SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

Medical Education 2011: 45: 407–414

Context  Previous research has demonstrated the influence of familiar symptom descriptions and entire case similarity on diagnostic reasoning. In this paper, we extend the role of familiarity to examine the influence of familiar non-diagnostic patient information (e.g. name and age) on the diagnostic decisions of novices, both immediately following training and after a delay. If an instance model (reliance on similar previously seen cases) has strong explanatory power in clinical reasoning, we should see an influence of familiar patient information on later cases containing similar identifying characteristics even though such information is objectively irrelevant.

Methods  Thirty-six participants (undergraduate psychology students) were trained to competence on four simplified psychiatric diagnoses and allowed to practise their diagnostic skills on 12 prototypical case vignettes, for which feedback was provided. One-third of participants were tested immediately, one-third following a 24-hour delay, and one-third following a 1-week delay; all were tested on novel cases. Test cases were created to have two equiprobable diagnoses, both of which were supported by two novel symptom descriptions. However, one diagnosis was also supported by non-diagnostic patient information similar to information on a patient seen in the training phase. A deviation from an equal assignment of diagnostic probability, in support of the familiar patient information, demonstrates a reliance on the familiar, non-diagnostic information, and therefore indicates an instance model of reasoning.

Results  Participants assigned significantly higher diagnostic probability to the diagnosis cued by the familiar patient information (52.6%) than to the plausible alternative diagnosis (38.9%). Participants also reported a higher number of clinically relevant symptoms to support the diagnosis associated with the familiar patient information than to support the plausible alternative diagnosis. The influence of familiar patient identity was consistent across delay periods and cannot be accounted for by the forgetting of diagnostic rules.

Conclusions  Participants were clearly relying on familiar patient identity information as evidenced by their diagnostic conclusions and differential reporting of clinically relevant features. These results support an instance model of reasoning which is not limited by whole case similarity or similarity of diagnostic information.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

When acquiring expertise in a complicated domain such as medicine, it is generally accepted that a student needs both formal knowledge and experience. There is something to be gained from each type of learning: analytical knowledge of basic mechanisms, signs and symptoms, epidemiology and so forth, which is primarily acquired in the pre-clinical years, and the experiential knowledge that is slowly absorbed in clinical settings during clerkship, residency and practice. Evidence has accrued from studies of clinical reasoning for the use of basic science knowledge,1 consideration of base rates or Bayesian reasoning (e.g. 2), and the role of similarity to previously seen examples (for a review, see3). Yet, although substantially more years are spent on acquiring clinical experience than formal knowledge, we understand little about how this experience contributes to expertise. One possibility is that clinical experience leads to the development of a large number of prior cases, or instances, on which to draw when faced with the challenge of diagnosing a new patient. This paper will examine the role of similarity to previously seen cases, specifically investigating the influence of familiar non-diagnostic information on the diagnostic decisions of novices.

One line of research3 has demonstrated that clinical diagnostic reasoning can be strongly influenced by specific experiences. The use of prior instances has been demonstrated in both doctors and students;4–6 however, the majority of these studies have used a visual domain, dermatology, and similarity was based on overall visual appearance. This demonstration of the reliance on similarity in a relatively expert population supports the notion of exemplars, or instances, in decision making, whereby we maintain information about previous cases in memory and use these memories to aid our current decision making. Although the overall visual similarity of cases has been demonstrated to influence clinical reasoning,4–6 more recent research has attempted to delineate the role of similarity in written case materials7,8 (Young ME, Norman GR, Brooks LR, unpublished data, 2009) to ensure the generalisability of an instance theory of clinical reasoning beyond the visually rich domains of medicine.

Further, in areas of medicine such as dermatology, it is difficult to determine whether the influence of previously seen cases is driven by the familiarity of the clinically relevant (symptoms or features of a lesion) or clinically irrelevant (patient appearance or context) components of a case. Similarity to a previous case can occur on any or multiple stimulus dimensions; little distinction is made between encoding critically relevant and irrelevant information in memory.9,10 However, anyone applying a diagnostic rule makes a clear distinction between what is relevant and what is not; it is the function of a rule to point to what is important in a diagnostic case. Some recent studies have shown that familiar symptoms (clinically relevant information) can influence the diagnostic decisions of novices; novice diagnosticians rely more heavily7,8 or differentially report (Young ME, Norman GR, Brooks LR, unpublished data, 2009) familiar manifestations of clinical features. More specifically, participants assign more diagnostic weight to a symptom description they have encountered before and are more likely to diagnose a fictional patient with the diagnosis supported by the familiar symptom description.7,8 This functional role of similarity, both holistic and specific, may be what helps an early learner to understand how to apply a diagnostic rule and may represent a very adaptive use of previous experience.

An instance theory (or exemplar theory, here used interchangeably) does not make a distinction between familiar clinically relevant and clinically irrelevant information and, thus, familiar non-diagnostic (or non-clinically relevant) information may influence diagnostic decision making. If an instance model is indeed a reasonable description of clinical reasoning, then non-diagnostic information that is similar to a previous experience may well bias diagnosis in a manner that is consistent with the prior case, despite the fact that the retrieval cues are objectively irrelevant to the diagnosis. Although the role of similarity has been established in clinical reasoning (see 3), few studies have investigated the role of familiar non-diagnostic information.11 As one study demonstrated that patient identity can influence the interpretation of electrocardiographs (ECGs) in residents (but not medical students),11 it is possible that previous research investigating the role of familiar patient identity in ECG interpretation11 may have inadvertently used what may be stereotypical profiles of cardiac patients (e.g. relatively young males in high-stress, mentally demanding jobs; e.g.12) to bias participant responses. Further, this prior study used identical age and occupation descriptions and therefore the generalisability of the findings to everyday clinical contexts may be limited. One might then conclude that the impact of patient identity could be driven by a representativeness heuristic,13 by relying on medical stereotypes,12 or simply ‘matching’ to an identical patient – a rare occurrence in medical practice – rather than by reliance on a previously seen case. Finally, Hatala et al.11 found that reliance on familiar patient identity information was limited to residents and did not influence medical students. It may be possible that the biasing impact of familiar patient identity information is limited to populations with specialised medical knowledge, suggesting that the exemplar model of reasoning may come about as expertise develops rather than as a fundamental component of decision making at more junior levels.

Previous research has demonstrated support for exemplar-based reasoning within novice populations,7,8 but these demonstrations have thus far been limited to clinically relevant information, which may reflect a learning strategy for recognising manifestations of a diagnostic rule, rather than an automatic memory retrieval of an entire previously encountered case. The research presented here was conducted to investigate the role of familiar non-diagnostic information, presented in the form of familiar patient identity, on the diagnostic decisions of novices. As patient identity is rarely diagnostically informative, we should not expect to see reliance on a similar previous case when cases match on an irrelevant dimension.

To extend earlier investigations, we carefully manipulated aspects of patient information to maintain similarity, but not identity, in order to ensure that these data were not relevant to the diagnosis and did not conform to any popular stereotypes or base rate probabilities regarding the specific disorders used (e.g. if a disorder was popularly believed to be more present among males, female patient identities were used). We hypothesised that, because of the role of non-analytic (see3), exemplar-based reasoning, the clinically irrelevant or non-diagnostic information would bias participants towards the diagnosis supporting that sense of familiarity. More specifically, we hypothesised that participants faced with equiprobable diagnoses would assign significantly higher diagnostic probability to the disorder supported by familiar patient characteristics (e.g. age, name, marital status, employment) consistent with a previously seen training case. Finally, we examined the effects of 24-hour and 1-week delay periods in order to examine whether these experience-based effects were transient or whether they remained strong after a relatively long delay.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

Participants

Thirty-six students enrolled in a first-year undergraduate psychology course participated for course credit. The McMaster University Research Ethics Board approved this study. Informed consent was obtained from all participants.

Stimuli

Four pseudo-psychiatric disorders were created for the purposes of this research. Similar methods and materials have been used before (see 8) to examine the role of familiarity within the application and use of a diagnostic rule. The disorders were drawn from disparate areas of psychiatry and included mania, obsessive compulsive disorder, schizophrenia and paranoid personality disorder. The rules were versions of the rules found in the Diagnostic and Statistical Manual for Mental Disorders, 4th edn (DSM-IV),14 which had been simplified to ease learning. Table 1 shows the modified psychiatric rules used in this experiment. Participants were told that these disorders were fictional and were not representative of the real-world disorders of the same name and thus that they should rely only on the material presented in the experiment. This instruction was included to ensure that participants would be less likely to rely on medical stereotypes if any were inadvertently included in the patient description information, or on other prior knowledge of psychiatric conditions.

Table 1.   Adapted medical diagnoses and diagnostic rules used in this study
Diagnostic categoryDiagnostic rule
ManiaDecreased need for sleep
Increased energy
Inflated self-esteem
More talkative
SchizophreniaHallucinations
Delusions
Disorganised speech
Disorganised behaviour
Paranoid personality disorderConcerns over fidelity of spouse
Distrustful of others
Reads hidden meaning into daily events
Reluctant to confide in others
Obsessive compulsive disorderRepetitive behaviours
Recurring thoughts
Difficulty ignoring thoughts
Interferes with day-to-day life

Each disorder was characterised by four unique, non-overlapping, characteristic symptoms. During practice and test phases of this experiment, each symptom was presented using a patient voice or a unique symptom description (e.g. the rule feature ‘concern over fidelity of spouse’ would be presented as ‘she has been worried ever since her husband started a new job with an old girlfriend, so she has been secretly checking his e-mail regularly’). Each symptom description was unique and participants received feedback during training regarding the appropriate ‘translation’ of the symptom description. During practice and test phases of the experiment, stimuli were presented as case vignettes, with personal identifying information (e.g. name, age, type of employment, familial situation) and unique symptom descriptions. Participants were asked to report their diagnoses in the form of percentages assigned to each of the four diagnostic options. Participants were instructed to distribute the diagnostic percentages as they saw fit, with the only restriction being that they must sum to 100%. Following the presentation of a written case vignette and the assignment of diagnostic percentages, the case was removed and participants were asked to report the clinically relevant features, or symptoms, present in the case.

During the practice phase of the experiment, participants were shown prototypical written case vignettes, containing all four features of the diagnostic rule for that disorder, presented as unique symptom descriptions. During the test phase, each case vignette included two symptom descriptions supporting one diagnosis, and two symptom descriptions supporting an alternative diagnosis, creating a case with two equiprobable diagnoses. Additionally, each case contained a patient description similar to one seen in the practice phase of the experiment. Similarity was defined along five dimensions, the values for each of which were similar, but never identical. Patients had names starting with the same phoneme, were within 5 years in age (always within the same decade), had similar but never identical professions, had similar but never identical familial situations, including martial status and number of children, and lived in similar areas surrounding McMaster University (Hamilton, ON, Canada). An example of similar, but not identical, patients can be seen in Table 2.

Table 2.   Sample of familiar patient identity pairs. Patients were constructed to have similar, but never identical, ages, names, familial situations, professions and areas of residence
NameAge, yearsOccupationFamilial situationOther (including area of residence)
Allen65Accountant nearing retirementFour children, each with a minimum of two childrenCaregiver for his wife, who has Alzheimer’s disease
Albert64Retired financial consultantTwo children, each with a minimum of three childrenCaregiver for his sister, who has Parkinson’s disease

For example, if, during training, a participant met Allan as an individual diagnosed with schizophrenia, Albert’s case vignette would include two symptom descriptions that would support a diagnosis of schizophrenia, and two symptom descriptions that would support another diagnosis, such as obsessive compulsive disorder. If participants were influenced by the familiarity of the patient identity information, then we would expect to see a deviation from an equal split in diagnostic probabilities in favour of the diagnosis supported by the familiar patient identity.

Procedure

This experiment was programmed and presented using RunTime Revolution 2.1 (RunTime Revolution Ltd, Edinburgh, UK) and presented on a computer screen. The experiment included a learning phase, a practice phase and a test phase. Twelve participants were tested immediately following training, 12 were tested 24 hours after training and 12 were tested 1 week after training.

Learning phase

The learning protocol was identical for each disorder, but the order of learning was randomised across participants. Participants were given a list of the four symptoms characteristic of the disorder to study. Participants were asked to correctly identify the four features characteristic of the disorder from the full list of 16 symptoms. If participants were unable to do this, they were asked to review the symptom list and re-take the quiz. If they could correctly identify the appropriate features in the diagnostic rule, they saw an initial prototypical case vignette and were asked to identify the features present in the patient vignette. This protocol was repeated for each of the four modified psychiatric disorders. Before participants moved to the practice phase, they were again required to identify the four correct features for each disorder. A pass criterion of 15/16 was set; if participants were unable to meet this criterion, they returned to the beginning of the learning phase. If participants passed this strict learning criterion, included to ensure competency, they progressed to the practice phase of the experiment.

Practice phase

The practice phase included 12 prototypical case vignettes, presented in random order. Each case contained patient identity information (name, age, area of residence, employment and familial situation, etc.) and four unique symptom descriptions indicative of a single diagnosis. There were three practice cases for each of the four disorders, and participants were asked to assign diagnostic probabilities to each of the four possible diagnoses and report the relevant features in the case vignette. After participants had reported the clinically relevant features in the case, they received feedback regarding not only the correct diagnosis, but also the relevant features present in the case. The practice phase in this experiment fulfilled two purposes: firstly, it allowed participants to hone their diagnostic abilities in relation to the pseudo-psychiatric disorders, and, secondly, it was designed to expose participants to a variety of unique patients that functioned as familiar patients in the test phase.

Test phase

Participants were randomly assigned to the immediate test group (which went directly from the practice phase to the test phase), the 24-hour delay group (which returned to the laboratory 24 hours after the practice phase to complete the test phase) or the 1-week delay group (which returned to the laboratory 1 week after the practice phase to complete the test phase). A total of 12 unique case vignettes were presented in a randomised order. After participants had assigned their diagnostic probabilities, the case vignette was removed from the screen and participants were asked to report the clinically relevant information in the case.

Post-test assessment

Following the test phase of the experiment, participants in the 24-hour and 1-week conditions were tested on their memory of the diagnostic rule. They were asked to generate the four features characteristic of each disorder in a free recall task. This was added for these groups of participants to ensure that they could indeed recall the diagnostic rule following a substantial delay period.

Analysis

Diagnostic probabilities

Mean diagnostic probabilities for the disorder supported by the familiar patient identity and the plausible alternative diagnosis were computed. Diagnostic probabilities assigned were analysed using a mixed-design analysis of variance (anova) in which mean diagnostic probability assigned (mean diagnostic probability assigned to the diagnosis supported by the familiar patient identity compared with mean diagnostic probability assigned to the plausible alternative diagnosis) was the within-subject comparison of interest, and delay (immediate test, 24-hour delay, 1-week delay) was the between-subjects factor. The interaction between delay and mean diagnostic probability was examined using a mixed-design anova to investigate the longevity of the influence of familiar non-diagnostic information.

Analysis of clinically relevant features identified

Following the assignment of diagnostic probabilities, participants were asked to report the clinically relevant features, or symptoms, present in each case. Responses were coded by an independent research assistant blinded to the individual biasing manipulation. Features were coded as: (i) consistent with the diagnosis supported by the familiar patient identity; (ii) consistent with the plausible alternative diagnosis, and (iii) other. The number of features identified congruent with the biasing manipulation and with the plausible alternative were averaged across the 12 cases to create the mean number of each type of feature identified by the participants. To examine whether participants were more likely to report features congruent with the familiar patient identity than with the alternative diagnosis, a mixed-design anova was conducted, in which the type of diagnosis supported by the features identified was the within-subjects factor of interest, and the delay (immediate testing, 24-hour delay, 1-week delay) was the between-subjects factor of interest.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

Diagnostic probabilities

Mean diagnostic probabilities assigned for the immediate test group, 24-hour delay and 1-week delay groups are shown in Fig. 1.

image

Figure 1.  Mean diagnostic probabilities assigned to the diagnosis supported by the familiar patient identity and the plausible alternative diagnosis. The mean diagnostic probabilities assigned to the other two of the four diagnostic options were collected, but are not displayed here. Data are presented across test group (immediate test, 24-hour delay, 1-week delay). Error bars represent the standard error of the mean

Download figure to PowerPoint

Participants assigned significantly higher diagnostic probabilities to the disorder supported by the familiar patient identity (mean diagnostic probability assigned to the disorder supported by familiar patient identity = 52.6%, standard deviation [SD] = 4.3%; mean diagnostic probability assigned to the plausible alternative diagnosis = 38.9%, SD = 7.2%; F1,36 = 71.40, p < 0.001). A delay by familiarity interaction was also found, with the strength of the familiarity effect increasing across delay period, as seen in Fig. 1 (F2,36 = 3.40, p < 0.05).

Identification of clinically relevant features

Participants reported slightly, but significantly, higher numbers of features or symptoms for the diagnosis supported by the familiar patient identity than the plausible alternative (mean number of symptoms identified for the diagnosis congruent with the familiar patient identity = 1.87; mean number of symptoms identified for the plausible alternative diagnosis = 1.72; F1,34 = 6.55, p < 0.05). The number of features identified for the diagnosis supported by the familiar patient identity or the plausible alternative diagnosis did not interact with delay (F2,34 = 0.56, p = 0.64).

Post-test recall of the diagnostic rules

Participant responses to the free recall of diagnostic rules were coded and performance was compared across delay periods (24 hours and 1 week). No significant difference in performance was found between the two delay conditions (mean performance for 24-hour delay = 88.5% correct; mean performance for 1-week delay = 87.5% correct; t[22] = 0.24, p = 0.81).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

Novice diagnosticians assigned significantly higher diagnostic probability to the disorder supported by the familiar non-diagnostic patient identity information than to the equally valid alternative diagnosis. If a case vignette included a patient who had a similar name, occupation, age and familial situation to a previously encountered case, participants were more likely to consider the patient to have a similar diagnosis. This pattern of response occurred across participants who were tested immediately following training and those who were tested after 24-hour and 1-week delays, and the reliance on non-diagnostic information increased across delay periods. The increased impact of similar non-diagnostic information was not accounted for by participants’ forgetting of the diagnostic rule, as performance on post-test assessment remained near 90%. Additionally, as each feature mentioned in the rules had an equal probability of occurring in support of the diagnosis cued by the familiar patient identity and the plausible alternative diagnosis, small decrements in post-test recall for the diagnostic rule cannot fully account for the influence of familiar patient identity across delay periods.

The results presented here further support an exemplar, or instance, theory of clinical reasoning. The fact that the effect of prior cases was based entirely on irrelevant and non-stereotypical factors, such as patient name, strongly supports the claim of a qualitatively different thinking process that is not restrained by diagnostically relevant information. These results also indicate that the influence of a single case can be relatively long-lasting, at least up to 1 week, which suggests that the exemplar-based effects previously demonstrated are not transient.3,11 Further, the specific strategy of designing the similar, but not identical, patient identity information contributes to the ecological validity of exemplar-based reasoning by showing that retrieval is not limited to identical features.

Participants also reported significantly more symptoms in support of the disorder associated with the familiar patient identity than in support of the plausible alternative diagnosis, although the numerical difference was small. One interpretation of these findings is that diagnostically irrelevant information may function as a cue for the retrieval of a similar previous case and therefore influence current decision making. The increased reporting of features that support the diagnosis cued by the familiar patient identity information may indicate an interactive influence of diagnostic and non-diagnostic information, or more analytical (rule-based) and more exemplar-like (familiarity-based) processes. Although this remains speculative, it may suggest that participants are rapidly recalling a previous instance, gaining a sense of fluency (e.g. 15) or are being cued to a particular diagnosis and are relying on supporting information for that diagnosis, which is, in essence a possible demonstration of a type of confirmation bias (e.g. 16). Although this conclusion remains speculative, the data presented here support the influence of non-diagnostic information on the reporting of clinically relevant features. In a more common medical education framework, these data suggest a more interactive or iterative use of non-analytic and analytic reasoning strategies,3 as the presence of familiar non-diagnostic information (traditionally considered to influence non-analytic reasoning) influences the detection, or reporting, of clinically relevant symptoms (traditionally the role of analytic reasoning). The possibility of an iterative interaction of reasoning strategies is not frequently discussed as this contrasts with a more classic either/or approach or conceptualisation of reasoning strategies, and is a novel theoretical contribution of this research.

This study has some limitations. The study of the influence of a single prior case and the role of specific previous experience requires control of prior experience. With clinicians, this range of experience is both extensive and idiosyncratic, which is a unique challenge to this type of research. As a solution to this problem we chose to use novices and to teach them simple medical rules, thereby controlling their relevant experience. The disadvantage, however, is that by studying exemplar effects in novices, we can only infer that the phenomena demonstrated in this paper can be equated with, or related to, those seen in more expert populations.4–6,11 Further, the exemplar-based reasoning demonstrated here may add to the literature demonstrating that non-analytic reasoning becomes more prevalent with increased levels of expertise11 by showing that it is also present and enduring in novices. Finally, static case presentation using written materials is not perfectly analogous to clinical encounters, although the style of presentation and type of material are quite common in all levels of medical education. Indeed, restricting the presentations to formal written materials may, if anything, reduce the effect of idiosyncratic irrelevant information and may even have biased against the observed effects.

This research has demonstrated the role of non-diagnostic, or diagnostically irrelevant, information on the diagnostic decisions of novices using a task that has strong rules and decision criteria, and so should be an ‘analytic’ task. This research suggests that early medical learners do not rely solely on analytic reasoning and objectively relevant features of a case. It may be simplistic to consider the development of expertise simply as a matter of acquiring and applying the ‘rules’ of clinical medicine. Integrating the learning of medical rules with clinical cases may not only aid in the learning of a medical rule, but may support the retention and later use of a diagnostic rule.

Contributors:

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

MEY contributed to the design, conceptualisation, and execution of the study. She also analysed the results, and drafted the manuscript. LRB contributed to the conceptualisation and design of the study, and aided in the development of the manuscript. GRN contributed to the conceptualisation of the study, analysis of the results, and aided with the development of the manuscript.

Acknowledgements:

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

The authors would like to thank Elizabeth Howey for her tireless aid, and the reviewers of this paper for their many helpful suggestions.

Ethical approval:

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References

this study was approved by the Research Ethics Board of McMaster University, Hamilton, ON, Canada.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Contributors:
  8. Acknowledgements:
  9. Funding:
  10. Conflicts of interest:
  11. Ethical approval:
  12. References
  • 1
    Woods NN, Brooks LR, Norman GR. The value of basic science in clinical diagnosis: creating coherence among signs and symptoms. Med Educ 2005;39:10712.
  • 2
    Elieson SW, Pappa FJ. The effects of various knowledge formats on diagnostic reasoning. Acad Med 1994;69:813.
  • 3
    Norman GR, Young ME, Brooks LR. Non-analytic models of clinical reasoning: the role of experience. Med Educ 2007;41:11405.
  • 4
    Brooks LR, Norman GR, Allen SW. Role of specific similarity in a medical diagnostic task. J Exp Psychol Gen 1991;120:27887.
  • 5
    Allen SW, Norman GR, Brooks LR. Experimental studies of learning dermatological diagnosis: the impact of examples. Teach Learn Med 1992;4:3544.
  • 6
    Kulatunga-Moruzi C, Brooks LR, Norman GR. Coordination of analytic and similarity-based processing strategies and expertise in dermatological diagnosis. Teach Learn Med 2001;13:1106.
  • 7
    Dore K, Weaver B, Norman GR. Does a rose by any other name smell as sweet? The importance of variation in case presentation to diagnostic reasoning. Presented at the Meeting for Research in Medical Education, 911,November 2005, Washington, DC.
  • 8
    Young ME, Brooks LR, Norman GR. Found in translation: the impact of familiar symptom descriptions on diagnosis in novices. Med Educ 2007;41:114651.
  • 9
    Brooks LR. Non-analytic concept formation and memory for instances. In: RoschE, LloydBB, eds. Cognition and Categorization. Hillsdale, NJ: Lawrence Erlbaum Associates 1978;169211.
  • 10
    Brooks LR. Concept formation and particularising learning. In: HansonPP, ed. Information, Language and Cognition. Vancouver, BC: University of British Columbia Press 1990;141167.
  • 11
    Hatala R, Norman GR, Brooks LR. Research basic to medical education: influence of a single example on subsequent electrocardiogram interpretation. Teach Learn Med 1999;11:1107.
  • 12
    Clarke JN. Cancer, heart disease, and AIDS: what do the media tell us about these diseases? Health Commun 1992;4:10520.
  • 13
    Tversky A, Kahneman D. Judgement under uncertainty: heuristics and biases. Science 1974;185:112431.
  • 14
    American Psychiatric Association. Diagnostic and Statistical Manual for Mental Disorders, 4th edn. Washington, DC: APA Press 2000.
  • 15
    Kelley CM, Jacoby LL. Memory attributions: remembering, knowing and feeling of knowing. In: RederLM, ed. Implicit Memory and Metacognition. Mahwah, NJ: Lawrence Erlbaum Associates 1996;287307.
  • 16
    Bower GH. Contacts of cognitive psychology with social learning theory. Cognit Ther Res 1978;2:12346.