current perspectives

What every teacher needs to know about clinical

reasoning
Kevin W Eva

CONTEXT One of the core tasks assigned to clinical

teachers is to enable students to sort through a clus- INTRODUCTION
ter of features presented by a patient and accurately
assign a diagnostic label, with the development of an A 43-year-old woman is brought to the Emergency
appropriate treatment strategy being the end goal. Room by her husband at 0200 in the morning
Over the last 30 years there has been considerable because of acute shortness of breath. The dyspnea
debate within the health sciences education literature had occurred suddenly at 1100 pm and had awo-
regarding the model that best describes how expert ken the patient from sleep. She had felt nauseated
clinicians generate diagnostic decisions. and vomited a small amount of bile. She com-
plained of retrosternal chest pain that was worse on
PURPOSE The purpose of this essay is to provide a deep breathing. For several days she had coughed
review of the research literature on clinical reasoning up small amounts of blood. For 4 days she had felt
for frontline clinical teachers. The strengths and unwell and had had a sore throat and sinus
weaknesses of different approaches to clinical rea- congestion that resolved. She complained of hav-
soning will be examined using one of the core divides ing experienced fever and chills on several occa-
between various models (that of analytic (i.e. con- sions in the past few days. The previous night she
scious ⁄ controlled) versus non-analytic (i.e. uncons- had woken with chest tightness, but this had settled
cious ⁄ automatic) reasoning strategies) as an after a short while. Her past history included
orienting framework. bronchitis.1

DISCUSSION Recent work suggests that clinical What is the most likely diagnosis? Analogous to
teachers should stress the importance of both forms determining ‘whodunit’ when reading a mystery
of reasoning, thereby enabling students to marshal story, the diagnostic challenge involves considering
reasoning processes in a flexible and context-specific each piece of available information and determining
manner. Specific implications are drawn from this the most plausible explanation for the illustrated
overview for clinical teachers. pattern. Doing so is not a straightforward task. It
often entails careful observation, appropriate elicita-
KEYWORDS education, medical, undergraduate ⁄ tion of historical information, accurate performance
*methods; clinical competence ⁄ *education; decision of physical manoeuvres, the generation of hypothe-
making; teaching ⁄ methods; review literature. ses, appreciation of the relationship between each
piece of data and each hypothesis, and attempting to
Medical Education 2004; 39: 98–106 confirm ⁄ disconfirm hypotheses through the appro-
doi:10.1111/j.1365-2929.2004.01972.x priate ordering of diagnostic tests. Unlike the reader
of a mystery story, the clinician is often faced with the
added task of determining if, when and how various
Program for Educational Research and Development, Department of
Clinical Epidemiology and Biostatistics, McMaster University,
pieces of information will be collected.
Hamilton, Ontario, Canada
Correspondence: Kevin W Eva, Program for Educational Research and The challenge facing clinical teachers is perhaps even
Development, Department of Clinical Epidemiology and Biostatistics, greater. Not only must clinical teachers be capable of
MDCL 3522, McMaster University, Hamilton, Ontario L8N 3Z5, performing all the tasks listed above, but they must
Canada. Tel: 00 1 905 525 9140 ext. 27241; Fax: 00 1 905 572 7099;
E-mail: evakw@mcmaster.ca also find a way to convey their knowledge and

98
Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

 99

on current understanding of the way in which

clinicians solve diagnostic challenges and the impli-
Overview cations arising from this understanding. Discussion
will focus on instructional techniques for maximising
What is already known on this subject the probability that students will become successful
medical problem solvers and on strategies for accu-
Traditional models of clinical decision making rately assessing whether or not students have in fact
provide insufficient descriptions of the way in developed the required competencies. To begin, we
which clinicians reason; non-analytic processes will undertake a more careful examination of how
(e.g. pattern recognition) play a major role, one might solve the diagnostic problem that opened
but can result in biases at all levels of exper- this article.
tise.

What this paper adds WHAT IS THE MOST LIKELY DIAGNOSIS

This paper orients clinical teachers to the
(AND HOW IS IT DERIVED)?
need to teach students multiple reasoning
One need not look very far to recognise that medical
strategies. It attempts to re-conceptualise the
educators have traditionally focused on what are
construct of expertise to that of an amorphous
known as ‘analytic’ models of clinical reasoning;
entity that might best be defined as flexibility
models that presume a careful analysis of the relation
regarding the ways by which solutions to
between signs and symptoms and diagnoses are the
clinical problems can be derived.
hallmark of clinical expertise. For example, Harrison’s
Principles of Internal Medicine presents shortness of
Suggestions for further research
breath as an indication of both pneumonia and
pulmonary thromboembolism (PTE). In the above
Further study of the optimal way in which
case, the additional features of sore throat, nausea
various reasoning strategies can be conveyed
and vomiting further suggest pneumonia,5 whereas
to medical trainees, contextual factors influ-
the additional feature of coughing up blood suggests
encing the co-ordination of analytic and non-
PTE.6 The implication, in both cases, is that the
analytic processes, and the nature of the
characteristic features are plainly evident and that
flexibility displayed by experts is required.
diagnostic reasoning involves understanding the
relationship between the features detected and the
underlying disorders. Generation of a differential list
of relevant diagnoses and application of an appro-
reasoning strategies to novice diagnosticians to priate diagnostic algorithm then allows each diagno-
nurture each pupil’s own expertise. Over the last half sis to be weighted in terms of its relative probability.6
century it has become clear that the ability to do so is
related to, but distinct from, expertise within the Clinical reasoning models that incorporate Bayes’
content area to be taught.2 The maintenance of theorem or regression analyses best represent this
clinical teaching expertise requires, in part, an form of reasoning.7,8 Briefly, these models assume
understanding of strategies expert clinicians use, that physicians are aware of the a priori probability
often unconsciously, to reason through diagnostic with which a particular diagnosis may present and the
case presentations like that which opened this conditional probability associating each piece of
article.3 Adding to the clinical teacher’s challenge is evidence (e.g. signs, symptoms and diagnostic tests)
the fact that the psychological mechanisms underly- with the diagnosis. The mathematical model com-
ing such reasoning tendencies are not always avail- putes a post-test assessment of the likelihood of each
able to introspection.4 diagnosis under consideration. This process, illustra-
ted in Fig. 1, continues to be promoted by individuals
This article was written to provide a review of the close to the evidence-based medicine movement.9,10
literature on clinical reasoning for frontline clinical While some have argued that the forward flow of
teachers. In doing so, few details will be provided information illustrated in Fig. 1 (i.e. reasoning from
regarding the research methods that led to each the evidence to diagnoses) best captures the essence
conclusion – references will be provided for those of ‘expert’ clinical reasoning,11,12 the posterior
who are interested. Rather, the focus will be placed probabilities presented could just as easily be used to

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

100 current perspectives

Presenting There is, however, another way to solve the problem

Clinical Diagnostic Posterior described above – one that has received increasing
Features Hypotheses Probability amounts of attention over the past 15 years. As an
illustration, attempt to diagnose the following case:
w(A,1)
A Dx1 Pr(Dx1) A 43-year-old woman is brought to the Emergency
Room by her husband at 02 : 00 in the morning
B Dx2 Pr(Dx2) because of acute shortness of breath. The dyspnea
had occurred suddenly at 11 : 00 pm and had
awoken the patient from sleep. She had felt
C Dx3 Pr(Dx3) nauseated and vomited a small amount of bile. She
w(C,3) complained of retrosternal chest pain that was
Figure 1 Analytic processes in clinical reasoning. Unique
worse on deep breathing. For several days she had
diagnoses are indicated by numbers, clinical features by coughed up small amounts of blood. For 4 days
letters. Each feature maintains a unique relationship (i.e. she had felt unwell and had had a sore throat and
weight) with each diagnosis. The magnitude of the weights sinus congestion that resolved. She complained of
is indicated by the size of the arrow. Two (e.g. weight of the having experienced fever and chills on several
relationship between feature A and diagnosis 1) are la- occasions in the past few days. The previous night
belled. The result of combining base rates and feature- she had woken with chest tightness, but this had
based weights is assignment of a conditional probability settled after a short while. Her past history inclu-
(Pr) to each diagnostic hypothesis that takes into account ded bronchitis.1
the cluster of features observed.
Even clinically naı̈ve readers will recognise that a
plausible differential diagnosis for this case includes
pneumonia and pulmonary thromboembolism. Why?
feed back onto the collection and analysis of addi- Because this case has been encountered before (at
tional data and the model would remain an analytic the beginning of this article). When asked, ‘What is
processing model as long as the notion of careful 120 divided by 10?’, most of us can quickly and
analysis is maintained. effortlessly respond ‘12.’ Clinicians are similarly often
in such a position that they need not ‘reason’ at all.
In fact, analytic processes have been used in many Referred to as pattern recognition in some circles,
ways across different models of expertise.13–15 At the more general form of ‘non-analytic reasoning’,
the heart of each approach, however, is the illustrated in Fig. 2, essentially amounts to comparing
fundamental belief that causal rules linking fea-
tures (e.g. signs and symptoms) to categories (e.g.
diagnoses) can be extracted from the world and
Presenting
that the development of expertise in clinical Filter
Clinical Diagnostic
reasoning consists of the development and elab- through prior
Features Hypotheses
oration of rules that become more and more Epidodes
attuned to reality. This view of clinical reasoning
suggests that the educator’s task is to facilitate the A, B, D, F Pr(Dx1)
A
development of such rules. As an example,
Elieson and Papa have shown the pedagogical B
B, D, G, R Pr(Dx2)
benefit of providing students with diagnostic aids C
that explicitly describe the probabilistic relation- D
ships between features and symptoms.16 One need C, F, G, H Pr(Dx3)
not maintain such an extreme view of the
usefulness of explicit probabilities, however, to Figure 2 Non-analytic processes in clinical reasoning.
Unique diagnoses are indicated by numbers, clinical
incorporate the import of analytical reasoning
features by letters. Each patient (represented by the
strategies into one’s clinical teaching – intuitive
rounded rectangles) presents with a cluster of features.
theories of the value of analytical reasoning are This cluster is compared (unconsciously) to examples that
espoused every time a clinical teacher admonishes have been encountered in the past, resulting in a probab-
a student to ‘be objective’ and ‘carefully consider ility being assigned to each hypothesis. The strength of the
all the evidence available before generating diag- match between the current patient and past experiences is
nostic hypotheses’. variable, as indicated by the size of the arrows.

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

 101

the current case to those that have been encountered group trained to use non-analytic reasoning, even
in the past and using these past experiences to make among relative novices.22,32 Non-analytic bases of
judgements regarding the probability that any par- judgement are not inferior to more analytic forms of
ticular case belongs within a particular diagnostic reasoning and clinical teachers should inform their
category.17,18 The example provided here is extreme students that similarity to past instances can serve as a
in that all features were presented in exactly the same useful guide. The potential for ‘grim consequences’
way both times the case was encountered, but, as is a specious argument, given that the final respon-
Fig. 2 shows, it is not necessary for all features to sibility for clinical care typically remains with the
correspond in order for a potential ‘match’ to be clinical teacher, not with novice trainees.
identified.

This form of ‘reasoning’ is hypothesised to occur with THE UNION OF ANALYTIC AND NON-
sufficient automaticity to make it often take place ANALYTIC REASONING STRATEGIES
without conscious awareness. Despite the tendency
we as humans have to offer explanations for our That being said, it does appear to be true that
actions, in reality the sources of our behaviour and excessive reliance on non-analytic approaches to
decisions are often unknown to us.4 Although this clinical reasoning can be a source of diagnostic error.
fact makes it impossible to assume that the responses First impressions, while useful, are often incorrect,
we get are valid when we simply ask clinicians if ⁄ when even among experienced clinicians.24 Contextual
they use pattern recognition, the evidence that factors, such as receipt of a diagnostic suggestion,
clinicians use non-analytic processes in reaching have been shown to decrease both the likelihood that
diagnostic decisions is indisputable.19 For example, features consistent with alternative diagnoses will be
in a series of studies Brooks et al. showed that identified25 and the relative weight assigned to
diagnostic accuracy is higher for dermatological cases features consistent with alternative diagnoses that are
that are similar to cases seen before relative to cases identified.26 Eva has provided evidence that these
that are perceptually quite distinct.20 Furthermore, biases derive from excessive reliance on non-analytic
Hatala et al. reported that even diagnostically irre- processing; simple instruction to explicitly list the
levant features of a case (e.g. being a banker) have an evidence present in a case (i.e. an instruction that can
impact on the diagnosis of subsequently presented be anticipated to promote more analytic processing)
cases in which the irrelevant piece of information is was found to be sufficient to eliminate this type of
similar.21 bias.27 A critical factor, however, was that the analytic
processing should be carried out in close temporal
It has been argued that the ability to use non-analytic relation to performing the actual task of diagnostic
bases of clinical decision making increases with judgement.
expertise and, as a result, the use of pattern recog-
nition should not be advocated among medical Where does this leave the clinical teacher? First, it
students for fear of ‘potentially grim conse- must be recognised that these two forms of
quences’.22 (p 699) At the extreme, it must be the case processing are not mutually exclusive. It is highly
that absolute novices have no past experience on probable that both forms of processing contribute
which to rely and, hence, are unable to utilise to the final decisions reached in all cases (for both
similarity-based reasoning strategies. In reality, how- novices and experts). The impact of similarity, in
ever, it has been shown that the strategy employed by some cases, will be to prompt an analytic consid-
even the most junior medical students is qualitatively eration of the current case that is analogous to
indistinguishable from that employed by experienced analyses that were performed on a similar case in
doctors – both groups generate hypotheses very the past. As a result, the optimal form of clinical
quickly, presumably based in part on non-analytic reasoning should be considered an additive model
reference to past experiences.23 More experienced in which both analytic and non-analytic processes
clinicians are more likely to generate the correct play a role. One such model is illustrated in Fig. 3.
response, however, as would be expected given that In this model, the clinician forms a mental repre-
they have a larger database to refer to. More directly, sentation of the case upon presentation of a patient
whenever the advantage of teaching students to and this mental representation leads to hypothesis
reason in an analytic manner has been explicitly testing, which in most cases will take the form of
compared to the influence of teaching students to history taking, physical examination and the order-
trust in their non-analytic judgements, diagnostic ing of diagnostic tests. Importantly, the direction of
accuracy has been at least as good if not better in the

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

102 current perspectives

Patient Case Hypotheses nostic errors, even among highly experienced clini-
Presents Representation Tested cians.24 To avoid either of these situations, clinical
teachers should promote both forms of reasoning in
combination. Further consideration of the implica-
Non-analytic Interactive Analytic
tions of this view will be outlined after a brief note on
the stability of specific diagnostic strategies.
Figure 3 A combined model of clinical reasoning. The
analytic and non-analytic models from Figs 1 and 2 are
combined in this model. Each type of processing is believed THE ‘STATE’ OF THE UNION
to interact with both the mental representation of the case
being presented and the hypotheses that are raised, but to From the above description of clinical reasoning it
different degrees depending on the context. is easy to infer the need for clinical teachers to
nurture students to become ‘good problem solvers’
or ‘good co-ordinators of analytic and non-analytic
reasoning is illustrated to proceed in both direc-
processing’. Such views of the diagnostic process are
tions; results from hypothesis testing will influence
outdated and inaccurate, because of the robust
the mental representation maintained by the clini-
finding that the successful solution of a particular
cian and the mental representation may have an
clinical problem does not accurately predict the
influence on the way a patient’s problems are
successful solution of another clinical problem,
perceived. The bi-directional flow can be expected
even within an area of specialisation.34,35 Reasoning
to occur in both novices and experts.28 In addition,
ability is not a ‘trait’ that can be assigned to an
it should be noted that while non-analytic process-
individual. Undoubtedly some individuals are better
ing is expected to dominate during the initial
diagnosticians than others, but clinical teachers
phases of considering a new case and analytic
must recognise that the context within which a
processing is expected to play a dominant role in
problem is being addressed (i.e. the ‘state’) has a
hypothesis testing, these two forms of reasoning
major impact on the accuracy of the decisions
should be viewed as being very interactive; rather
reached and the optimal balance between potential
than lying along a continuum, they are instead
reasoning ‘strategies.’ ‘Context’ includes both situ-
complementary contributors to the overall accuracy
ational factors (for example the clinical setting and
of the clinical reasoning process, each influencing
cases that have been present recently) and personal
the other.29–31
factors (for example the experience of the clinician
and current thought and opinion).
Recent work provides practical support for this
model.32 When teaching absolute novices (undergra-
Despite attempts to offer students a uniform curri-
duate psychology students) to diagnose electrocardi-
culum, no two students ever have exactly the same
ograms (ECGs), one group of participants were
experiences. Different students see different cases,
instructed during practice and test phases to trust
reflect upon different aspects of a given case, and
feelings of similarity (a non-analytic basis of clinical
derive different insights from such reflections. Each
reasoning), but to avoid ‘jumping the gun’ by explicitly
of these differences will have an impact upon the way
considering the specific features present on the ECG
an individual student approaches a given case (i.e.
(an analytic strategy). This group showed higher
specific situational factors will influence the ‘reason-
diagnostic accuracy than two other groups in which
ing strategy’ adopted). This state-based conception of
participants received either the non-analytic instruc-
clinical practice is perhaps the most fundamental
tion or the analytic instruction alone. The diagnostic
reason for ensuring that students are provided with
performances of the latter two groups were equivalent.
multiple strategies that might enable them to work
This result replicates and extends past work where
through a clinical problem. In some cases a heavy
combined instruction resulted in greater diagnostic
dose of pattern recognition is most likely to yield the
accuracy than did purely analytic instruction.33 In both
correct solution. In others, a more complete history,
studies it appeared clear that students who were
or application of a diagnostic algorithm, or consid-
instructed to use purely analytic techniques found
eration of the basic science underlying the patho-
themselves awash in a virtual torrent of clinical
physiology might be required. The more tools one
features, making it difficult to reconcile the observed
has in one’s workshop, the more likely it is that one of
pattern with a single diagnostic entity. Furthermore,
the tools will successfully enable completion of the
other work suggests that failure to perform an analytic
task at hand.
confirmation results in premature closure and diag-

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

 103

With this in mind, it becomes unlikely that any single earlier students begin to accumulate a mental
construct will ever fully define the term ‘expert’ in a database of cases, the sooner they will develop a
domain as broad as clinical reasoning. By way of firm foundation on which to allow non-analytic
analogy, consider the development of reading processes to contribute. This idea is not new to
expertise. In performing a component skills analysis medical educators – it is a fundamental principle of
of reading ability, Levy and Hinchley administered a good pedagogy. What is relatively new, however, is
series of 11 reading tests and found that while poor the recognition that a few complex and elaborate
readers were on average worse than good readers, any examples are likely to be suboptimal as effective
individual reader had some strengths and some teaching tools. Context specificity and the need to
weaknesses.36 In fact, 56% of the good readers build up an adequate database from which to reason
exhibited performance deficits on one or more tests by way of analogy demand that many examples be
and 58% of the poor readers showed superior skill on seen, that students be enabled to actively engage in
one or more tests. It is likely, therefore, that expertise the problem solving process, and that the examples
in clinical reasoning should be considered an amor- provide an accurate representation of the range of
phous entity that enables competent clinicians to ways in which specific conditions present.34 This
compensate for case-specific weaknesses. It allows one latter criterion has become increasingly important to
to adapt to the demands of the situation, flexibly consider as the evolution of the health system in
(albeit often unconsciously) utilising the full arma- many parts of the world has lessened the probability
ment available. that students will randomly encounter a large
number of some medical conditions during their
clinic-based learning experiences. As a result, great-
IMPLICATIONS FOR CLINICAL er awareness and creativity on the part of clinical
TEACHERS teachers and curriculum planners is required to
ensure that students receive adequate exposure to
In summary, a great deal of debate has taken place pedagogically useful cases.
within the medical education literature pertaining to
the structure of medical expertise. In the late 1970s, Second, clinical teachers should recognise that the
Elstein et al. presented the hypothetico-deductive traditional 2-stage approach to clinical teaching,
model of clinical reasoning – namely, that when dating back at least as far as Flexner,39 in which
faced with a new case, doctors generate a set of students are expected to master the basic biomedical
hypotheses that they later use to test against the data sciences before proceeding to consideration of clin-
presented.37 Since that time numerous frameworks of ical problems, may be inappropriate. There is evi-
knowledge representation have been developed, but dence to suggest that an understanding of basic
research performed in the last 15 years has called science mechanisms can assist diagnosticians in
into question whether or not any particular frame- generating accurate hypotheses and therefore should
work will prove correct.38 More recently still, evidence remain part of medical training.40,41 It must be
has begun to accumulate that suggests a more recognised, however, that this strategy provides a way
comprehensive approach to clinical teaching, an of reaching the correct diagnosis, not the way.
approach that includes recognition of the benefits of Similarly, simply working on a ward and interacting
both analytic and non-analytic approaches to clinical with a series of patients without additional focus on
reasoning and that can enable students to take the underlying principles of the cases may do
advantage of the best of both worlds.32 Further students a disservice by weakening one of the avenues
awareness of the prevalence of context specificity has by which they might be able to derive solutions to
highlighted the need to provide students with an future cases.
array of strategies that might better position them to
flexibly adapt as the situation demands. The Third, practice with cases should proceed in a way
remaining paragraphs of this article will outline some that mimics the eventual use of the resulting know-
of the implications that arise from this present ledge. Clinicians rarely encounter a novel case in
understanding in an effort to facilitate reflection on which the diagnosis is known. Working through
current pedagogical techniques and stimulate the textbook cases in which one already knows the
development of new approaches. diagnosis as a result of the chapter topic (or the topic
of the lecture) does not enable the student to
First and foremost this review highlights the determine whether or not they would be able to
importance of teaching around examples. The recognise the case if it were to show up on the ward.

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

104 current perspectives

If the patient presentation and case representation time required to perform an evaluation task, the less
outlined in Fig. 3 are fully entwined with a particular opportunity there is to have students complete the
diagnosis, practice with the critical hypothesis testing task multiple times. To ensure information that
phase is lost. In support of this statement, many reliably indicates student ability level is collected,
investigators have shown that ‘mixed practice’ in clinical teachers should continue to utilise tools such
which students see cases of multiple categories mixed as the objective structured clinical examination,
together (as opposed to blocked practice in which clinical reasoning exercise and multiple-choice
students work through a block of cases from one tests.47
diagnostic category before proceeding to the next
block of cases from a different diagnostic category) is
pedagogically optimal.42,43 Acknowledgements: the author would like to thank John
Bligh for providing the motivation to write this article, as
Furthermore, clinical teachers should not rely on well as Lee Brooks, Geoff Norman and Glenn Regehr for
students to make meaningful comparisons across insightful commentary provided during the creation of the
problems spontaneously. Students are much more manuscript.
This material was presented, in part, at the May 2004
likely to successfully reason by way of analogy when
Educational Workshop of the Association of Medical
they have been explicitly instructed to attempt to
School Microbiology and Immunology Chairs.
identify similarities in the underlying concepts of Funding: none.
superficially distinct problems.44 As such, principles Ethical approval: not required.
inherent in novel examples should be related back
to those inherent in past examples whenever poss-
ible. Adding to the benefit of such an educational REFERENCES
strategy is the by-product of the provision of better
information to the clinical teacher regarding where 1 Cunnington JPW, Turnbull JM, Regehr G, Marriott M,
students may be experiencing difficulties. It is well Norman GR. The effect of presentation order in clin-
known that experts have difficulty predicting the ical decision making. Acad Med 1997;72 (Suppl):
errors that others make.45 Providing students with 40–42.
an opportunity to reveal idiosyncratic mistakes 2 Wilkerson L, Irby DM. Strategies for improving teach-
enables clinical teachers to focus teaching efforts in ing practices: a comprehensive approach to faculty
a direction that is most likely to benefit individual development. Acad Med 1998;73:387–96.
students. 3 Irby DM. What clinical teachers in medicine need to
know. Acad Med 1994;69:333–42.
4 Bargh JA, Chartrand TL. The unbearable automaticity
Finally, the flexibility inherent in clinical reasoning
of being. Am Psychol 1999;54:462–79.
and the prevalence of context specificity has very real 5 Levison ME. Pneumonia, including necrotising pul-
implications for clinical teachers’ evaluation of monary infections (lung abscess). In: Braunwald E,
trainees. One should not assume that because a Hauser SL, Fauci AS, Longo DL, Kasper DL, Jameson
student has provided an accurate diagnosis and ⁄ or JL, eds. Harrison’s Principles of Internal Medicine. 15th
management plan, he or she fully understands the edn. New York: McGraw-Hill 2001;1475–85.
physiological mechanisms underlying the process. 6 Goldhaber SZ. Pulmonary thromboembolism. In:
Similarly, even if the student can explain the under- Braunwald E, Hauser SL, Fauci AS, Longo DL, Kasper
lying physiological mechanisms, one should not DL, Jameson JL, eds. Harrison’s Principles of Internal
assume that he or she would provide an accurate Medicine. 15th edn. New York: McGraw-Hill 2001;1508–
diagnosis upon encountering the next case. In 13.
7 Weinstein MD, Fineburg HU, Elstein AS et al. Clinical
domains that are afflicted by context specificity (i.e.
Decision Analysis. Philadelphia: WB Saunders 1980.
all domains), a ‘multiple biopsy’ approach to evalu-
8 Wigton RS. Use of linear models to analyse physicians’
ation is required to accurately assess a student’s decisions. Med Decis Making 1988;8:241–52.
performance. While no one would dream of assessing 9 Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical
knowledge with a single multiple-choice question, Epidemiology: A Basic Science for Clinical Medicine. 2nd
recognition of the need to broadly sample when edn. Toronto: Little Brown 1991.
assessing other characteristics of expertise has been 10 Fletcher R, Fletcher S, Wagner E. Clinical Epidemiology:
less forthcoming. In hindsight it has become clear The Essentials. 2nd edn. Baltimore: Williams & Wilkins
that context specificity is a major contributor to the 1996.
poor psychometric qualities of evaluation exercises 11 Patel VL, Goren GJ. Knowledge-based solution
like the triple jump, patient management problems strategies in medical reasoning. Cognit Sci 1986;10:
91–116.
and long essay and oral examinations.46 The more

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

 105

12 Patel VL, Arocha JF, Kaufman DR. Diagnostic reason- 30 Regehr G, Cline DJ, Norman GR, Brooks LR. Effect of
ing and medical expertise. Psychol Reason Motiv processing strategy on diagnostic skill in dermatology.
1994;31:187–252. Acad Med 1994;69:S34–S36.
13 Genero N, Cantor N. Exemplar prototypes and clinical 31 Kulatunga-Moruzi C, Brooks LR, Norman GR.
diagnosis: toward a cognitive economy. J Soc Clin Psychol Co-ordination of analytic and similarity-based process-
1987;5:59–78. ing strategies and expertise in dermatological diagno-
14 Bordage G. Elaborated knowledge: a key to successful sis. Teach Learn Med 2001;13:110–6.
diagnostic thinking. Acad Med 1994;69:883–5. 32 Ark T, Brooks LR, Eva KW. The best of both worlds:
15 Mandin H, Jones A, Woloschuk W, Harasym P. Helping adoption of a combined (analytic and non-analytic)
students learn to think like experts when solving clin- reasoning strategy improves diagnostic accuracy
ical problems. Acad Med 1997;72:173–9. relative to either strategy in isolation. Presented at the
16 Elieson SW, Papa FJ. The effects of various knowledge annual meeting at the Association of American
formats on diagnostic performance. Acad Med 1994;69 Medical Colleges; Boston, MA; November 2004.
(Suppl):81–3. 33 Norman GR, Brooks LR, Colle CL, Hatala RM. The
17 Brooks LR. Decentralised control in categorisation: the benefit of diagnostic hypotheses in clinical reasoning:
role of prior processing episodes. In: Neisser U, ed. experimental study of an instructional intervention for
Concepts and Conceptual Development: Ecological and forward and backward reasoning. Cognit Instruct
Intellectual Factors in Categorisation. Cambridge: Cam- 2000;17:433–48.
bridge University Press 1987;141–74. 34 Eva KW, Neville AJ, Norman GR. Exploring the
18 Jacoby LL, Baker JG, Brooks LR. The priority of the aetiology of content specificity: factors influencing
specific: episodic effects in picture identification. J Exp analogic transfer and problem solving. Acad Med
Psychol Learn Mem Cogn 1989;15:275–81. 1998;73 (Suppl):1–5.
19 Norman GR, Brooks LR. The non-analytical basis of 35 Eva KW. On the generality of specificity. Med Educ
clinical reasoning. Adv Health Sci Educ 1997;2:173–84. 2003;37:587–8.
20 Brooks LR, Allen SW, Norman GR. Role of specific 36 Levy BA, Hinchley J. Individual and developmental
similarity in a medical diagnostic task. J Exp Psychol Gen differences in the acquisition of reading skills. In: Carr
1991;120:278–87. TH, Levy BA, eds. Reading and its Development: Compo-
21 Hatala RM, Norman GR, Brooks LR. Influence of a nent Skills Approaches. San Diego: Academic Press, Inc.
single example on subsequent electrocardiogram 1990;81–128.
interpretation. Teach Learn Med 1999;11:110–7. 37 Elstein AS, Shulman LS, Sprafka SA. Medical
22 Coderre S, Mandin H, Harasym PH, Fick GH. Diag- Problem Solving: An Analysis of Clinical Reasoning.
nostic reasoning strategies and diagnostic success. Med Cambridge, Massachusetts: Harvard University Press
Educ 2003;37:695–703. 1978.
23 Neufeld VR, Norman GR, Feightner JW, Barrows HS. 38 Custers EJFM, Regehr G, Norman GR. Mental repre-
Clinical problem solving by medical students: a cross- sentations of medical diagnostic knowledge: a review.
sectional and longitudinal analysis. Med Educ Acad Med 1996;71 (Suppl):24–26.
1981;15:315–22. 39 Flexner A. Medical Education in the United States and
24 Eva KW. The ageing physician: changes in cognitive Canada. A Report to the Carnegie Foundation for the
processing and their impact on medical practice. Acad Advancement of Teaching. Boston: Carnegie Foundation
Med 2002;77 (Suppl):1–6. for the Advancement of Teaching 1910.
25 Brooks LR, LeBlanc VR, Norman GR. On the difficulty 40 Schmidt HG, Norman GR, Boshuizen HP. A cognitive
of noticing obvious features in patient appearance. perspective on medical expertise: theory and implica-
Psychol Sci 2000;11:112–7. tion. Acad Med 1990;65:611–21.
26 Eva KW, Brooks LR. The underweighting of implicitly 41 Woods NN, Brooks LR, Norman GR. The value of basic
generated diagnoses. Acad Med 2000;75 (Suppl):81–83. science in clinical diagnosis: creating coherence
27 Eva KW. The influence of differentially processing among signs and symptoms. Med Educ 2005;39:in press.
information on diagnostic decision making. Disserta- 42 McKenzie CRM. Taking into account the strength of
tion. Hamilton, Ontario: McMaster University 2001. an alternative hypothesis. J Exp Psychol Learn Mem Cogn
28 Eva KW, Brooks LR, Norman GR. Forward reasoning as 1998;24:771–92.
a hallmark of expertise in medicine: logical, psycho- 43 Hatala RM, Brooks LR, Norman GR. Practice makes
logical and phenomenological inconsistencies. In: perfect: the critical role of mixed practice in the
Shohov SP, ed. Advances in Psychological Research. Vol. 8. acquisition of ECG interpretation skills. Adv Health Sci
New York: Nova Science Publishers, Inc. 2002;41–69. Educ 2003;8:17–26.
29 Whittlesea BWA, Brooks LR, Westcott C. After the 44 Thompson L, Gentner D, Loewenstein J. Avoiding
learning is over: factors controlling the selective missed opportunities in managerial life: analogical
application of general and particular knowledge. J Exp learning improves case-based transfer. Organ Behav
Psychol Learn Mem Cogn 1994;20:259–74. Human Decis Process 2000;47:98–123.

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106

106 current perspectives

45 Norman GR, Brooks LR, Rosenthal D, Allen SW, 47 Swanson D, Norman GR, Linn RI. Performance-based
Muzzin LJ. The development of expertise in derma- assessment: lessons from health professions. Educ Res
tology. Arch Dermatol 1989;125:1063–8. 1995;24:5–12.
46 Nendaz MR, Tekian A. Assessment in problem-based
learning medical schools: a literature review. Teach Received 12 March 2004; accepted for publication 28 June
Learn Med 1999;11:232–43. 2004

Blackwell Publishing Ltd MEDICAL EDUCATION 2004; 39: 98–106