Asking to advance: a discourse analytic perspective on how Socratic questioning scaffolds medical house officers’ expertise

Abstract

Background

Housemanship is a transitional phase that prepares medical graduates to become safe, independent practitioners. Learning during this phase occurs through patient-case talk with senior doctors, a hybrid interaction that serves patient care and learning. Effective case-based learning has been linked to fostering participation, dialogue and strategic language use, with Socratic questioning recognised as a key scaffolding approach. However, discursive evidence of how such questioning unfolds in everyday training remains limited. This study examines how house officers and senior doctors formulate questions and responses to facilitate teaching and learning in two training contexts.

Methods

We performed a discourse analysis of naturally occurring patient-case talk between house officers and senior doctors in the Paediatrics Department of a teaching hospital in Kuala Lumpur. Audio-recorded interactions produced 21 units of patient-case talk from five Continuing Medical Education (CME) sessions and 78 units from 13 Ward Rounds (WRs). Drawing on concepts of questioning and responding practices, the analysis identifies how senior doctors design questions to scaffold learning and how house officers orient to these designs in producing responses.

Results

Senior doctors use varied questioning practices to scaffold developing expertise and promote self-correction. In CME sessions, they frequently use polar questions and designedly incomplete utterances, with house officers displaying interactional agency through repetitional and transformative responses. During ward rounds, senior doctors rely more on wh-questions, often paired with designedly incomplete utterances, to stimulate reasoning and self-initiated repair. Collectively, these practices foster reflection, critical thinking and discovery-based learning without prematurely providing answers.

Conclusion

This paper demonstrates how questioning functions as a key discursive resource for scaffolding expertise in Malaysian medical training. It extends understanding of Socratic questioning by showing how distinct question designs support self-correction, reasoning and learner agency across teaching contexts. Strategic questioning allows senior doctors to address knowledge gaps in a manner that aligns with principles of psychological safety. As a result, house officers engage in the interactional display of a clinical voice and demonstrate an orientation toward managing uncertainty as part of learning. These insights provide actionable, evidence-based direction for faculty development and clinical teaching practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12909-026-09364-8.

Background

Housemanship or internship training is a pivotal phase in Malaysia’s medical education system, during which new medical graduates known as house officers (HOs), develop clinical competence through guided participation under the supervision of senior doctors (SDs) across six postings over a period of 2 years. As a transitional period between undergraduate education and independent practice, housemanship is characterised by close supervision, division of labour and progressive responsibility, with SDs playing a central role in shaping HOs’ clinical reasoning skills, procedural skill acquisition, independent decision-making skills, and expertise development. Learning during housemanship unfolds across multiple settings including formal and informal teaching sessions, ward rounds and corridor discussions, and is supported by approaches such as peer-assisted learning, problem-based learning (PBL), and case-based learning (CBL). Among these, CBL remains one of the most established and widely used approaches in postgraduate medical training [1–3].

A key discursive site through which case-based learning occurs is patient-case talk: pedagogical exchanges in which patient cases are used as stimuli for learning across diverse settings. Existing literature associates effective postgraduate case-based learning with practices that foster participation, employ structured discussion methods, and facilitate dialogue, highlighting the strategic role of language in supporting these aims [3]. However, there remains limited empirical insight into how such practices unfold in authentic clinical teaching encounters.

Central to these exchanges is educators’ use of strategic questioning to stimulate learners’ thinking [4]. Questioning is widely recognised in health professionals’ education as a practice that promotes learner engagement, sustains concentration and enhances comprehension [5]. Research indicates that effective questioning extends beyond eliciting factual recall and instead, probes reasoning, extends dialogue and promotes deeper understanding through techniques such as the “reflective toss” [6], Initiation-Response-Feedback-Response-Feedback cycles [7], conceptual change questions [8] and follow-up strategies [9]. Among these approaches, Socratic questioning represents a central form of inquiry-based teaching.

Socratic questioning involves the systematic use of questions to prompt learners to make their thinking explicit, examine assumptions, consider alternatives and adjust their reasoning [10]. When used effectively, it enables educators to ascertain learners’ level of understanding and calibrate instruction through graded sequencing of questions [11]. By progressing from simple to complex questions, senior doctors can scaffold learners’ cognitive processes while regulating the level of challenge. However, if mishandled, particularly when incorrect responses are met with evaluative criticism or public embarrassment rather than constructive exploration, the practice risks slipping into what has been described as “pimping,” a questioning practice characterised by rapid, high-pressure and often unanswerable questioning of trainees widely critiqued in the medical education literature [12, 13]. Prior work suggests that the distinction between Socratic questioning and pimping lies in questioner’s intent and how that intent is perceived by learners [14–16]. This warrants attention because if the educational environment does not support learners’ recognition of Socratic teaching, even well-intentioned Socratic questions may be misinterpreted as pimping [15].

While ineffective questioning practices have been extensively discussed [15–18], comparatively less attention has been paid to how effective Socratic questioning is enacted moment by moment in workplace learning settings. Best-practice recommendations emphasise the importance of psychological safety [11, 15, 19] and strategies such as the use of wait time, yet little interactional evidence currently demonstrates how these practices are accomplished during clinical teaching. Existing research largely focuses on classifying question types according to pedagogical functions [20–22], the forms of Socratic dialogue [10], dimensions of reasoning [10, 23] or their cognitive levels [11]. While such classifications aid instructional planning, they offer limited insight into the interactional work through which questions are framed and adapted in authentic classroom or ward-round encounters. As a result, the mechanisms through which questions invite, withhold, probe, or narrow learners’ responses remain largely tacit, despite these being central to how expertise is scaffolded in practice.

Closer analysis of these communicative practices is warranted for two reasons. Firstly, claims that the effectiveness of Socratic questioning depends on both its philosophical underpinnings and the questioner’s intent which is reflected in the manner of discourse through which questioning unfolds [24], highlights the importance of how questions are delivered. Secondly, its successful implementation has been shown to rely on educators’ skilful use [25], underscoring the need for empirical work that captures how the pedagogical goals are accomplished through talk.

To address this gap, the present study analyses naturally occurring patient-case talk between Malaysian house officers (HOs) and senior doctors (SDs) in a university teaching hospital. Drawing on interactions from both formal teaching contexts and real-time clinical care, the study examines how questioning is discursively enacted within error-correction sequences to probe understanding, guide reasoning and balance the dual demands of assessment and support. This focus allows for a detailed examination of the communicative work through which expertise is scaffolded and how house officers navigate these interactions as novices negotiating learning and expertise acquisition. Therefore, this study is guided by the following research question: how do senior doctors and house officers collaboratively accomplish expertise development through question-answer sequences in CME and ward round sessions?

Findings from this analysis could potentially contribute empirical insights into the discursive resources through which Socratic questioning operates as a scaffolding tool in workplace learning. Making these practices visible offers two sets of implications. For educators, it identifies questioning strategies that may extend existing pedagogical tools particularly in light of ongoing concerns about limited formal pedagogical training and the need to strengthen educators’ questioning skills [19, 22]. For learners, it clarifies how questioning can be interpreted as a facilitative learning opportunity rather than as an evaluative threat. Greater transparency in the method to which questioning unfolds in practice may help mitigate misinterpretations, strengthen psychological safety and support learners’ engagement with questioning as an integral component of professional learning and development.

Methods

Overview and study design

This qualitative study employs an ethnographic design which is well suited for studying groups in education to understand their behaviours, language and shared patterns of interaction [26]. The study draws on naturally occurring interactions between house officers (HOs) and senior doctors (SDs) in the Paediatrics Department of a teaching hospital in Kuala Lumpur, to investigate how educators and learners engage in teaching and learning during patient-case talk.

Participants, context and ethical considerations

Paediatrics was selected for this study for two primary reasons. Firstly, the study benefited from the strong institutional support of the department’s clinicians, which facilitated access to naturally occurring interactional data. Secondly, Paediatrics is widely regarded as a particularly demanding posting for trainees, often characterised by a high workload and the emotional intensity of the rotation [27]. These pressures are further heightened by unique communication challenges, which are frequently cited as a source of conflict in paediatric encounters [28]. Given the complexity of these interactions (typically involving clinicians, patients and caregivers), the department provides a rich setting for examining how senior doctors use questioning to support clinical learning under pressure.

Using a convenience sampling approach based on voluntary participation, 20 house officers and 9 senior doctors including medical officers, specialists and consultants were recruited. To capture a diverse range of participants across various stages of training, recruitment was conducted over several rounds. Recruitment invitations were circulated via email and text message to all senior doctors and house officers rotating through the department during the data collection period. Information sessions were also held during departmental orientation and subsequent briefings to reach those at both early and late stages of their postings. Following these sessions, house officers who expressed interest were approached individually to facilitate the recruitment and consent processes.

To capture authentic educator-learner interactions across both formal and informal settings, patient-case talk was recorded in two key contexts: Continuing Medical Education (CME) sessions and ward rounds (WRs). CME sessions are structured, bi-weekly departmental meetings where practitioners discuss selected patient cases for collective professional development. These sessions typically comprise several units of discussion on selected cases presented by doctors representing their respective wards. As presenters were scheduled in advance, they were contacted ahead of time via email and text message for consent.

In contrast, ward rounds involved more opportunistic teaching, with learning moments emerging during routine patient care. To manage the logistical and ethical complexities of real-time clinical care, data collection was centered at a specific general paediatric ward, selected for its stable patient profile and lower clinical risk relative to high-acuity environments. This ward was served by a core group of five senior doctors and two rotating specialists during the study period, all of whom provided informed consent to participate. Recording these sessions required a coordinated consent process involving not just participating doctors but also caregivers. Caregivers were approached in the morning prior to the ward rounds, where the study was explained and consent was obtained before any recording was initiated. This approach ensured that the data remained authentic to the natural flow of the hospital’s daily operations while prioritising the ethical requirements of real-time patient care.

Dataset and data collection

This paper draws on a subset of data from a larger project, A Discourse-Based Model of House Officer Socialisation Towards Enhancing Interprofessional Collaboration in Medical Practice, which investigates how house officers (HOs) are socialised into the medical profession through participation in interactional practices involving senior doctors (SDs) during housemanship training. The dataset comprises interactions extracted from Continuing Medical Education (CME) sessions and ward rounds (WRs), which represent contrasting yet complementary sites of educator-learner interaction. Each CME session and ward round typically consists of several discrete units of patient-case talk, with each unit centered on discussion of an individual patient.

For clarity in analysis and reporting, we treated each patient-case discussion as a separate unit of analysis and assigned it a unique label. This labelling system indicates both the teaching context and the specific patient-case unit. For example, CME-PCT3 refers to the third unit of patient-case talk from the CME session, while WR-PCT17 refers to the seventeenth unit of patient-case talk from the ward round. This system enables excerpts to be clearly traced to their clinical teaching context and position within the dataset. Overall, the dataset comprises approximately 22 h of naturally occurring interaction, including 21 units of patient-case talk from five CME sessions and 78 units from 13 ward rounds (see Fig. 1).

Fig. 1.

Data types extracted from two training contexts

Prior to recording, we attended several CME sessions to familiarise ourselves with the setting and to observe how discussions were typically structured. We undertook a similar familiarisation process in one of the patient wards, where we conducted daily observations to understand the organisation of rounds, including aspects such as team composition, spatial layout and overall workflow. These preliminary observations informed the development of a data collection strategy that minimised disruption to routine clinical practice and supported ethical and effective recording. We subsequently recorded the interactions in two settings. CME sessions were recorded between December 2023 and February 2024 in a seminar room, while routine ward rounds were recorded between January and February 2024 in one of the patient wards. Both settings were located within the hospital’s Women and Children Health Complex.

Analytical framework

This study employs an approach within the field of Applied Linguistics known as Discourse Analysis (DA), which encompasses a range of approaches for examining language as it is used in context. In particular, it uses Theme-Oriented Discourse Analysis [29–31], which provides a framework for analysing discourse in relation to both its interactional features and the professional context. Within this approach, the analysts engaged systematically with the discourse, identifying focal interactional practices which are relevant to accomplishing the goals of a particular discourse. Through systematic analysis, we identified question-answer sequences as key interactional devices through which teaching and learning are accomplished. Guided by this identification, the analysis narrows its focus to question-answer practices within error-correction sequences between HOs and SDs during patient-case talk, drawing on interactional research on question and response design [32–44]. To guide the analysis, a coding framework was developed incorporating key question designs typically observed in pedagogical interactions (Table 1) and the response designs through which participants engaged with them (Supplementary Material, Table S1). Utterances were coded as questions when they showed either formal or functional features in design. Formal questions displayed interrogative features at the lexical (e.g., who, why), morphological (e.g., question particles or clitics) or prosodic (e.g., rising intonation) level. Functional questions, by contrast, were those treated by participants as seeking information, confirmation or agreement, regardless of forms.

Table 1.

Core coding scheme of question design in the analysis of patient-case talk

Coding of Questions (Stivers & Enfield, 2010):
For an utterance to be coded as a question, it was required to meet at least one of the following criteria:
i. Formal or functional question design:
• Formal: displays lexical, morphological, syntactic or prosodic features conventionally associated with interrogatives in the language. Lexical marking, such as “who” or “why” questions; morphological marking including question particles or clitics; prosodic marking such as rise in intonation. Rhetorical questions used to express stance were also included.
• Functional: interactionally treated as seeking information, confirmation or agreement, regardless of whether it is marked as interrogative. Newsmarks such as ‘‘Really?’’ or ‘‘Is it?’’ were also included.
ii. Logical-semantic structure of questions: type of answers they made relevant:
• Polar questions make relevant confirmation or disconfirmation of a proposition
• Content questions make relevant an explanatory or informational response
Question Type	Definition	Interactional Purpose	Example
Wh-Question Ask for information or reflection	Questions using wh-formats to target missing information or challenge, with their function dependent on context.	To elicit specific informational details, or to challenge prior claims.	Content-elicitation Why did you admit the patient? Challenge Why is that a reason to admit?
Polar Question Ask for confirmation or revision	Questions seeking confirmation, disconfirmation or reconsideration, with their function dependent on context.	To test or verify a proposition through a “yes” or “no” response, or to prompt revision of a prior response.	(Dis)confirmation-seeking Did you repeat the bloods? Revision-seeking Is it clear?
Designedly incomplete utterance (DIU) Ask for completion	Unfinished utterances that withhold key information to encourage the supply of missing elements.	To prompt learner completion or self-correction.	So the source was? Hypoper___?

Questions were further sub-categorised into wh-questions, polar questions and designedly incomplete utterances (see Table 1 for an overview of the coding scheme and Supplementary Material for detailed descriptions and illustrative examples). Responses to these questions were also coded according to their design. Three main response types were identified: repetitional, which reproduce the question’s key terms; transformative, which transform the question’s agenda or design; and phrasal or clausal, which supply brief or elaborated answers. This framework provided a systematic basis for examining how questions and responses were designed and deployed during patient-case talk, and how these practices facilitated scaffolding and corrective work that supported the progressive development of clinical expertise.

Transcription

We prepared detailed transcripts of the interactions using Jeffersonian conventions [45, 46], capturing features of speech delivery including intonation, pauses, emphasis and volume to show not only `what’ was said but `how’ it was said. The transcription symbols used in this study are listed at the end of this paper.

In addition to prosodic features, we attended to features of Malaysian English (MalE) to preserve the interactional meanings through which teaching and learning were accomplished. MalE is characterised by distinct phonological and intonation patterns, localised lexical items and variation in syntactic and pragmatic features [47]. Within the dataset, we identified frequent use of discourse particles such as “lah/la,” “ah,” “what” and “lor.” These particles are well documented as central to spoken interaction in Malaysia and are used to signal politeness or tentativeness, display affiliation, and diffuse tension or deflect potential confrontation [48, 49]. Where such particles shaped meaning (for example, in softening corrective feedback), their function was briefly explained in the analysis to aid interpretation.

It is also common for speakers of MalE to produce utterances with missing auxiliary verbs or subjects in ways that are locally grammatical and pragmatically unproblematic, as well as to code-mix and switch between English and Bahasa Melayu. These features were italicised in the transcripts and accompanied by a Standard English equivalent to retain interactional nuances relevant to teaching and learning while ensuring that the analysis remains accessible to readers unfamiliar with local language practices. For clarity, we used a two-line transcription format, with the Standard English equivalent presented beneath the original.

Analytical procedures

Following the identification of question-answer sequences as central to teaching and learning, we conducted the analysis iteratively across three sequential stages:

Stage 1: development of the coding framework

Firstly, transcripts and audio recordings were reviewed repeatedly to identify question-answer sequences between house officers (HOs) and senior doctors (SDs). We developed a coding framework informed by relevant interactional literature [32, 50–53], beginning with broad operational definitions (e.g., statements, questions, responses) and progressing to specific question designs and sub-designs (see Table 1 for the core scheme and Supplementary Material, Table S1 for the detailed framework).

Coding was undertaken independently by the first author, with a subset of transcripts reviewed by the second and third authors (trained linguists) to ensure analytic consistency. Differences in coding or interpretation were discussed through regular analytic meetings and resolved through consensus, with reference back to the original transcript and, where necessary, the audio recording.

Stage 2: sequential analysis of interactional patterns

Secondly, we performed a close, line-by-line analysis of the transcripts to examine how senior doctors employed various question designs to scaffold house officers’ developing expertise. This stage moved beyond categorisation to examining the sequential and interactional unfolding of question-response exchanges, allowing us to identify how specific designs functioned in context. Based on this analysis, we identified a recurrent pattern of questioning following Socratic principles, leading us to narrow our focus specifically to error-correction sequences. In parallel, we analysed how house officers designed their responses, focusing on how they displayed, negotiated, or recalibrated their knowledge during moments of revision to ensure a balanced view of the interactional accomplishment of learning.

Stage 3: salience mapping and selection of excerpts

Following coding, a salience mapping was undertaken across the dataset to examine the distribution of question designs and identify recurrent patterns within and across contexts. Frequencies were tabulated to determine the extent to which particular practices featured more prominently in CME and ward-round interactions. This systematic comparison informed the selection of excerpts for detailed discussion based on their analytic richness and representativeness, ensuring they accurately reflected the broader patterns of error correction identified across the dataset.

Enhancing rigor and trustworthiness

To further strengthen analytic rigour, instances of divergence and points of contention were examined in detail with reference to the interactional context of the excerpts. These were resolved through discussion with the wider research team involving clinical experts from both the host department and an external specialty. This combination of internal and external clinical perspectives ensured that our analysis remained grounded in the specific departmental context while also reflecting broader clinical practice. By integrating these diverse viewpoints, we provided the necessary contextual grounding to ensure that our linguistic interpretations remained consistent with actual clinical reality until consensus was reached. In sum, several steps were taken to enhance trustworthiness. These included repeated engagement with the data, iterative refinement of the coding framework, collaborative discussion of analytic decisions among the research team and constant checking of claims against the wider dataset to ensure that interpretations remained grounded in the recorded interactions.

Figure 2 summarises the methodological steps followed across the multi-phase analysis, adapted from Creswell and Guetterman’s (2024) framework for analysing and interpreting qualitative data.

Fig. 2.

Visual model of the iterative coding strategy and analytic workflow

The following section presents the findings and highlights the interactional and discursive practices through which clinical expertise was scaffolded, that is, how senior doctors used questions to elicit participation and initiate corrective actions, and how house officers responded to negotiate learning and expertise acquisition.

Results

Across both contexts, senior doctors consistently employed specific question designs grounded in Socratic principles to scaffold house officers’ developing expertise during patient-case talk. These questioning practices enabled house officers to identify inaccuracies, reassess their responses and independently generate corrective actions, positioning self-initiated repair as a central mechanism of the learning process.

Overview of findings

Based on the salience mapping conducted across the dataset, we identified distinct interactional patterns in how error-correction sequences were organised in each teaching context. Table 2 provides a descriptive overview of the distribution of these question designs.

Table 2.

Distribution of questioning practices within error-correction sequences across CME and ward round session

Question Design	CME Sessions (n = X)	WR Sessions (n = Y)
Polar question (declarative, interrogative, turn-final element, reversed polarity,	12 (48%)	8 (23.5%)
Wh-question (content, challenge)	8 (32%)	14 (41.20%)
Designedly incomplete utterances (DIUs)	4 (16%)	12 (35.20%)
Alternative	1 (4%)	0

As shown in Table 2, polar questions accounted for the largest proportion of question designs used in CME error-correction sequences (48%). In ward rounds, however, there was a noticeable shift toward wh-questions (41.2%) and designedly incomplete utterances (35.2%). These distributions indicate context-sensitive differences in how senior doctors design questioning to support error correction: while CME interactions frequently foreground clarification through polar formats, ward round teaching more often mobilises wh-questions and incomplete utterances to constrain the reasoning space and orient house officers toward subsequent inferential steps.

In what follows, we present a detailed analysis of patient-case talk in both CME (Section A) and ward round (Section B) sessions. The excerpts selected are representative of the recurrent patterns identified in the dataset and enable close examination of how these question designs facilitate Socratic dialogue, manage error correction and accomplish pedagogical work moment-by-moment. We begin with Section A by examining these situated practices as they unfold within CME sessions.

A. Scaffolding expertise through elicitation of self-correction in CME sessions

As noted in the overview, error-correction sequences in CME sessions are characterised primarily by the frequent use of polar questions. These designs are often employed in combination with other question designs to prompt HOs to recognise gaps in their initial formulations and refine their presentations accordingly. The following excerpts illustrate how these questions function as tools in practice.

Table 3 (CME-PCT3) demonstrates how SD1 uses polar questions to pursue greater precision in case reporting. While discussing a patient’s management plan, SD1 extends the exchange through a series of polar questions that require HO1 to reevaluate and clarify her initial presentation of the clinical management.

Table 3.

Of the abdomen you mean? (CME-PCT3)

Turn	Speaker	Talk
59	HO1:	so the management of consideration is detailed er: ultraSOUND,
60	SD1:	of the: abdomen you mean?
61	HO1:	of the abdomen or KUB, and then your echo, karyotype, and growth monitoring surveillance
62	SD1:	so A: LL children with a single umbilical artery must have A::LL of these?
63	HO1:	not all, but usually we’ll give the: detailed ultrasound such as like abdomen and KUB, and then we’ll follow up monitoring in the neonatal clinic

In response to HO1’s plan, SD1 formulates a polar question with a turn-final element (60-of the abdomen you mean?) to signal that the proposed management requires further specification. As a known-information question, it holds little instructional value if answered affirmatively (“yes”), indicating that its pedagogical aim is to prompt revision. Through this formulation, SD1 steers HO1 toward producing a more precise and clinically appropriate account than the generalised formulation initially offered.

HO1 responds with a repetitional agreement used as leverage for disagreement. She asserts epistemic authority of her clinical knowledge with a turn-initial repeat (61-of the abdomen), which she then expands into a subsequent disagreement by expanding the list (“or KUB, and then your echo, karyotype, and growth monitoring surveillance”). In doing so, HO1 confirms SD1’s presupposition before clarifying that the plan extends beyond the abdomen, exercising her emerging voice as a novice clinician with developing expertise.

SD1 continues to guide the level of specificity in HO1’s recommendations with a reversed polarity question (62-so all children with a single umbilical artery must have all of these?). The emphasis on “all” conveys a negative assertion that signals the need for revision. This prompts HO1 to reconsider the generalisation implicit in her list and to distinguish between routine and case-specific investigations.

HO1 responds by refining the proposed plan with a negated repetition. She first incorporates negation into SD1’s prior turn (63-not all…) before extending the response by specifying the investigations that are typically performed when managing infants with a single umbilical artery (“detailed ultrasound such as abdomen and KUB” and “neonatal clinic monitoring”). Through this self-initiated repair, HO1 not only clarifies the management pathway but also demonstrates an increasing level of independent management reasoning, showing how SD1’s questioning prompts her to refine an initial response into a more contextually appropriate formulation.

A similar pattern is evident in Table 4 (CME-PCT11), in which SD1 uses polar questions to prompt greater precision and elicit a more accurate clinical descriptor of the patient’s signs and symptoms.

Table 4.

Bilateral high stepping gait (CME-PCT11)

Turn	Speaker	Talk
43	HO1:	so it’s a high stepping gait
44	HO1:	and > further examination< also [we]
45	SD1:	[so ] is it both legs, one leg?
46	HO1:	bi- bilateral (.) bilateral high stepping gait

In this exchange, SD1 interrupts HO1’s presentation with a polar interrogative followed by a polar declarative (45-so is it both legs, one leg?). While these questions appear to seek confirmation, they function to highlight a missing clinical specification in HO1’s prior turn. By embedding the required detail within the alternatives provided, SD1 invites HO1 to supply the correction independently, initiating repair without overt criticism. The emphasis placed on the first alternative (“both legs”) further draws attention to it as a candidate correction, signalling overall aim of promoting accuracy while maintaining interactional support for the learner.

HO1’s transformative response in line 46 resists the question constraints yet displays alignment with SD1’s mitigated criticism. Instead of offering a simple confirmation (“both”), she uses the clinically appropriate terminology (“bilateral”) indicating her interpretation of SD1’s intervention as a prompt to refine the precision of her earlier formulation rather than providing additional case details. In doing so, she orients towards facilitating agreement with the question posed but also demonstrates agency as a learner and exercises judgment of her own.

Besides that, polar questions are also observed to be paired with designedly incomplete utterances (DIUs) to facilitate real-time correction during CME patient-case talk. An illustrative example is drawn from CME-PCT20, presented in Table 5, where two senior doctors (SD1 and SD2) guide HO2 in describing a brain magnetic resonance imaging (MRI) finding.

Table 5.

Hyperintense? (CME-PCT20)

Turn	Speaker	Talk
29	SD2:	how to describe it, in imaging TE: RMS how do you describe it, in imaging TERMS
30	HO2:	oh erm (.) it’s a hyperintense lesion,
31	SD2:	hyper[intense? ]
32	SD1:	[HYPER? ]
33	HO2:	(.) sorry hypointense (.) ah hypointense lesion seen at the:: suprasellar region
34	SD2:	°alri: ght°

Following SD2’s prompt for imaging terminology, HO2 initially describes the finding as a “hyperintense lesion” (turn 30). This immediately triggers a corrective intervention in the form of SD2’s reversed polarity question in line 31 (hyperintense?). Using selective repetition, syllabic emphasis on the prefix “hyper” and rising intonation, SD2 marks this segment as a trouble source. Rather than directly supplying the correction, this question momentarily pauses the progression of the patient-case talk and redirects HO2’s attention to the accuracy of the descriptor.

SD1 then overlaps with a designedly incomplete utterance (32-hyper?), narrowing the locus of trouble to the prefix itself and further scaffolding the opportunity for self-correction. Following a micropause, HO2 revises the term to the accurate imaging descriptor (33-sorry, hypointense), explicitly marking the correction with emphasis on the prefix “hypo”. SD2’s minimal acknowledgement (34-alright) accepts the revised description and concludes the repair sequence without extending the exchange.

These examples show that senior doctors strategically design questions not merely to evaluate house officers’ knowledge, but to create interactional conditions that promote learner-led correction. Pedagogically, the use of these question formats enables SDs to scaffold HOs’ developing expertise in a way that prioritises learning over the public exposure of deficiencies. When errors occur during CME patient-case talk, questions are finely calibrated to be just sufficient to stimulate reflection and invite revision, rather than to impose overt evaluation, contributing to a psychologically safe learning environment. In doing so, this approach also allows HOs to remain active epistemic agents within the exchange.

However, the same indirectness can also reduce its instructional effectiveness. As shown Table 6, when HOs do not recognise the corrective intent embedded in a polar question, the opportunity for repair may be missed, necessitating a shift to more explicit instructional intervention.

Table 6.

Why suddenly became consolidation? (CME-PCT6)

Turn	Speaker	Talk
47	HO3:	so, > from the x-ray we can see < tha::t the: there is widening of the mediastinum, (.) and then there’s some consolidation a: hh over bilateral lung fields
48	SD1:	yes ah?
49	SD1:	why suddenly became consolidation? ((why did it suddenly become consolidation? ))*
50	SD1:	WHEre is the consolidation?
51	HO3:	[right side la, MORE on the right side]
52	SD1:	[so is it CONSOLIDATION or is it ] mediastinum widening, which one now?
53	HO3:	ah just widening: of mediastinum
54	SD1:	so there’s WIDENING of the mediastinum

*Note-((standard English version))

To draw attention to the inaccuracy in HO3’s description of a patient’s X-ray, SD1 begins with a reversed polarity question (48-yes ah?), formatted as a no-preferring question, to signal potential disagreement without making the correction explicit. SD1 then further problematises it using a wh-question used as challenge (49-why suddenly became consolidation), with emphasis placed on the word suddenly. This formulation directly targets the problematic aspect of HO3’s description by implying that there is no clinical basis for the claimed radiological finding. Such wh-questions are interactionally designed to project negative assertions (“no reason”), indirectly conveying that the feature described by HO3 lacks evidential support.

In line 50, SD1 continues without the pause typically associated with a transition relevance place and issues another wh-question to challenge (50-where is the consolidation?). This rapid continuation further strengthens the corrective stance and makes explicit that the finding is, in fact, absent. Together, these question designs display SD1’s orientation to the turns as repair-initiating and corrective rather than information-seeking, effectively conveying that the originally reported finding cannot be substantiated on the x-ray.

However, HO3’s response indicates that she oriented to SD1’s question as information-seeking rather than corrective (51-right side la, more on the right side), elaborating on location instead of revising the accuracy of the finding. When the indirect repair initiations fail to prompt self-correction, SD1 shifts to a more explicit approach by posing an alternative question (52-so is it consolidation or is it mediastinum widening?), requiring HO3 to select between two diagnostic possibilities drawn from her prior description.

This forced-choice format successfully elicits the intended correction, with HO3 selecting the accurate option. This indicates that the corrective intent of the earlier indirect questions only became clear to HO3 once she is explicitly required to select between two competing interpretations, one of which is implicitly marked as problematic. The louder delivery of the term “consolidation” further appears to have directed HO3’s attention to this element as the target of correction. SD1’s subsequent evaluative confirmation of the response (54-so there’s widening of the mediastinum) marks the answer as accurate and closes the repair sequence.

Overall, these analyses demonstrate that through the strategic use of polar questions and their varied combinations, senior doctors systematically design interactional spaces that invite HOs to engage in self-correction within error-repair sequences during CME patient-case talk. This questioning approach supports reflection and critical thinking without direct answer provision, reducing reliance on overly didactic instruction. In doing so, it preserves HOs’ epistemic autonomy and supports the exercise of an emerging clinical voice, even at an early stage of training. The indirect nature of these corrective practices also serves as a form of mitigated criticism, making correction face-saving and psychologically safe for HOs navigating high-stakes spaces where uncertainty and knowledge display are both publicly accountable.

We now turn to Section B to examine how Socratic questioning operates differently in ward rounds (WRs).

B. Scaffolding Expertise through Elicitation of Self-Correction in Ward Rounds

In contrast to the interactional patterns observed in CME sessions, error correction during ward rounds exhibits a shift toward more open-ended and collaborative question designs. As shown in Table 2, wh-questions (41.2%) and designedly incomplete utterances (35.2%) are the most frequent question designs observed in these encounters. These designs appear to be adapted to the situated nature of bedside teaching, where senior doctors use DIUs and wh-questions to ‘hand over’ the inferential work to the house officer. Furthermore, wh-questions in this context are used both to invite knowledge display and to challenge learner formulations, with the overarching aim of fostering clinical reasoning within routine patient management.

Tables 7, 8, 9 and 10 demonstrate how MO1 mobilises these question formats to strategically scaffold expertise during a ward round. Although drawn from a single interactional episode, the sequence is presented in segmented form to enhance clarity and readability.

Table 7.

We didn’t give any antiemetic, did we? (WR-PCT17)

Turn	Speaker	Talk
153	MO1:	why patient don’t vomit after admitted? ((why did the patient stop vomiting after admission? ))*
154	HO1:	(0.2)
155	MO1:	we didn’t give any:: antiemetic what. ((we didn’t give any antiemetic, did we? ))*
156	HO1:	(0.2) well rest

*Note-((standard English version))

Table 8.

What is the problem with dengue? (WR-PCT17)

Turn	Speaker	Talk
157	MO1:	so when you ge: t dengue, what > what is the problem< with dengue?
158	HO1:	increased muscular permeability?
159	MO1:	so, intravascular volume?
161	HO1:	reduced
162	MO1:	so, you::r human body: will try to preserve which organ?
163	HO1:	(0.2) brain, kidney
164	MO1:	kidney [dies   ], it dies.
165	HO1:	[heart? ]
166	MO1:	bra: in and heart
167	HO1:	mm
168	MO1:	these two organs they prioritise, others if they die, never mind ((they prioritise these two organs, if the other organs die, it doesn’t really matter))*
169	HO1:	mm
170	MO1:	so, > that’s why < you get a bit of AKI,
171	HO1:	mm

*Note-((standard English version))

Table 9.

Then you get? (WR-PCT17)

Turn	Speaker	Talk
172	MO1:	then:, you get?
173	HO1:	(0.2)
174	MO1:	hypoperfusion of your?
175	HO1:	(.) renal, kidney?
176	MO1:	°gut lah°*
177	HO1:	oh gut lor*

*Note-lah and lor are discourse particles common in Malaysian English

Table 10.

So? (WR-PCT17)

Turn	Speaker	Talk
178	HO1:	oh oh: h oh::h oh (.) so (.) gut become not well perfused? ((oh so the gut became not well perfused? ))*
179	MO1:	mm
180	HO1:	So the function [is (.)] depressed
181	MO1:	[mm: ]
182	MO1:	so you (.) vomit
183	HO1:	mm > that’s why< cannot cannot reabsorb water:? ((mm that’s why it cannot reabsorb water? ))*
184	MO1:	↑yes
185	MO1:	so now, you (.) restore the perfusion
186	HO1:	Mm
187	MO1:	so?
188	HO1:	the blood circula- er I mean like the perfusion of the gut is impro-vi-ng,
189	MO1:	↑yes
190	HO1:	so they start functioning °again°
191	MO1:	Yes

*Note-((standard English version))

To draw HO1 into causal clinical reasoning (see Table 7), MO1 turns the patient’s physiological response following admission into a learning opportunity by initiating a wh-content question (153-why patient don't vomit after admitted?). When HO1 shows difficulty formulating an explanation (2-second pause, see turn 154), MO1 follows up with a tag question (155-we didn’t give any antiemetic what?), steering the discussion toward a more constrained explanatory trajectory.

Although formatted as a tag question, the utterance functions interactionally as a statement [54], designed to prompt HO1’s reasoning. It asserts a key proposition (that no pharmacological intervention was given) while mitigating its assertive force through hedging [55] using the invariant tag “what” [56]. This framing treats HO1 as already aware that no antiemetic was administered, projecting this as shared information that can be used for reasoning. By ruling out medication as a causal factor of the patient’s improvement, MO1 strategically narrows the explanatory space and redirects HO1 to reconsider the original question: if vomiting ceased in the absence of pharmacological intervention, what alternative mechanisms may explain the patient’s improvement?

HO1 responds with a brief explanation (156-well rest), following 2-second pauses that suggest continued uncertainty. In response to that, MO1 initiates a self-correction trajectory in subsequent turns, shifting the interaction from a surface-level explanation toward deeper physiological reasoning, as shown in the excerpts that follow.

In the sequence presented in Table 8, MO1 follows up with a wh-content question (157-so when you get dengue, what is the problem with dengue?), which functions to launch a targeted line of causal reasoning concerning the patient’s improvement in the absence of antiemetic treatment (a medication that prevents and treats nausea and vomiting). HO1 responds with a tentative phrasal explanation (158-increased muscular permeability?) delivered with rising intonation, indicating uncertainty and a search for confirmation. MO1 then advances the reasoning trajectory using a designedly incomplete utterance (159-so intravascular volume?), which both affirms the direction of HO1’s thinking and narrows the focus to the next relevant physiological parameter. This move scaffolds HO1’s reasoning by prompting him to connect increased vascular permeability to its physiological consequence (reduced intravascular volume), thereby requiring retrieval and application of relevant knowledge. HO1 responds minimally (161-reduced), completing the implied causal link.

MO1 continues the reasoning chain with a further wh-content question (162-so your human body will try to preserve which organ?), building on HO1’s prior response. HO1 initially names two organs (163-brain, kidney), following a 2-second pause that suggests uncertainty. MO1 immediately signals the inaccuracy of this formulation (164-kidney dies, it dies), leading HO1 to offer a revised response (165-heart?) delivered with rising intonation and inviting MO1’s assessment. MO1 then confirms the accurate prioritisation by restating the preserved organs (166-brain and heart). The reasoning chain is subsequently completed when MO1 links this physiological prioritisation to the development of acute kidney injury (AKI), explicitly articulating the causal consequence (170- so that's why you get a bit of AKI). HO1’s minimal acknowledgements across the following turns (see 167, 169 and 170) indicate that the explanation was understood.

As the extract in Table 9 shows, MO1 continues prompting learner reasoning with another designedly incomplete utterance (172-then you get?), inviting HO1 to extend the causal chain. The absence of an immediate response (a 2-second pause in line 173) indicates difficulty in generating the next step in the explanation. To facilitate progress, MO1 produces a second designedly incomplete utterance as a syntactically fitted continuation (174- hypoperfusion of your?), narrowing the focus toward reduced or inadequate blood flow to an organ or tissue as the next relevant pathophysiological process.

Following a micropause, HO1 produces a tentative formulation (175-renal, kidney?) with the rising intonation indicating uncertainty of the affected system. This prompts MO1 to perform a corrective action (176-gut lah), delivered softly, redirecting the focus toward the gastrointestinal consequences of reduced perfusion. The discourse particle “lah” mitigates this correction by reducing social distance, allowing more accurate information to be introduced supportively. HO1 responds with a minimal acknowledgement (177-oh gut lor), accompanied by the particle “lor,” to index the information as retrospectively obvious.

Subsequently, HO1 demonstrates developing understanding by integrating reduced gut perfusion into the evolving clinical reasoning sequence (see Table 10). Through a tentative formulation (178-oh oh oh oh so gut become not well perfused?), HO1 displays emerging causal understanding of hypoperfusion, which is then developed into a functional inference that inadequate blood flow leads to impaired gastrointestinal function (see turn 180). These inferential steps are supported by MO1’s minimal confirmations (turns 179 and 181), validating the reasoning trajectory while preserving HO1’s agency and active participation.

MO1 then brings the discussion back to the patient’s clinical presentation by reintroducing vomiting (182-so you vomit). This reconnects the focus on gut perfusion with the original presenting symptom, reinforcing the causal pathway and ensuring that HO1 recognises the relevance of the reasoning process to the original clinical problem. HO1 advances the reasoning chain by linking the compromised perfusion to a specific functional consequence (183-mm that’s why cannot cannot reabsorb water?). In doing so, he extends the explanation from pathophysiological mechanism (hypoperfusion) to functional impairment (reduced absorptive capacity), demonstrating an emerging ability to apply biomedical knowledge in anticipating physiological outcomes. The rising intonation marks this contribution as tentative, indicating that confirmation is still being sought.

MO1 minimally confirms (185- so now you restore the perfusion) before inviting further participation from HO1 through another DIU (186-so?), directing HO1 to consider the treatment effects. HO1 responds by articulating the expected recovery trajectory (188-the perfusion of the gut is improving), and linking this to the return of normal function (190-so they start functioning again). Through this reasoning sequence, HO1 demonstrates understanding of the patient’s vomiting resolving after admission without the use of antiemetics (see line 153).

Besides DIUs and wh-content questions, MO1 also used wh-challenge questions during WRs. In Table 11 (WR-PCT20), this strategy occurs at the patient’s bedside during caregiver education on the management of future seizure episodes. Upon explaining that patients are recommended to be positioned in the left lateral position during a seizure, MO1 redirects the explanation into a learning opportunity for HO1 by posing a wh-content question (673-why usually we ask left lateral?).

Table 11.

The tongue is in the centre, isn’t it? (WR-PCT20)

Turn	Speaker	Talk
673	MO1:	why usua::lly we ask (.) left lateral? ((why do we usually ask carers to place the patient in the left lateral position))*
674	HO1:	mm::
675	MO1:	why not right lateral?
676	HO1:	the position of the tongue?
677	MO1:	the tongue centre what. ((the tongue is in the centre, isn’t it? ))*
678	HO1:	mm
679	MO1:	your tongue is centre what. ((your tongue is in the centre, isn’t it? ))*
680	MO1:	so kiri kanan takde masalah what. ((so there is no issue with left or right position, is there? ))*
681	HO1:	gastric
682	MO1:	Ha: [::h ]**
683	HO1:	> [stomach] stomach<
684	MO1:	stoma::ch
685	HO1:	stomach reflux

*Note-((standard English version))

**minimal agreement token

When no immediate response is produced, MO1 follows up with a second wh-content question (675-why not right lateral?). By invoking a negatively framed alternative, this follow-up implicitly marks the right lateral position as non-preferred and foregrounds a contrast that invites inferential reasoning. Through this design, MO1 embeds the clinical rationale within the questioning sequence itself, prompting HO1 to reason through the underlying considerations informing standard seizure management practices rather than merely reproducing procedural knowledge.

In response, HO1 produces a phrasal answer (676-the position of the tongue?), delivered with rising intonation, displaying uncertainty and inviting MO1’s confirmation. MO1 responds by launching a sequence of tag questions (677-the tongue centre what?; 679-your tongue is centre what?; 680-there is no issue on the left or right, what?), produced in rapid succession. The absence of transition silences between the turns indicates that these questions are not oriented toward soliciting new information, but rather to problematise HO1’s prior claim and stimulate deeper reasoning. Across this sequence, MO1 repeatedly uses the invariant tag “what” [56], a form commonly used by speakers of Malaysian English. As shown earlier (Table 7, turn 155), “what” allows assertions to be formatted as questions, softening delivery and mitigating overt correction by treating HO1 as already aware that the tongue is centred and that it should not determine whether a patient is lifted from the left or right. By using these negatively framed contrasts, MO1 foregrounds the inadequacy of tongue position as a justification for left lateral positioning and redirects HO1 toward alternative explanations for the recommended seizure management.

Following this guided redirection of reasoning, HO1 offers a revised answer (681-gastric), which MO1 acknowledges with a smile-voiced “hah.” This prompts HO1 to refine his response to “stomach” repeated twice with accelerated delivery, indicating increased certainty. MO1 then confirms the revised explanation (684-stomach), marking it as correct before closing the sequence. HO1 further reiterates (685-stomach reflux) to better support his answer by showing his understanding that the underlying reason informing that management strategy is that the left, rather than the right, lateral position reduces gastro-oesophageal reflux by keeping gastric contents away from the oesophageal junction, which in turn minimises the risk of it going into the airway. In doing so, HO1 demonstrates the ability to justify procedural advice by attending to its underlying physiological rationale.

In sum, these analyses demonstrate that by systematically deploying wh-questions and designedly incomplete utterances, SDs construct a tightly scaffolded reasoning environment that keeps the learner in a state of guided inquiry. These formats serve distinct but complementary pedagogical functions: highlighting gaps, narrowing the reasoning space and directing HOs toward the next inferential step without supplying answers prematurely. Through this structured sequencing of questions grounded in Socratic principles, SDs maintain HOs in active reasoning, prompting retrieval and application of relevant knowledge while enabling self-correction and ensuring that understanding is generated rather than given. Collectively, these practices demonstrate how questioning operates as a core pedagogical mechanism in WR patient-case talk, enabling HOs to refine their reasoning processes and arrive at sound clinical judgements through their own efforts.

Discussion

This study demonstrates that questioning serves as a key pedagogical resource through which senior doctors (SDs) scaffold house officers’ (HOs) developing expertise during Continuing Medical Education (CME) sessions and ward rounds (WR). Across both settings, questioning was systematically designed to promote self-correction and active reasoning. By mobilising different question designs, senior doctors (SDs) guided house officers (HOs) to identify knowledge gaps, advance clinical reasoning and refine their formulations without direct answer provision. A recurrent feature of this practice was the stance of professed ignorance, which positioned senior doctors (SDs) as co-inquirers while still steering the trajectory of reasoning. By detailing how these questioning practices draw on Socratic principles and unfold in real-world training contexts, the study offers empirical insight into a form of clinical teaching widely recognised by clinicians and educators yet rarely described in detail in medical education research.

The findings also show that questioning practices were adapted to the demands of different learning contexts. In CME sessions, questioning worked to sharpen the precision (see Tables 3 and 4) and accuracy (see Tables 5 and 6) of formulations, with polar questions, wh-questions and designedly incomplete utterances creating structured opportunities for self-repair. During ward rounds, questioning supported causal reasoning amid ongoing patient care (see Tables 7, 8, 9, 10 and 11), helping learners narrow the problem space and retrieve relevant knowledge using wh-questions and designedly incomplete utterances. Despite these contextual differences, a consistent pedagogical function remained: questioning provided corrective guidance while sustaining reflective engagement and active learner involvement. By keeping HOs engaged in “working out the case,” SDs balanced instructional support with the interactional space needed to display more independent clinical reasoning.

Although these questioning practices ultimately aimed at accurate formulations, its primary pedagogical function was formative rather than evaluative. By probing HOs’ thinking and surfacing both existing knowledge and gaps, SDs were able to determine where HOs were situated along a developmental continuum and calibrate support accordingly. Across excerpts, inaccuracies were treated as learning opportunities: instead of issuing evaluative judgments, SDs produced follow-up questions that enabled refinement without punitive consequence.

These findings align with and extend existing work on the use of Socratic questioning in medical education. Consistent with prior studies, the analysis demonstrates that Socratic-oriented questioning can be adapted across both classroom-based and workplace-based learning contexts [12]. The present study adds interactional detail by showing how SDs scaffold learning through questions that surface gaps in knowledge and provide targeted support as HOs progress toward more advanced understanding [19]. In particular, educators were observed using a range of question designs to model clinical reasoning, guide learners through their thinking processes and prompt them to verbalise existing knowledge (see Tables 8, 9, 10) [57].

The findings further contribute to ongoing efforts to distinguish Socratic teaching from “pimping.” Prior work suggests that this distinction rests on both the questioner’s intent and the learner’s perception of that intent [15, 16]. While research indicates that intent is conveyed through the manner of discourse [24], the present study offers empirical detail on how such intent is interactionally signalled through the design, sequencing and delivery of questions. Under these conditions, interactional features that suggest a climate of psychological safety [15, 18] were observable: learners locally displayed uncertainty (see turns 158 and 165 in Table 8; turn 175 in Table 9), were provided with the interactional space to articulate existing knowledge and build conceptual connections (see Tables 8, 9, 10) and received corrective feedback designed to identify gaps without the observable manifestations of face-threat or humiliation (see turn 164 in Table 8; turn 176 in Table 9). These observations align with recommendations to create learning environments in which learners are challenged and supported simultaneously [11, 14, 15].

A related implication concerns how medical educators might move beyond traditional Initiate-Respond-Evaluate (IRE) patterns [19]. Rather than issuing direct evaluative judgements, senior doctors (SDs) used third-turn responses in the form of polar questions, wh-questions and designedly incomplete utterances to prompt learners to reconsider and respond at higher-order thinking levels (see Tables 3, 8 and 11). These designs not only served as constructive alternatives to evaluation but also softened the face-threatening potential of correction, allowing inaccuracies to surface in a mitigated rather than punitive manner. Corrective feedback thus operated as a cognitive prompt rather than a final judgement of error, aligning the exposure of gaps with interactional displays associated with psychological safety. However, the interpretation of such corrective practices may vary across contexts. In the present Malaysian setting, learners routinely orient to these questions as invitations to refine their answers, suggesting that the perceived softness of corrective moves rests not only on design but also on the interactional management of cultural norms within the training environment.

Beyond this, the analysis also highlights the pedagogical value of wait time between questions and responses. While optimal durations remain debated [58–60], the present findings show how strategically maintained silence within supportive questioning sequences can facilitate engagement by allowing learners time to retrieve knowledge and reorganise their reasoning (see turn 156 in Table 7; turn 163 in Table 8; turns 173 and 175 in Table 9). This aligns with what is recognised as a best practice in the medical education literature [11, 19], while providing interactional evidence of its effects in practice.

Our findings offer practical insights for medical educators, particularly in utilising questioning techniques to facilitate learning. Rather than relying on single, isolated questions or direct correction, senior doctors can structure sequences of incrementally focused questions that guide learners toward more precise and accurate formulations. Different question designs can be deployed strategically: for example, polar questions to check or challenge assumptions, wh-questions to probe underlying reasoning and incomplete utterances to prompt learner completion and active engagement.

Besides, the analysis shows that effective questioning is inherently responsive and learner-centred. The design of each subsequent question is shaped by how learners respond: whether they display uncertainty, partial understanding or emerging insight. This highlights that questioning in clinical teaching is not a fixed, uniform technique, but an interactional practice that unfolds moment-by-moment. As such, effective use of questioning depends on educators’ ability to attend closely to learners’ contributions and adapt their questioning in real time.

By making visible how questioning is sequentially organised and contingently shaped by learner responses, this study provides a more fine-grained understanding of how clinical teaching is accomplished in practice. This interactional perspective offers educators a practical framework for using questioning not simply to test knowledge, but to actively guide learners’ participation, support revision and sustain engagement during patient-case talk. To translate these findings into practice, such insights could be integrated into faculty development modules. By using authentic examples from real-world training grounds, they could offer an impactful training framework, helping educators recognise and deploy question designs more effectively in real time.

Finally, the study builds upon existing models of Socratic questioning by bridging the gap between educational theory and real-world clinical talk. While current frameworks describe Socratic questions by their reasoning dimensions, forms of dialogue and pedagogical purposes, they often do not specify the actual question formats that can be used to achieve these goals in practice. Our study addresses this gap by mapping specific question designs onto both their reasoning goals (such as precision, accuracy, logic) and their pedagogical functions (such as probing, clarifying, challenging) (see Table 12). By identifying these recurring patterns, the study offers a practical guide that connects abstract Socratic principles to the everyday reality of clinical teaching, This not only advances our conceptual understanding of Socratic pedagogy but also offers medical educators a clear, evidence-based toolkit for using these techniques effectively in clinical teaching.

Table 12.

Socratic question designs and their associated reasoning targets and pedagogical functions in clinical teaching

No.	Targeted Reasoning Dimensions (Paul & Elder, 2019; Oyler & Romanelli, 2014)	Socratic Question Type (Hu, 2023; Mason, 2011; Paul, 1990)	Question Sample	Question Design & Subdesign
1.	Relevance	Questions that challenge assumptions	your tongue is centre what? ((your tongue is in the centre, isn’t it? ))*	Polar question (invariant tag “what”)
			so left right no problem what? ((so there is no issue with left or right position, is there? ))*
2.	Logic		why did it suddenly become consolidation?	Wh-question (challenge)
			where is the consolidation?
3.	Accuracy	Questions that clarify	hyper?	DIU
			yes ah?	Polar question (reversed polarity)
			hyperintense?
4.	Precision		so all children with a single umbilical artery must have all of these?
			of the abdomen you mean?	Polar question (turn-final element)
			so is it both legs, one leg?	Polar question (interrogative, declarative)
5.	Breadth	Questions that explore implications and consequences	when you get dengue, what is the problem with dengue?	Wh-question (content elicitation)
			your human body will try to preserve which organ?
6.	Depth	Questions that examine evidence or reasons	so, intravascular volume?	DIU
			then you get?
			hypoperfusion of your?

*Note-((standard English version))

Limitations

However, several limitations must be considered when interpreting the findings. Firstly, the dataset was drawn from a single paediatric ward and biweekly CME sessions collected over three months, which limits the range of interactional contexts represented. Therefore, the patterns identified may not reflect the diversity of practices across other departments or institutions and limit broad generalisability, although they offer transferable insights for comparable settings.

Secondly, the study did not systematically account for participants’ demographic and professional backgrounds, such as years of clinical experience or formal teaching training. These factors may shape individual questioning styles, interactional positioning and the management of authority and support during teaching encounters. Furthermore, the voluntary nature of participation introduces the possibility of self-selection bias, whereby clinicians who are more engaged in teaching or more experienced may have been overrepresented, potentially limiting the range of interactional styles captured. As such, the findings should be interpreted as reflecting situated interactional practices rather than universal norms.

Thirdly, the analysis focused exclusively on verbal interactions. By excluding multimodal features such as gaze, gesture and spatial positioning, the study may present a partial or potentially skewed understanding, not fully capturing the embodied complexity of clinical interaction.

Finally, a key methodological consideration is the observer’s paradox. Despite efforts to minimise disruption, the presence of recording devices may have subtly influenced participants’ behaviour toward more formal or ‘idealised’ interactional performances. While we employed prolonged engagement with the site to acclimate participants and capture interactions as close to routine practice as possible, this remains an inherent challenge in ethnographic and discourse-analytic research. Therefore, it is an important factor when assessing how far the data represent the “unobserved” reality of everyday talk.

Conclusion

This study contributes to the understanding of Socratic pedagogy by analysing how questioning practices unfold in routine educator-learner interactions during the postgraduate medical education of HOs or interns. The analysis demonstrates that Socratic questioning functions as a vital pedagogical resource in supporting HOs’ developing expertise, particularly through practices that elicit learner contributions and facilitate self-initiated repair. Senior doctors use a range of question designs including polar questions, wh-questions and designedly incomplete utterances to prompt learners to engage in higher-order thinking, systematically work through their clinical considerations and revise their interpretations as required. These strategies help SDs balance supervision with opportunities for learner agency, ensuring HOs receive timely guidance while being encouraged to exercise greater agency over their clinical claims. When enacted in line with best practices, such questioning makes knowledge gaps visible through mitigated interactional moves that surface areas for development without the stigma of deficiency. Collectively, the findings highlight the pedagogical value of structured, dialogic questioning as an evidence-informed approach for strengthening clinical learning and supporting the transition of early-career doctors navigating complex workplace learning environments.

Transcription symbols

[ ] : Overlapping talk

(.) : Discernable silence but less than 0.2 of second

(0.3) : Silence within turns or in talk

. : Closing intonation

, : Slightly rising intonation

? : Rising intonation

wo: rd : Elongation of preceding sound

word : Emphasis

WORD : Spoken more loudly

°word° : Spoken more softly

↑ ↓ : Marked increase/decrease in pitch

> word< : Talk is speeded up

((word)): Transcriber’s notes

£word£ : Smiley voice or suppressed laughter

Abbreviations

HO

House Officer

SD

Senior Doctor

CME

Continuing Medical Education

WR

Ward Round

Authors’ contributions

AyS conceptualised and designed the study. AzS conducted the analysis and wrote the first manuscript draft. AAK and MFMT facilitated data collection. AyS, PSA, JPE and AAK reviewed and edited the draft. AyS, PSA, JPE, AAK and MFMT contributed to validation of the analysis. All authors read and approved the final manuscript.

Funding

This research was funded by the Ministry of Higher Education Malaysia through the Fundamental Research Grant Scheme with the Project Code: FRGS/1/2022/SSI09/UM/02/1.

Data availability

The dataset supporting the conclusions of this article are included within the article.

Declarations

Ethics approval and consent to participate

All procedures were conducted in accordance with the ethical standards of the Declaration of Helsinki and approved by the Universiti Malaya Medical Centre-Medical Research Ethics Committee (ID: 202317 − 11936). Written and verbal informed consent was obtained prior to recording from all individuals involved in the interactions, including house officers, senior doctors and caregivers who were present during data collection. Participants were informed of the study in advance, advised that participation was voluntary and that they could withdraw at any time. Identifiable information was accessible only to members of the linguistic research team via password-protected storage, and all textual and audio data were anonymised prior to analysis and wider team access.

Consent for publication

Not Applicable.

Competing interests

AAK is the head of the department in which the study was conducted. However, appropriate measures were taken to ensure analytic independence and this role did not influence the study design, analysis or interpretation of the findings.