Semantics – relating to meaning in language or logic.

In the didactic teaching days of old, written exams required recall and regurgitation of factual information, BUT oh how things have changed, and for the better.  In keeping with socratic methods, the RACGP exams now require contextual interpretation of knowledge.  That’s where semantics come in.   The stem of an exam question contains the  information necessary to define both the problem and its context, and if we misinterpret those critical key features, then our clinical knowledge will not be applied appropriately.  The key features are required to apply effective clinical reasoning. Our interpretation of the features can be influenced by our clinical, personal and affective biases.

Clinical Reasoning

Previous posts have explored clinical reasoning in detail, but to reiterate simplistically, it involves the ‘push and pull’ between the dual processes of non-analytical (pattern recognition) and analytical (hypothetico-deductive) approaches.   We apply clinical reasoning to both the diagnostic and therapeutic processes.   We tend to use non-analytical strategies when we perceive the problem to be easier (perhaps in clinical areas where we feel more confident), and save the analytical processes for the harder stuff.   This can be a trap, as if our perception of difficulty is based on misinterpreted information, it is likely that we will use the incorrect technique (1).  Then there’s also causal reasoning, which is going back to first principles (or pathophysiology) and this can be a useful backup strategy to approach a topic that you’re less familiar with.

Essentially when we are seeing a patient, or answering an exam question, we take the following steps:

  1. Gather pertinent information including context
  2. Develop a problem representation
  3. Consider a list of differentials
  4. Compare each differential to the pertinent information, our knowledge bank, and ‘illness scripts’ including those developed from experiences.
  5. We use frameworks such as surgical sieves and Murtagh’s PROMPT to make sure we don’t miss anything.

Step 1 is the rate limiting step – even if we are experienced at steps 2-5, if we misinterpret information at Step 1, we will be reasoning with the wrong information.

The literature (2) identifies additional points at which clinical reasoning can go astray:

  1. Difficulties in generating hypotheses, identifying cues and directing data gathering – i.e. failure to identify the key features required to inform the diagnosis.
  2. Premature closure, i.e. focusing on some key features more than others.
  3. Difficulties in prioritizing, e.g. failure to identify the impact of context.
  4. Difficulties in painting an overall picture of the clinical situation, e.g. poor problem representation.
  5. Difficulties elaborating a management plan, e.g. not considering the influence of the Domains of General Practice.

All of above hinge on the semantic interpretation of the information given.


How do we interpret, link and organise information as clinicians?

We do this using syntax and semantics.

Syntax – The arrangement of words and phrases to create well-formed sentences in a language.

As you can appreciate, if English is your second language, the syntax part can be even more challenging.   We use syntax to define which symptoms and signs belong to what diagnosis.  This is called Syntactic theory, which also includes interpretation of timelines given, e.g. acute vs subacute vs chronic (3).

For example:  A patient presents telling you that they have had “non-stop diarrhoea for the past eight days”.

We can interpret this in several ways:

Most simplistically, the patient has had ‘non-stop diarrhoea” and ‘for the past eight days”.

OR to interpret more comprehensively,  the patient has had constant diarrhoea of recent onset.

The patient then tells you they have had ‘high fevers’ – which favours a bacterial versus viral diagnosis.   If the patient had noted “loose and more frequent bowel movements for the past six months” you may interpret the problem as a “chronic but variable bowel habit” and you might consider an alternative diagnosis of Inflammatory Bowel Disease.

Doctors with high diagnostic accuracy tend to have a large number of Semantic Axes, i.e.  multiple ways to consider and compare the information, as opposed to ‘weak diagnosticians’ who use the presenting facts with minimal consideration as to how those facts relate to each other.

The difference between good and poor diagnostic reasoning is not  lack of knowledge, but difficulty developing the semantic connections between the facts.  Doctors are intelligent people and have well formed study techniques by the time they reach GP Fellowship exams.  It is uncommon to have difficulty with the exams due to lack of knowledge, rather it’s lack of exposure to, and practice of semantic analysis and clinical reasoning, that is the contributor to unsatisfactory exam attempts.

So what’s the semantic solution?

How do we connect the dots and increase our semantic axes?

  1. Recognise the role and importance of semantics.
  2. Identify our clinical and cognitive biases and how they might influence our semantic analysis.  There is an ebook available that explores the common clinical biases.
  3. Make semantics explicit when answering an exam question:
    • Write down the predominant key features – age, gender, presenting complaint.
    • Generate a differential list.
    • Write down the remaining key features.
    • Narrow your differential list.
    • Write a Problem Representation – this is a one-sentence summary that highlights the defining features of the case.   Now check back and see how it fits with your differentials.
  4. Ensure that you’re not adding any additional information or making any assumptions (usually based on clinical experiences).

Studies suggest merit in the 4-step approach above (4).   Propositional analysis of a problem is essentially analysing it in its entirety, akin to reading the stem of the question once, and developing a differential, i.e. pattern-recognition (5).   Structural semantic analysis requires considering the information presented in a sequential manner, i.e. hypothetico-deductive reasoning.

If you are studying with a colleague – compare your problem representations, and justify to each other why you have chosen a particular differential, and why you may have excluded others.  If you have attempted the exams previously, there is significant benefit in working with a Medical Educator or GP Supervisor to explore the mechanics of how you thought through a question and articulated the answers. By engaging in this review process, you will increase the linkages between the information you have gathered – and you should have honed your semantics.

The purpose of the RACGP written exams is to assess how you apply clinical reasoning in context.   The exams assess how you utilise the information given to move fluently between pattern-recognition and hypothesis testing, as that is the skill required of an effective and safe GP.

If you don’t think about semantics, your approach and answers are more likely to reflect pattern-recognition, and that’s just not enough to get across the pass line.


  1. Arthur S Elstein, and Alan Schwartz“Clinical problem solving and diagnostic decision making: selective review of the cognitive literature.” BMJ (Clinical research ed.) vol. 324,7339 (2002): 729-32. doi:10.1136/bmj.324.7339.729
  2. Marie-Claude Audétat, Suzanne Laurin, Gilbert Sanche, Caroline Béïque, Nathalie Caire Fon, Jean-Guy Blais & Bernard Charlin (2013) Clinical reasoning difficulties: A taxonomy for clinical teachers, Medical Teacher, 35:3, e984-e989, DOI: 10.3109/0142159X.2012.733041
  3. David Steward, Georges Bordage, Madeleine Lemieux (1991) Cognitive structures of Experts and Novices,  Academic MedicineVolume 66 – Issue 9 – pp. S70-S72
  4. Madeleine Lemieux and George Bordage (1992) Propositional versus structural semantic analyses of Medical Diagnostic ThinkingCognitive Science, 16, 185-204.
  5. Jose Arocha, Dongwen Wan, Vimla Patel (2005) Identifying reasoning strategies in medical decision making:  A methodological guide, Journal of Biomedcial Informatics, Volume 38, Issue 2, pp. 154-171.

Please feel free to comment.

Scroll to Top
%d bloggers like this: