). An error management perspective
on audit quality: toward a multi-level model. Account. Organ. Soc. 62, 21–42. doi: 10.1016/j.
aos.2017.08.004
Sitkin, S. B. (1992). “Learning through failure: the strategy of small losses” in Research
in Organizational Behavior. eds. B. M. Staw and L. L. Cummings (Greenwich, CT: JAI
Press), 231–266.

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 15 frontiersin.org
Smeets, L., Gijselaers, W. H., Meuwissen, R. H. G., and Grohnert, T. (2021).
Beyond psychological safety–the role of direct supervisor behavior in fostering learning
from errors at the workplace. Vocat. Learn. 14, 533–558. doi: 10.1007/s12186-021-
09272-6
Smeets, L., Gijselaers, W., Meuwissen, R., and Grohnert, T. (2022). Exploring the link
between learning from error climate and professionals’ engagement in social learning
activities aer errors. Balt. J. Manag. 17, 413–428. doi: 10.1108/BJM-07-2021-0278
Spilker, B. C. (1995). e eects of time pressure and knowledge on key word selection
behavior in tax research. Account. Rev. 70, 49–70.
Stefaniak, C., and Robertson, J. C. (2010). When auditors err: how mistake signicance
and superiors’ historical reactions inuence auditors’ likelihood to admit a mistake. Int. J.
Audit. 14, 41–55. doi: 10.1111/j.1099-1123.2009.00402.x
Storbeck, O. (2020). How wirecard misled investors over KPMG’s special audit. Financ.
Time s. Available at: https://www..com/content/63d8a230-ce3c-4088-989f-7741632d89a2
(Accessed October 24, 2022).
Sutclie, A., and Rugg, G. (1998). A taxonomy of error types for failure analysis and risk
assessment. Int. J. Hum. Comput. Interact. 10, 381–405. doi: 10.1207/s15327590
ijhc1004_5
Tulis, M., Steuer, G., and Dresel, M. (2016). Learning from errors: a model of individual
processes. Front. Learn. Res. 4, 12–26. doi: 10.14786/r.v4i2.168
van Dyck, C. (2009). e tragic 1996 Everest expedition: a tale of error culture. Neth. J.
Psychol. 65, 22–34. doi: 10.1007/BF03080124
van Dyck, C., Frese, M., Baer, M., and Sonnentag, S. (2005). Organizational error
management culture and its impact on performance: a two-study replication. J. Appl.
Psychol. 90, 1228–1240. doi: 10.1037/0021-9010.90.6.1228
Vera, D., and Crossan, M. M. (2016). Strategic leadership and organizational learning.
Acad. Manag. J. 28, 1035–1057. doi: 10.1002/job.496
Zhang, J., Patel, V. L., Johnson, T. R., and Shortlie, E. H. (2019). Toward an action based
taxonomy of human errors in medicine. Proc. Twenty-Fourth Annu. Conf. Cogn. Sci. Soc.,
970–975. doi: 10.4324/9781315782379-202
Zhao, B. (2011). Learning from errors: the role of context, emotion, and personality. J.
Organ. Behav. 463, 435–463. doi: 10.1002/job
Zhao, B., Emby, C., and Sieweke, J. (2019). Audit senior modeling fallibility: the eects
of reduced error strain and enhanced error-related self-ecacy on audit juniors’ responses
to self-discovered errors. Behav. Res. Account. 31, 17–30. doi: 10.2308/bria-52471
Zhao, B., Lu, Y., Wang, X., and Pang, L. (2022). Challenge stressors and learning from
failure: the moderating roles of emotional intelligence and error management culture.
Technol. Anal. Strateg. Manag., 1–14. doi: 10.1080/09537325.2022.2028765
Zhao, B., and Olivera, F. (2006). Error reporting in organizations. Acad. Manag. 31,
1012–1030. doi: 10.5465/amr.2006.22528167

van Mourik et al. 10.3389/fpsyg.2023.1033470
Frontiers in Psychology 16 frontiersin.org
Appendix
Open-ended questions in the experiential questionnaire
Questions used for determining error consequence.
– If applicable, please indicate the impact of the described error for the audit itself (e.g., led to adjustments or additions in audit procedures).
– If applicable, please indicate the impact of the described errors for youpersonally in terms of upcoming engagement performance evaluations.
– If applicable, please indicate the impact of the described error for people around youin terms of colleagues (subordinates, peers, and superiors)
and the client.
Questions used for determining error type.
– With your chosen example in mind, please describe the following: (I) the task youwere performing, (II) the error youmade during the task.
– Please explain how the described error was discovered.
– What steps did youundertake to handle the described error?