Conversations and interviews are not simply monologues with two voices. Their meaning is distributed across turns, reactions, interruptions, questions, corrections, shared assumptions and changes of position. A listener can understand every sentence and still lose the examination mark by attaching an idea to the wrong speaker, treating politeness as agreement, missing why a question was asked, or failing to notice that a view changed after new evidence.
This guide is SETC-SEC-LISTEN-090, a specialist audio-architecture spoke inside the Secondary English Listening Examination Performance lane. It does not replace the general listening system, the question-strategy owner, the note-taking owner or narrow conversational mechanisms elsewhere in the eduKate ecosystem. Its canonical job is narrower and deeper: how a secondary learner performs when meaning is jointly built by two or more voices under examination conditions.
The general SETC listening foundation remains English Listening Performance. The question-control layer remains Listening Comprehension Test Strategies. Detailed note mechanics remain in Listening Exam Note-Taking. For the narrow mechanics of how speakers enter, hold and yield the conversational floor, use eduKateSG’s How English Works | Conversational Floor Management. This page integrates those dependencies specifically around conversation and interview examination performance.
Maren reads the architecture of a conversation: proposal, objection, repair, agreement, decision. Iona tracks who owns each claim, which evidence supports it and how strongly a speaker commits. Leonie protects timing and re-entry when rapid turn changes make one missed phrase especially costly.
IDENTIFY THE VOICES → TRACK THE TURN JOB → BIND CLAIMS TO SPEAKERS → NOTICE RESPONSE RELATIONSHIPS → UPDATE STANCE → PRESERVE DECISIONS → ANSWER FROM THE FINAL SPEAKER MODEL → RESET
Part I | Why conversations create a different listening problem
1. Meaning is distributed across people
In a monologue, one speaker usually owns the whole discourse even when several viewpoints are reported. In a conversation, source changes repeatedly. The same noun can be repeated by different people for opposite purposes. One speaker may raise an idea, another challenge it, and a third accept only part of it. The listener therefore needs a live source map. “Cost was mentioned” is not enough. The useful representation is closer to “Speaker A worries about cost; Speaker B accepts the cost but questions timing.” This source binding is part of comprehension, not bookkeeping added afterward.
2. Turn order is not argument order
Conversation alternates voices, but the reasoning may cross several turns. A asks a question, B answers partly, A clarifies the question, and B gives the decisive response. The correct answer can depend on a relationship built across this four-turn sequence. A learner who treats each turn as independent sentences may remember local facts without understanding what the exchange accomplished. Training should therefore track conversational moves—question, answer, challenge, repair, concession, decision—rather than only vocabulary.
3. Short turns can carry high information value
A one-word response such as “Exactly”, “Not quite”, “Actually” or “Maybe” can change the status of the previous proposition. The acoustic duration is small but the discourse effect is large. Examination questions exploit this because learners often devote attention according to sentence length. Conversation listening requires semantic weighting: brief turns can confirm, reject, soften, redirect or reopen an issue. The listener should ask what each short response does to the shared model.
4. Politeness can hide disagreement
Real speakers often disagree indirectly. “I see why you’d think that” may be followed by a counterargument. “That’s one option” can signal weak commitment rather than endorsement. A listener who classifies attitude from polite vocabulary alone can miss the proposition. The safe method is to listen through the entire response and distinguish social acknowledgement from intellectual agreement. Tone can support the interpretation, but the wording and subsequent reasoning remain essential.
5. Conversation contains unfinished language
Speakers hesitate, restart, abandon phrases and complete one another’s ideas. Examination recordings vary in naturalness, but even carefully produced dialogue may include self-repair and interrupted syntax. A listener should not treat an unfinished fragment as a final claim. Wait for proposition completion. If a speaker begins “I thought the best option was—well, after seeing the figures…” the initial direction is provisional. The final commitment emerges later.
6. Shared knowledge creates omission
Conversational speakers can omit information because both already know the context. Pronouns, ellipsis and shorthand references therefore become more common. “That won’t work” requires the listener to know what “that” refers to. “Neither did I” depends on a prior proposition. The listener must maintain an entity and proposition map across turns, not simply decode each sentence in isolation.
7. Response meaning depends on the previous turn
“Probably” can answer a factual prediction, a suggestion or an invitation. “I’d rather not” can express preference, refusal or caution depending on the question before it. Conversation listening is relational: the function of one utterance is partly determined by what it responds to. Review should therefore include the neighbouring turn whenever a response seems ambiguous.
8. The listener must track alignment and distance
Speakers continually align with or distance themselves from prior claims. They agree, partly agree, concede, reject, correct, qualify or defer. These moves create the social and argumentative structure of the conversation. A strong listener hears not only “what Speaker B says” but “how Speaker B positions this statement relative to Speaker A.” That relation often determines attitude and purpose questions.
9. Decisions emerge from negotiation
Many dialogue tasks contain options, objections and final decisions. The first suggestion can be memorable but obsolete. A later compromise can differ from every original proposal. Training should represent negotiation as a state machine: proposal → response → constraint → revised proposal → agreement. The answer may target the final decision, the reason for change or the speaker who introduced the decisive constraint.
10. Conversation speed is partly turn density
A dialogue can feel faster than a monologue even at the same words per minute because the listener must repeatedly reset source, intention and relationship. The cognitive cost comes from switching, not only acoustic pace. When a learner says “the conversation was too fast,” diagnosis should check whether speaker transitions and source binding were the real bottleneck.
11. Interviews add question architecture
An interview is a conversation with asymmetric roles. The interviewer frames topics, requests clarification, challenges assumptions or invites elaboration. The interviewee supplies content, perspective and evidence. Meaning can depend on why a question was asked, not merely the literal answer. The listener benefits from tracking interviewer function: opening, narrowing, probing, challenging, redirecting, summarising or closing.
12. Interview answers often move from direct response to story
An interviewee may answer briefly, then give an anecdote, example or explanation. Examination questions can target the direct answer, the example’s purpose or the broader attitude revealed by the story. The learner must maintain hierarchy. The anecdote is not automatically the main point simply because it is vivid.
Part II | Build a live speaker model
13. Give every speaker a stable identity
At the start of a conversation, establish who is who using names, roles or simple labels. In notes, initials are usually sufficient when writing is useful. The goal is not perfect biography; it is stable attribution. If speakers are introduced by role—student, teacher, customer, organiser—use that role consistently. Changing labels mid-task increases binding errors.
14. Record role before opinion when roles matter
Some tasks become easier when the listener understands each speaker’s function. A manager may approve a decision, while an adviser only recommends it. A customer can request; a staff member can explain policy. The same sentence carries different institutional weight depending on role. Speaker model therefore includes both identity and conversational authority where the audio supports it.
15. Track the main current position
Each speaker can be represented by a compressed stance: supports later start because transport; opposes because staffing; undecided pending evidence. This is more useful than verbatim notes. The model should update when the stance changes. If Speaker B moves from opposition to qualified support, the old label must not remain visually or mentally equal to the new one.
16. Separate position from reason
Two speakers can agree on the decision for different reasons. If notes only record “A yes, B yes,” a matching question about reasons becomes impossible. Keep stance and rationale separate: A → yes / cost; B → yes / access. This simple distinction prevents many multi-speaker errors.
17. Separate reason from evidence
A speaker may support a plan because of cost and then cite a statistic as evidence. The reason is the criterion; the statistic supports it. Questions can target either. A precise speaker model can therefore include stance, reason and evidence at different levels without turning into a transcript.
18. Track confidence and degree
“I support it”, “I think it could work”, “I’m willing to try it” and “I suppose it’s the least bad option” all point in a generally positive direction but differ in commitment. Advanced listening requires degree. Store the smallest accurate label: strong support, cautious support, provisional acceptance or reluctant compromise.
19. Track changes explicitly
A speaker may change position after receiving new information. Use a directional model: B: against → after data → cautiously for. This preserves trajectory, which can answer questions about final stance, change trigger or development of opinion.
20. Track unresolved status
Not every speaker reaches a decision. “I need to see the full report first” is not opposition and not support. The model should allow undecided, conditional or deferred states. Forcing every stance into yes/no categories creates false certainty.
21. Track reported views separately
“My parents think the course is too specialised, but I disagree.” This sentence contains two owners. Notes might read parents → too specialised; speaker → disagrees. Reported opinion should never automatically overwrite the current speaker’s stance.
22. Track group views separately from individual views
An interviewee may say “Most of the team wanted the earlier date, although I preferred waiting.” The group and individual occupy different nodes in the source map. This matters because examination questions frequently contrast personal view with consensus.
23. Track role changes in hypothetical speech
A speaker can temporarily imagine another perspective: “If I were the organiser, I’d choose Saturday.” The proposition is hypothetical, not necessarily the speaker’s actual recommendation. Context determines whether the imagined role reveals true stance or merely explores an option.
24. Speaker models should remain lightweight
If the model becomes a full biography, note-taking will steal attention. Use only information that helps answer likely jobs: identity, current stance, reason, key evidence, change. Everything else can remain in the general meaning model unless the task requires it.
Part III | Understand turn functions
25. Questions do more than request facts
A question can challenge, invite elaboration, redirect topic, check understanding or expose a contradiction. In interviews, recognising the question’s function helps predict the kind of answer that follows. “Are you saying the policy failed?” is not a neutral request for a date; it invites clarification or defence.
26. Answers can be direct or indirect
“Can you lead the event?” “I’m already committed to two deadlines that week.” The second speaker does not say “no”, but the pragmatic function is a refusal or strong reluctance. Training should distinguish literal form from conversational action while keeping inference conservative.
27. Clarification repairs shared meaning
One speaker may say “Do you mean the first trial or the final version?” The clarification reveals that the original reference was ambiguous. The listener should update the model after the answer rather than treating both referents as equally active.
28. Confirmation locks a proposition
Responses such as “Exactly”, “That’s right” or “Yes, Friday at three” can stabilise a plan. In negotiation audio, confirmation often marks the point at which provisional information becomes final. This is especially useful for schedule and decision questions.
29. Correction changes proposition status
“No, not Thursday—Friday.” The correction can be explicit or embedded in a longer response. The listener must replace the old state rather than store both without labels. This is a conversation-specific application of final-state control.
30. Expansion adds evidence without changing stance
A speaker may agree and then elaborate with reasons. The new material supports the existing position rather than revising it. Distinguish expansion from correction. Otherwise the listener may incorrectly infer a change merely because the speaker continues.
31. Concession acknowledges without surrendering
“You’re right that it costs more, but the reliability matters more to me.” The speaker grants one point and retains the larger position. A concession is neither full agreement nor reversal. It changes the argument’s nuance.
32. Challenge questions increase evidential demand
An interviewer may ask “What evidence do you have for that?” The answer that follows should be processed as support, not automatically as a new claim. Question function guides note hierarchy.
33. Topic shifts reset the local model
“Let’s move on to…” or a new interview question can create a fresh evidence corridor. A listener stuck on the previous topic should use the shift as a re-entry landmark. Topic shifts therefore support both comprehension and recovery.
34. Summaries compress the shared model
An interviewer or speaker may summarise: “So the main issue is timing rather than cost.” Such turns are high-value because they reveal how the participants themselves organise the conversation. Still verify that the other speaker accepts the summary; a following correction can change it.
35. Backchannels signal attention, not always agreement
“Mm-hm”, “right”, “I see” and “okay” can simply show that the listener is following. Do not convert every backchannel into endorsement. Examine whether the speaker later accepts, challenges or qualifies the proposition.
36. Silence and hesitation have limited evidential value
A pause can reflect uncertainty, thought, politeness or natural speech rhythm. It should rarely support a strong psychological inference on its own. Combine hesitation with lexical content and context before classifying attitude.
Part IV | Conversation structure and negotiation
37. Proposal → response
Many exam dialogues begin with a suggestion and response. The learner should distinguish proposal content from response stance. “We could meet Friday.” “That might work, but only after four.” The final state is conditional acceptance, not simple agreement.
38. Proposal → objection → revision
A proposal meets a constraint and is modified. Track which feature changes and which remain stable. If time changes but place does not, update only the time field. Overwriting the whole plan creates new errors.
39. Proposal → alternative
A rejected option can still be true as something discussed. A question about final choice must distinguish considered alternatives from adopted action. Notes should visually mark status.
40. Alternative → comparison → choice
Speakers may compare options on cost, convenience and risk. The final choice may depend on only one decisive criterion. Do not assume the longest-discussed criterion caused the decision.
41. Constraint negotiation
One speaker can accept the goal but impose conditions: “I can do Wednesday if we finish before five.” Conditions are part of the final agreement. An answer that omits the condition can be too broad.
42. Compromise
The final solution may not equal either original proposal. A says Monday, B says Thursday, and they settle on Wednesday. Training should model negotiation as dynamic state, not a binary contest.
43. Deferred decision
Some conversations end without resolution: “Let’s check the figures and decide tomorrow.” The final state is deferred. Learners who expect every dialogue to produce a clear choice may invent one.
44. Decision ownership
A speaker can propose while another has authority to decide. Questions may ask who made the final decision or whose recommendation influenced it. Role and conversational function interact.
45. Reason for change
A revised decision can be triggered by new evidence, a practical constraint, another speaker’s argument or a change in priority. Track the trigger, not just the before-and-after states.
46. Resistance without refusal
“I’m not sure that solves the problem” keeps negotiation open. The speaker resists but does not reject. Degree matters for attitude questions.
47. Agreement without enthusiasm
“Fine, let’s do that” can mark acceptance even when tone is reluctant. The final action and emotional stance can differ. A question about decision should not be answered with attitude, and vice versa.
48. Social face-saving
Speakers may soften disagreement to preserve relationships. “Perhaps another approach would be easier” can function as criticism. Pragmatic listening should remain evidence-based and avoid over-reading social nuance beyond what the dialogue supports.
Part V | Interview architecture
49. The interviewer frames the evidence space
Interview questions define local topics. A new question can create a clean section boundary even if the interviewee links back to earlier material. The learner should use question turns as organisational markers.
50. Opening questions establish identity and context
Early questions often invite background, role or origin stories. These details can matter later because they explain perspective. However, not every biographical fact is answer-bearing. Keep the hierarchy.
51. Probing questions ask for depth
“Can you explain what you mean by that?” or “Why was that important?” signals that the next answer will develop a previous claim. The listener should connect the expansion to its parent idea.
52. Follow-up questions can challenge assumptions
An interviewer may ask “But didn’t the second trial show the opposite?” The next answer often contains correction, defence or qualification. This is a high-value zone for attitude and evidence questions.
53. Redirect questions create a topic reset
“Let’s turn to…” marks a new section. Use it as a mental chapter boundary and re-enter if attention drifted.
54. Interviewer summaries can be contested
“So you regret the decision?” “Not exactly…” The interviewer’s summary is a hypothesis, not authoritative truth. The interviewee’s correction controls her actual stance.
55. Anecdotes serve functions
A story can illustrate a claim, justify an attitude, provide evidence or humanise the topic. Ask what job the anecdote performs rather than treating every narrative detail as equal.
56. Examples can be representative or exceptional
An interviewee may offer one unusual case. Do not generalise it to all situations unless the speaker does. Scope remains important even in conversational storytelling.
57. Interviewees can self-correct
“It was in 2018—no, 2019.” The same final-state rules apply. Because interview speech often feels spontaneous, corrections may be frequent and less formally signalled.
58. Interviewees can reframe the question
“I wouldn’t describe it as a failure; it was more of a partial result.” The answer challenges the interviewer’s premise. The learner should capture the reframing, which often carries stance.
59. Interviewees can refuse a premise
“That assumes cost was the main issue, which it wasn’t.” Here the response addresses not only content but the question’s framing. Advanced listening should track this meta-level correction.
60. Interview conclusions may be distributed
The final view can emerge over several answers rather than one explicit closing statement. A global attitude question may require integrating the entire interview trajectory.
Part VI | Speaker attitude, intention and pragmatics
61. Intention is not identical to content
A speaker can mention an idea in order to reject it, joke about it, question it or ask for clarification. The listener must distinguish proposition content from communicative purpose.
62. Request versus suggestion
“Could we move it to Friday?” can function as a proposal rather than a literal question about possibility. Context determines speech act.
63. Suggestion versus decision
“We could…” remains an option until participants commit. Do not promote suggestions into final plans.
64. Advice versus obligation
“You might want to…” is weaker than “You must…” even if both concern the same action. Modality controls force.
65. Complaint versus explanation
A speaker describing a problem may be venting, explaining a decision or requesting change. Questions about purpose require the surrounding interaction.
66. Reassurance
“Don’t worry, the deadline hasn’t changed” performs emotional and factual work. The learner should not let emotional language obscure the concrete proposition.
67. Sarcasm and irony
Where examination material genuinely includes irony, interpret through convergence of wording, context and delivery. Do not infer sarcasm from unusual intonation alone.
68. Enthusiasm
Repeated positive evaluation, energetic commitment and specific reasons can support enthusiasm more strongly than one adjective. Degree should come from cumulative evidence.
69. Reluctance
A speaker may agree while expressing reservations. Reluctant acceptance differs from rejection. Final action and emotional stance should be represented separately.
70. Uncertainty
“I think”, “probably”, “I’m not entirely sure” and hesitation can reduce commitment. Preserve uncertainty rather than forcing certainty.
71. Frustration
Frustration may be signalled by repeated problem descriptions and evaluative language. Avoid jumping from mild dissatisfaction to anger without support.
72. Curiosity
Follow-up questions, requests for examples and exploratory language can signal curiosity rather than disagreement. Function matters.
Part VII | Examination question families in conversations and interviews
73. Who said what?
Matching speakers to propositions is the most direct source-binding task. Use initials or a small matrix when needed. The deep rule is speaker before claim.
74. Who agreed with whom?
Agreement can be proposition-specific. Two speakers may agree about the problem and disagree about the solution. Identify the exact claim under comparison.
75. Why did the speaker change view?
Track initial stance, trigger and final stance. The answer is usually the change mechanism, not merely the final opinion.
76. What was finally decided?
Ignore obsolete proposals unless the question asks for them. Track until the negotiation closes.
77. What is the speaker’s attitude?
Integrate lexical evaluation, qualification and context. Use the smallest accurate stance.
78. What does the response imply?
Interpret indirect meaning conservatively. A refusal can be implied; a detailed motive may not be.
79. Why did the interviewer ask that?
Identify function: clarify, challenge, redirect, request evidence, invite example or test consistency.
80. What is the main disagreement?
Speakers may share many beliefs. Find the specific proposition or criterion that separates them.
81. What do the speakers have in common?
Look for shared conclusion, concern, experience or value even when reasons differ.
82. Which speaker mentions the evidence?
Distinguish claim owner from evidence provider when conversation is collaborative.
83. Which option was considered but rejected?
Track status. A rejected option can be heard clearly and still be wrong as final answer.
84. What is the purpose of the example?
Examples can support, challenge, clarify or personalise a claim. Function questions operate above detail.
Part VIII | Failure modes specific to conversational audio
85. Content without source
The learner remembers the proposition but cannot identify who owns it. Repair source labels.
86. Source without time
The learner knows Speaker B’s opinion but not that B later changed it. Add stance trajectory.
87. Agreement flattened
Partial agreement becomes total agreement. Repair proposition scope.
88. Politeness misread
Social acknowledgement becomes intellectual endorsement. Listen through the response.
89. Reported view captured as current view
Quotation or report is mistaken for ownership. Use nested source notation during training.
90. Backchannel misread
“Right” or “I see” becomes agreement without later evidence. Treat backchannels cautiously.
91. Interrupted sentence treated as complete
An abandoned phrase becomes a false proposition. Wait for completion.
92. First proposal trap
The first suggestion is selected despite later negotiation. Mark status.
93. Longest turn bias
The speaker who talks most is assumed to be most important. Question relevance, not duration, controls attention.
94. Interviewer authority bias
The interviewer’s paraphrase is treated as fact even when the interviewee corrects it. Source authority does not override content.
95. Anecdote salience
A memorable story becomes the main point. Restore hierarchy.
96. Rapid speaker-switch cascade
One missed turn causes the listener to lose who is speaking next. Re-entry landmarks include name, role, question-answer relation and new topic.
Part IX | Diagnostic tests
97. Source-binding test
Play a two-speaker clip and provide the propositions afterward. Ask who owns each. If meaning is accurate but attribution is weak, the source model is the first target.
98. Reported-speech test
Use nested opinions and ask whose view is whose. Keep vocabulary simple.
99. Turn-function test
After each short turn, ask what it does: question, answer, challenge, clarification, concession, correction or decision.
100. Negotiation-state test
Ask for initial proposal, constraint and final decision. If the learner remembers values but not status, final-state control is weak.
101. Politeness-versus-agreement test
Use indirect disagreement and ask whether the speaker actually accepts the proposition.
102. Stance-trajectory test
Ask for initial and final attitude plus the trigger. This separates static from dynamic speaker models.
103. Backchannel test
Use “right”, “okay”, “I see” in contexts where they do and do not signal agreement. Ask what evidence beyond the backchannel determines stance.
104. Interview-question function test
Provide several interviewer turns and classify their purpose. This can be trained without full audio first.
105. Anecdote-function test
Ask whether a story illustrates, proves, challenges or contextualises the main claim.
106. Speaker-switch recovery test
Introduce one difficult turn, then measure whether the learner correctly identifies the next speaker and proposition.
107. Multi-speaker density test
Compare two-speaker and four-speaker versions with similar language. A large drop indicates source interference.
108. No-note versus speaker-matrix test
Compare conditions. If a tiny matrix sharply improves accuracy, working memory binding may be the bottleneck.
Part X | Practice architecture
109. Start with two voices and one dimension
Speaker A supports, B opposes. Keep language simple. Once source binding is stable, make reasons more similar.
110. Add same stance, different reason
Both speakers support but for different reasons. This removes polarity as a shortcut.
111. Add reported speech
One speaker reports another person’s view. The learner now tracks nested ownership.
112. Add stance change
One speaker changes opinion. Source now has a temporal dimension.
113. Add three speakers
Increase source density after two-speaker control is reliable.
114. Add similar voices
Rely less on acoustic contrast and more on turn structure, names and content.
115. Add faster turn exchange
Shorter turns increase switching cost without necessarily increasing speech rate.
116. Add negotiation
Proposal, constraint, revision and decision create state tracking.
117. Add interview probes
Interviewer follows up, challenges and redirects. The learner tracks question function.
118. Add anecdotes
Maintain hierarchy between story and main answer.
119. Add close attitude options
Differentiate cautious support, reluctant acceptance, uncertainty and opposition.
120. Integrate under target timing
Only after mechanisms are stable should full conversational sections be used to test coordination.
Part XI | Conversation drills
121. Speaker-label drill
Allow only initials and three-word stance summaries. This prevents transcript-like notes.
122. Turn-function drill
Pause after turns during early training and label function. Later remove pauses.
123. Agreement-boundary drill
Use pairs where one speaker says “I understand” but later disagrees. Ask what exact proposition is accepted.
124. Reported-view drill
Use “my teacher thinks”, “people say”, “I agree” and “I disagree” in different combinations.
125. Decision-chain drill
Track proposal → objection → alternative → final choice across practical scenarios.
126. Conditional-agreement drill
Use “yes, if…”, “only if…”, “as long as…” so conditions remain attached to acceptance.
127. Stance-degree drill
Compare enthusiastic support, cautious support, reluctant acceptance, uncertainty and rejection.
128. Interview-probe drill
Identify whether a follow-up requests example, evidence, clarification or defence.
129. Anecdote-compression drill
After a story, summarise its function in one phrase: evidence for X, example of Y, reason for change.
130. Speaker-switch recovery drill
Mask or complicate one turn, then score accurate re-entry on the next speaker.
131. Cross-turn inference drill
The evidence needed for an inference is split across two speakers. Require the smallest supported conclusion.
132. Final-state interview drill
The interviewee revises a claim during probing. Ask for the final position and the evidence that caused the revision.
Part XII | Interview drills
133. Question-map drill
Write interviewer questions as headings and summarise each answer underneath. This turns a long interview into a structured map.
134. Probe-chain drill
Track an initial answer followed by two follow-ups. Identify how each probe changes depth or direction.
135. Premise-challenge drill
The interviewee rejects the assumption in the question. Ask what premise is being corrected.
136. Evidence-source drill
Separate personal experience, reported data and opinion inside the interviewee’s answer.
137. Story-versus-claim drill
Identify which sentence contains the general claim and which details merely illustrate it.
138. Global-attitude drill
Integrate stance across several answers rather than using one phrase.
139. Interviewer-intention drill
Classify question functions without relying on fixed wording.
140. Topic-shift drill
Use interviewer questions as re-entry landmarks after an attention lapse.
141. Compression drill
After a three-minute interview, produce five lines: topic, interviewee position, two reasons, one qualification.
142. Contradiction-check drill
Compare an early and later answer to see whether the interviewee actually contradicted herself or merely added nuance.
143. Quote ownership drill
Track when the interviewee quotes another person or group and whether she endorses the quote.
144. Closing-answer drill
Ask whether the final answer summarises, changes or merely adds to the overall position.
Part XIII | Resident casebook
145. Maren: structure without source
Maren can outline the conversation perfectly but writes “cost concern” without a speaker label. A matching question exposes the weakness. Her repair is to bind every evaluative point to a source before compressing structure.
146. Iona: reported view becomes owned view
Iona hears a sophisticated argument and remembers it accurately, but it belonged to a colleague quoted by the interviewee. Her evidence discipline must include provenance, not just content.
147. Leonie: rapid turn exchange
Leonie misses one short objection and remains unsure who is speaking. She uses the next interviewer question as a re-entry landmark and rebuilds the source map instead of replaying the lost turn mentally.
148. Maren: over-writing negotiation
She records every proposal. The final agreement arrives while she is still writing the second alternative. A compact state diagram outperforms prose.
149. Iona: politeness over-inference
“That’s an interesting idea” sounds positive, so Iona selects agreement. The following “but” changes the stance. Her rule becomes: social acknowledgement is provisional until the full response closes.
150. Leonie: first speaker shock
A new voice causes heavy notes and missed turns. She trains a first-minute ceiling: no transcription, only source, topic and final-state anchors.
151. Maren: anecdote becomes main point
A vivid interview story occupies half her notes. The question asks the interviewee’s reason for changing policy. Maren learns to label the anecdote as evidence under the reason, not as a separate main idea.
152. Iona: cautious support versus uncertainty
Two options seem plausible. The interviewee says the plan “could work if the trial data hold”. Iona compares conditional support with undecided stance and chooses the option that preserves the explicit condition.
153. Leonie: one missed name
She tries to spell an unfamiliar guest’s name and loses the next answer. If exact spelling is not required, she learns to use an initial and role.
154. Maren: answer scope after rich conversation
She understands three reasons but the question asks for one final objection. The conversation model is rich; the answer must still be narrow.
155. Iona: interviewer premise accepted too quickly
The interviewer says “So you regret it?” but the interviewee replies “Not really.” Iona learns that interviewer summaries are propositions to test, not authoritative interpretation.
156. Leonie: difficult turn before easy turn
A dense technical answer is followed by a simple scheduling decision. Recovery training lets the first remain partly unresolved without losing the second.
Part XIV | Transfer across conversation types
157. Casual conversation to formal interview
Turn-taking remains, but register becomes more structured. The listener transfers source and stance tracking while adapting to longer turns.
158. Interview to panel discussion
Speaker count rises and interviewer control weakens. Source binding becomes harder; conversational floor cues matter more.
159. Service dialogue to academic discussion
Practical decisions become conceptual claims. The same agreement, correction and cause relationships still apply.
160. Two-speaker to multi-speaker
Expand source map gradually. Do not change topic difficulty simultaneously at first.
161. Familiar topic to unfamiliar topic
Use question structure and discourse relationships as fallback when vocabulary is thinner.
162. Clear disagreement to qualified disagreement
Move from explicit “I disagree” to concession and hedging. Stance degree becomes the transfer target.
163. Explicit final decision to implied resolution
Not every conversation ends with “we agree”. The final action can be inferred from commitments, provided the evidence is sufficient.
164. Scripted clarity to more natural hesitation
As material becomes more natural, unfinished phrases and backchannels increase. The listener should preserve proposition status.
165. One interviewer style to another
Some interviewers probe aggressively; others invite long stories. Track function, not personality labels.
166. Paper questions to digital layouts
Source mapping must survive interface changes where the target assessment requires them.
167. Multiple choice to short answer
Correct understanding must become concise output without losing attribution.
168. Short clip to sustained interview
Section structure and endurance become important. The speaker model must persist over minutes rather than seconds.
Part XV | Stress tests
169. Similar-speaker stress
Use voices that are not easily distinguished acoustically. Test whether content and turn structure support source tracking.
170. Similar-position stress
Speakers agree on outcome but differ in reason. The learner must bind rationale accurately.
171. Changing-position stress
One speaker revises stance twice. Track final position and triggers.
172. Reported-speech stress
Several third-party views are quoted. Preserve ownership hierarchy.
173. Fast-turn stress
Use short rapid exchanges without artificially speeding pronunciation. Test switching cost.
174. Long-turn stress
One interviewee gives a long answer before the interviewer returns. Test section mapping and main-idea hierarchy.
175. Unknown-name stress
Introduce unfamiliar proper names and see whether the learner can track roles without spelling fixation.
176. Politeness stress
Use indirect disagreement and soft refusals. Test pragmatic inference degree.
177. Anecdote stress
Use a vivid story whose details are not the answer. Test salience control.
178. First-miss stress
Place one difficult turn early and measure downstream source accuracy.
179. Late-interview stress
Test whether speaker models and question jobs remain stable after sustained listening.
180. Unfamiliar-topic panel stress
Combine several speakers with new subject matter only after individual dimensions are stable.
Part XVI | Measurement
181. Speaker-attribution accuracy
Track source mistakes separately from general comprehension. A learner can understand the proposition and still fail attribution.
182. Stance accuracy
Measure whether degree and qualification are preserved, not just positive/negative direction.
183. Final-state accuracy
Track whether negotiation updates and changed opinions are correctly retained.
184. Reported-view accuracy
Count cases where quoted or reported views are incorrectly assigned to the speaker.
185. Turn-function accuracy
Can the learner recognise challenge, clarification, concession and decision?
186. Recovery cost
Count downstream errors after a difficult conversational turn.
187. Note efficiency
Compare note volume with attribution and later-item accuracy.
188. Interview-map quality
Can the learner reconstruct topic sections, main position and key qualifications from sparse notes?
189. Confidence calibration
High confidence on wrong speaker or wrong stance signals blind spots worth prioritising.
190. Late-conversation decay
Compare source and stance accuracy early versus late in sustained interviews.
191. Transfer penalty
Compare familiar speaker/topic conversations with unfamiliar ones at comparable level.
192. Independent diagnosis
Can the learner state whether the error was source, stance, timing, inference or answer transfer?
Part XVII | Teacher protocol
193. Observe before explaining
Ask who the learner thought said what before showing the transcript. Source errors disappear once names are visible.
194. Use transcript corridors
Reveal only the relevant exchange. Preserve the original auditory route.
195. Model source maps briefly
Show how an expert compresses speakers and stances, then let the learner perform.
196. Contrast partial and full agreement
Use paired examples differing in one proposition.
197. Fade initials gradually
Visible speaker labels are scaffolds. Reduce them when binding becomes stable.
198. Keep dialogue authentic
Artificial lines that exist only to contain a connector can teach brittle cue spotting. Use meaningful interaction.
199. Test function, not theatre
Do not reward dramatic interpretation of tone without evidence. Keep inference anchored.
200. Build multi-speaker difficulty gradually
Speaker count, topic novelty and lexical load should not all increase simultaneously early.
201. Use small-group comparison
Different students can explain who they thought owned a claim. Contrasting source maps reveals reasoning.
202. Route narrow weaknesses outward
If the problem is pure note-taking, use the note owner. If it is general conversational floor mechanics, use the eduKateSG specialist. Keep this page’s role integrated.
203. Preserve unseen conversation material
Repeated dialogues become scripts. Transfer requires first-contact source tracking.
204. End with new speakers
A session should finish with source and stance work on unfamiliar voices when transfer is the goal.
Part XVIII | Self-study protocol
205. First listen without transcript
Build a source map from the audio alone.
206. Record speaker hypotheses
Use initials and short stance labels before checking.
207. Replay only with a question
“Who owns this view?” is better than replaying everything.
208. Use transcript after reconstruction
Compare source, wording and turn function.
209. Write one portable rule
Example: “acknowledgement is not agreement until the proposition closes.”
210. Test with a new dialogue
Transfer immediately.
211. Return after delay
Use another conversation days later without announcing the target.
212. Rotate speaker types
Use interviews, negotiations, discussions and service dialogues.
213. Rotate topics
Prevent source strategies from becoming tied to familiar subject scripts.
214. Track confidence
Notice whether unfamiliar voices feel worse than they actually perform.
215. Build a small active ledger
Keep only current conversational risks.
216. Integrate into full listening sections
Do not leave conversation skills isolated from the rest of the examination.
Part XIX | Parent-supported listening
217. Ask “who thinks what?”
After a short interview or discussion, ask the learner to distinguish speakers and reasons. Keep it conversational.
218. Ask “did they actually agree?”
This supports proposition-specific stance without technical jargon.
219. Ask “what changed their mind?”
Encourage trajectory thinking.
220. Ask “what was finally decided?”
Useful for everyday planning conversations or appropriate audio.
221. Avoid quizzing every interaction
Home conversation should remain communication. Structured practice can be brief and intentional.
222. Use varied real voices
Interviews and discussions expose the learner to natural speaker diversity.
223. Encourage evidence language
“What made you think that?” supports listening inference without turning the discussion into a test.
224. Support broad world knowledge
Interesting interviews build topic knowledge that later reduces listening load.
Part XX | Exam-day operating model for conversations and interviews
225. Before the audio
Identify whether the task cares about speaker, reason, stance, decision, change or purpose. Prepare the smallest useful source structure.
226. When the first speaker begins
Establish identity and current role. Do not over-note introductory small talk unless the question targets it.
227. At every speaker switch
Reset source mentally. New voice, new owner until evidence says otherwise.
228. When speakers overlap in viewpoint
Track the discriminating reason or degree, not just shared conclusion.
229. When politeness appears
Wait for the complete response before assigning agreement.
230. When a speaker changes view
Update the model. Old stance becomes historical, not current.
231. When you miss a turn
Use the next question, name, topic shift or speaker switch as a re-entry landmark.
232. When an unfamiliar name appears
Track function before spelling unless exact spelling is required.
233. When the interviewer challenges
Expect defence, clarification, correction or qualification. Listen for what changes.
234. When an anecdote begins
Maintain the parent claim. Ask what the story is doing.
235. When two answers seem plausible
Compare source, time, degree and final status. These are common conversational discriminators.
236. At the end
Check answer scope. Rich conversation does not require rich output if the question asks for one precise point.
Part XXI | Technical appendix: compact conversation map
Speaker A: role → current stance → reason → evidence → changes. Speaker B: role → current stance → reason → evidence → changes. Shared state: unresolved issue → options → constraint → final decision. This map can be mental or minimally written. It is not intended to become a full transcript.
Part XXII | Technical appendix: turn-function dictionary
Proposal: introduces possible action. Question: requests or frames information. Clarification: resolves ambiguity. Challenge: tests a claim. Concession: grants part without surrendering the whole. Correction: replaces a proposition or field. Expansion: adds support. Backchannel: signals attention. Decision: commits. Deferral: leaves unresolved. Summary: compresses shared state. Redirect: changes topic.
Part XXIII | Technical appendix: error ledger for conversation audio
Useful categories include source binding, reported-view confusion, stance degree, change tracking, politeness misread, turn-function error, negotiation final-state error, anecdote salience, rapid-switch recovery, note overload and answer-scope transfer. Record only recurring or high-cost patterns.
Part XXIV | Technical appendix: transfer ladder
Level 1: two clear speakers, familiar topic. Level 2: same stance/different reason. Level 3: reported views and stance change. Level 4: unfamiliar speakers and topic. Level 5: fast turn exchange, subtle pragmatics and mixed question families under target timing. Each level should preserve the core source-and-stance model while increasing novelty.
Part XXV | Technical appendix: final checklist
- I know who each speaker is.
- I keep claims attached to sources.
- I distinguish a speaker’s view from a reported view.
- I do not treat every backchannel as agreement.
- I wait for the full response before classifying stance.
- I track initial and final positions separately.
- I preserve conditions on agreements.
- I know the difference between proposal and decision.
- I can identify why an interviewer asked a follow-up.
- I keep anecdotes under the larger claim.
- I can recover after a rapid or missed turn.
- I can answer narrowly even when the conversation is rich.
Closing synthesis | Listen to relationships, not just voices
Conversation listening becomes reliable when the learner stops treating dialogue as a stream of alternating sentences and begins to hear a changing system of ownership, response and commitment. One speaker proposes. Another qualifies. A third reports a view. Someone asks for evidence. An objection changes the plan. A polite phrase acknowledges without agreeing. A final turn converts possibility into decision.
Maren preserves the architecture of that movement. Iona preserves provenance and evidential degree. Leonie preserves the live stream when rapid turns create pressure. Their shared task is to keep the speaker model current enough that the question can be answered from the actual final conversation rather than from the most memorable phrase.
The examination does not reward hearing voices as separate audio objects. It rewards understanding what people are doing with language together.
Previous: SETC-SEC-LISTEN-080 | Secondary Listening Examination Training Plan · Hub: Secondary English Listening Examination Performance · Next: SETC-SEC-LISTEN-100 | Listening Comprehension Talks and Monologues
Advanced Conversation and Interview Casebook | Source, Stance, Negotiation and Recovery Under Pressure
The main article builds the operating system. This advanced casebook stress-tests it. Each case isolates a conversational architecture that repeatedly causes otherwise capable listeners to lose marks: apparent agreement that is only acknowledgement, shared conclusions with different reasons, source nesting inside reported speech, question-answer pairs that change function halfway through, interviews in which an anecdote becomes more memorable than the claim it supports, and negotiations where a final decision is assembled rather than announced. The point is not to memorise cases. It is to practise seeing the same speaker-and-relationship mechanics when the surface changes.
Casebook 1 | The agreeable opening that becomes disagreement
Speaker B begins, “I can see why that approach appeals to people,” which sounds positive. She then says that the plan ignores implementation cost and would be difficult to sustain. A learner who commits after the first clause selects “supports the plan”. The stronger model waits for the complete turn and separates acknowledgement from stance. In review, ask which proposition B accepts: that the plan has appeal. Then ask which proposition B rejects: that it is practical. The repair rule is portable: polite or positive framing does not determine the final intellectual position until the whole response closes.
Casebook 2 | Two supporters, two incompatible reasons
Speaker A supports extending library hours because students need evening access. Speaker B supports the same extension because workers cannot visit during the day. A matching item asks which speaker focuses on employment schedules. Both speakers share the final recommendation, so “supports longer hours” has no discriminating value. The listener must bind the reason to the source. In training, deliberately give several speakers the same conclusion and remove easy polarity shortcuts. The speaker model becomes “A → student access; B → work schedules”, which is both shorter and more useful than a transcript.
Casebook 3 | The speaker who quotes a critic accurately
An interviewee says, “One criticism is that the programme is expensive, and that criticism is fair in the short term. But I think the long-term savings outweigh it.” Three viewpoints are present: the external criticism, the interviewee’s acknowledgement that the short-term cost is real, and the interviewee’s final support. A careless source map compresses all three into “programme expensive”. The accurate model preserves layers: critics → expensive; speaker → accepts short-term cost; speaker final → supports because long-term savings. This distinction is central to reported-view questions and also to attitude calibration.
Casebook 4 | The interviewer’s summary is wrong
The interviewer says, “So the project failed mainly because of poor planning?” The interviewee responds, “Not really. Planning was difficult, but the decisive problem was the supplier delay.” If the learner assumes interviewer authority, “poor planning” becomes the answer. The correct listening rule is that interviewer summaries are hypotheses inside the conversation, not official interpretations. The interviewee can accept, modify or reject them. A useful drill gives learners interviewer paraphrases followed by corrections and asks which proposition survives after the exchange.
Casebook 5 | A decision emerges through compromise
Speaker A wants Monday. Speaker B cannot attend Monday and suggests Thursday. A says Thursday is too late but Wednesday afternoon is possible. B agrees to Wednesday at three. No single early turn contains the final answer. The listener must maintain a temporary state rather than committing to the first plausible date. Notes can use “Mon ✗ → Thu? ✗ → Wed 3 FINAL”. The architecture is negotiation, not recall. Transfer the same structure to budgets, locations, methods and recommendations so the rule becomes “track until shared commitment stabilises”.
Casebook 6 | The reluctant yes
“It’s not my first choice, but I can work with it.” The speaker accepts the plan while remaining unenthusiastic. An attitude question may ask for emotional stance; a decision question may ask whether the plan was accepted. Those are different jobs. The listener needs two dimensions: action state and attitude state. A compact model could be “accepts / reluctant”. Training should deliberately pair questions that target the same turn differently so learners stop assuming one label can answer every operation.
Casebook 7 | The story that changes nothing
An interviewee tells a long story about a difficult first week, including several vivid details, then concludes that the experience did not alter her belief that the programme was worthwhile. The anecdote dominates memory because it is concrete, but the attitude remains positive. Maren’s structural rule is to attach stories beneath the proposition they serve. Ask: does the story explain a change, support a stable claim, provide an exception or simply add colour? If the story does not change the final stance, it should not be allowed to replace it.
Casebook 8 | The story that changes everything
In a different interview, the speaker begins supportive but describes one incident that made her reconsider the policy. Here the anecdote is the turning point. The same surface form—storytelling—has a different discourse function. The listener should not use a generic “ignore anecdotes” rule. Instead, ask what the story does to the current speaker model. Does stance remain stable, become qualified or reverse? Function determines importance.
Casebook 9 | The incomplete answer completed by the next question
An interviewee says, “The main reason was probably…” and trails off. The interviewer asks, “The funding issue?” The interviewee replies, “Yes, exactly.” The decisive proposition is distributed across turns. Treating the unfinished first turn as a complete statement loses meaning. This case trains cross-turn proposition assembly. The answer belongs to the interviewee even though the interviewer supplies the lexical phrase. Source ownership follows endorsement, not simply who pronounced the words.
Casebook 10 | The interviewer supplies a wrong candidate answer
“Was it the funding issue?” “No, actually it was staffing.” Here the interviewer again supplies language, but the interviewee rejects it. The learner must track status, not lexical presence. A high-value drill pairs this case with Casebook 9 so the same question structure leads to opposite outcomes. The discriminating evidence is the response relation—confirmation versus correction.
Casebook 11 | Agreement on problem, disagreement on solution
Both speakers believe congestion is serious. A wants to increase public transport capacity; B wants to change working-hour patterns. A question asking “What do the speakers agree about?” should target the shared problem. A question asking “How do their views differ?” should target the solution. The same conversation contains both agreement and disagreement. Train proposition-specific stance instead of global labels such as “the speakers agree”.
Casebook 12 | Agreement on solution, disagreement on rationale
Both speakers support the same policy. A justifies it through cost savings; B through fairness. This architecture is especially dangerous for matching tasks because listeners remember the shared conclusion and blur reasons. Use a reason matrix and later remove it as source binding becomes automatic. The learner should be able to answer “who values what?” even when everyone reaches the same decision.
Casebook 13 | A question that is really a challenge
The interviewer asks, “But if the scheme was so successful, why did participation fall?” The literal surface is a request for explanation, but the discourse function is challenge: the interviewer exposes tension between claim and evidence. The interviewee’s next turn is likely to defend, qualify or revise the original claim. Recognising this function prepares attention for correction and evidence. The learner should hear the question as part of the argument, not merely as a prompt for a fact.
Casebook 14 | A question that is really an invitation
“What happened next?” may simply invite narrative continuation rather than challenge. Same grammatical form—question—different interactional role. Training interview function improves prediction without content guessing. The listener anticipates story continuation, explanation, evidence or defence based on what the interviewer is doing.
Casebook 15 | The speaker who softens refusal
“I’m not sure I can commit to that timetable” may function as refusal or strong hesitation depending on context. If the conversation immediately moves to an alternative plan, the refusal interpretation gains support. If the speaker later checks availability, the state may remain undecided. The examination answer should match the evidence trajectory, not one phrase in isolation. Pragmatic interpretation is temporal.
Casebook 16 | The speaker who appears uncertain but is actually precise
A speaker says “I think the meeting begins at nine” because she is speaking casually, yet later confirms the schedule from a document. The initial hedge may reflect conversational style rather than factual uncertainty. The final evidence state should replace the preliminary one. This case warns against treating every hedge as permanent stance. Track how certainty evolves.
Casebook 17 | The confident speaker who is corrected
Speaker A confidently states a date; Speaker B immediately corrects it with documentary evidence. Confidence of delivery does not make A’s proposition current. Examination listening should prioritise final supported state over vocal certainty. This is another reason to distinguish attitude evidence from factual status.
Casebook 18 | The shared assumption that is never stated
Two colleagues discuss “moving it to the second week” without naming the event because both know the context. The learner must resolve the referent from earlier turns. This tests discourse memory and entity binding rather than vocabulary. In training, use pronouns and ellipsis only after the underlying source model is stable, because otherwise several variables fail at once.
Casebook 19 | One speaker completes another’s sentence
A begins, “The main problem is the lack of—” and B says, “Staff?” A replies, “Exactly.” The proposition becomes shared. A initiated it and endorses B’s completion; B supplied the word. Depending on the question, both may be associated with the idea, but the original position belongs to A. This is a useful advanced source-boundary case because conversational ownership can be collaborative rather than neatly isolated.
Casebook 20 | The interruption that changes direction
A is explaining why the first plan is workable when B interrupts with new information that invalidates an assumption. A then abandons the defence and revises the plan. The interruption is not merely noise; it causes state change. The listener should track the effect of the interrupted turn on the final model. Long unfinished reasoning that is later abandoned should not retain equal weight.
Casebook 21 | Backchannel overload
During a long interview answer, the interviewer repeatedly says “right”, “yes” and “I see”. None necessarily indicate agreement; they can simply maintain the conversational floor. A learner counting these as endorsements may infer consensus that does not exist. Train by muting or mentally ignoring backchannels and asking whether the interviewer ever actually states a position.
Casebook 22 | The apparently neutral interviewer with a loaded frame
An interviewer asks, “Why did management ignore the warning signs?” This presupposes that warnings existed and were ignored. An interviewee may accept the frame, challenge it or partially accept it. Advanced listeners should recognise that question wording can embed propositions. The answer may depend on whether the interviewee adopts the premise, not merely on the surface topic.
Casebook 23 | The speaker who separates personal and professional view
“Personally I liked the original design, but as project lead I had to choose the safer option.” One speaker holds two evaluative perspectives tied to roles. A question about personal preference differs from a question about final decision. The source is the same person, but viewpoint role differs. This is an advanced version of source binding.
Casebook 24 | The speaker who distinguishes past and present view
“At the time I thought the delay was unnecessary; looking back, I understand why they did it.” Same speaker, different temporal stance. A global attitude question may target current retrospective judgement. The listener must time-index opinion, just as changed logistics are time-indexed.
Casebook 25 | The conditional supporter
“I’d support the plan if the funding were guaranteed.” This is neither unconditional support nor current rejection. The stance is conditional. Multiple-choice distractors may deliberately omit the condition. Train learners to treat conditions as part of the proposition rather than optional decoration.
Casebook 26 | The speaker who agrees with the goal but not the method
“We definitely need to reduce waste, but banning all disposable packaging immediately would be unrealistic.” A learner may classify the speaker as anti-environmental or pro-ban depending on which phrase is remembered. The accurate model is pro-goal, opposed to this method. Advanced stance listening often requires separating end from means.
Casebook 27 | The interviewer’s repeated question
An interviewer asks essentially the same question twice using different wording because the first answer was incomplete. The repetition signals unresolved information. The second answer may contain the missing evidence. Learners should not dismiss repeated questions as redundancy; they can mark a gap in the shared model.
Casebook 28 | The interviewee answers a different question
A speaker sidesteps the interviewer’s question and discusses a related issue. A task asking “Did the speaker explain why X happened?” may require recognising that the answer was not actually provided. Listening comprehension includes relevance, not just fluent speech.
Casebook 29 | The source who changes only one part of a position
A speaker keeps supporting the overall plan but changes her preferred implementation after hearing cost data. A learner who records “changed opinion” too broadly may think she now opposes the plan. Update fields selectively: stance stable; method changed. Conversation models should support partial revision.
Casebook 30 | The source who changes degree, not direction
A speaker moves from enthusiastic support to cautious support after a risk is raised. Direction remains positive; commitment decreases. An attitude item may distinguish “enthusiastic” from “qualified”. This is why stance trajectories need degree, not only sign.
Casebook 31 | The source who changes direction, not reason
A speaker values affordability throughout but initially believes option A is cheaper. After corrected figures, she switches to option B for the same reason. The underlying criterion is stable while choice changes. Questions can target either the reason or the final choice. Preserve both layers.
Casebook 32 | The source who changes reason, not choice
A speaker initially prefers option A for convenience, then learns that convenience advantage does not exist but retains A because it is safer. Final choice is stable; rationale changes. Again, the conversation model must be field-specific rather than a single binary label.
Casebook 33 | The interviewer who supplies evidence
Sometimes the interviewer introduces data: “The survey found that most users preferred the new system.” The interviewee then interprets it. A question about evidence source should not attribute the survey finding to the interviewee merely because she discusses it. Track who introduces the evidence and who evaluates it.
Casebook 34 | Collaborative reasoning
Two speakers jointly build a conclusion. A proposes one cause, B adds another, A integrates both. The final explanation is shared rather than owned by one person. This architecture is common in discussions. If the question asks what the speakers conclude, use the shared state; if it asks what B contributed, preserve the individual addition.
Casebook 35 | Competing memories inside a conversation
Two speakers recall an event differently. Neither is immediately confirmed. The learner should not decide which memory is true based on confidence or detail. Track both as claims until the conversation provides evidence or leaves the issue unresolved. Listening is not fact-checking beyond the source.
Casebook 36 | The final joke that does not reverse the conclusion
A speaker ends with a light joke after making a serious recommendation. Tone changes, but proposition may not. Learners who over-weight emotional surface can misclassify stance. The final model should integrate humour without treating it as automatic evidence of rejection or unseriousness.
Casebook 37 | The joke that does reveal criticism
In another context, irony functions as criticism because the literal praise conflicts sharply with the described failure. Use wording, context and subsequent response together. The training goal is not a sarcasm detector; it is converging evidence.
Casebook 38 | The pause before the important answer
Hesitation can increase listener attention, but it does not define stance by itself. A speaker may pause because the question is complex. Do not let prosodic drama substitute for proposition analysis. The words after the pause remain decisive.
Casebook 39 | The response that corrects a misunderstanding
“No, I meant the earlier trial, not the final study.” This is a meta-conversational repair: the speaker corrects how a previous utterance was interpreted. The listener must revise the referent, and any answer built from the old referent becomes obsolete. Such repairs are high-value because they reveal the intended proposition explicitly.
Casebook 40 | The response that narrows scope
“I’m only talking about the first phase; the later phases worked well.” A previous broad interpretation is narrowed. Multiple-choice distractors may generalise the criticism to the entire project. Scope is updated through conversation, not merely vocabulary.
Transfer Lab 1 | Same conversation mechanism, five topics
Take conditional agreement and practise it in school scheduling, travel planning, project management, environmental policy and event organisation. The wording changes, but the deep structure remains “speaker accepts only if condition C holds”. The learner should identify stance and condition without relying on topic scripts. When accuracy survives all five, the mechanism is more likely abstract rather than memorised.
Transfer Lab 2 | Same topic, five interaction patterns
Use one familiar topic but vary interaction: clear disagreement, partial agreement, negotiation, reported opinion and stance change. Holding topic constant reduces background-knowledge variation and reveals whether conversational function itself is understood. This is useful when a learner appears comfortable only because the subject is familiar.
Transfer Lab 3 | New speakers, stable script architecture
Use several clear voices across comparable two-speaker discussions. Measure first-contact source accuracy and adaptation. If the listener knows what each speaker said but source errors rise with new voices, speaker binding—not semantic comprehension—is the transfer boundary.
Transfer Lab 4 | New registers
Move the same stance and negotiation mechanics from casual conversation to workplace discussion, interview and panel exchange. The language becomes more formal, but agreement, correction and final-state tracking remain. Transfer is successful when the operating system survives register without needing a new strategy for every setting.
Transfer Lab 5 | Longer turns
Begin with short dialogue, then extend each speaker’s turns. The learner must retain source identity across longer reasoning while avoiding excessive notes. This bridges conversation and monologue architectures without losing attribution.
Transfer Lab 6 | Faster speaker switching
Keep individual sentences simple but shorten turns. Switching cost rises even though speech rate does not. If errors increase, practise source resets rather than speed drills. This laboratory is especially useful for learners who describe rapid dialogue as “fast” when the true challenge is turn density.
Transfer Lab 7 | Delayed questions
Occasionally reveal a question only after a conversation to test whether the learner built a robust general speaker model rather than listening only for previewed details. Use this diagnostically, not as a replacement for format-faithful practice. Strong performance suggests the internal model itself is coherent.
Transfer Lab 8 | No-note conversation
Short dialogues can expose whether notes are habit rather than necessity. Compare source accuracy with no notes, initials only and free notes. If initials-only wins, the best external memory may be extremely small.
Transfer Lab 9 | Sparse-note sustained interview
For longer interviews, allow only section labels, stance changes and decisive evidence. The learner should reconstruct the whole interview afterward. This teaches that long audio requires hierarchy rather than transcription.
Transfer Lab 10 | Full examination integration
Place conversation and interview tasks among monologues, announcements and visual listening. The learner must identify the audio architecture quickly and adjust the model: source map for conversation, section map for monologue, action map for instructions, visual anchor map for graphic tasks. This cross-architecture switching is one reason the lane uses separate specialist owners.
Teacher Laboratory | Diagnosing identical wrong answers
Two learners choose the same wrong speaker. One misheard a name at the start; the other understood every turn but failed to update after a stance change. The correction cannot be identical. Ask each learner to reconstruct the speaker model before teaching. The first needs acoustic/source anchoring; the second needs time-indexed stance updating. Small-group teaching remains individual at the level of mechanism.
Teacher Laboratory | Fading the speaker matrix
Visible columns are useful early, but permanent dependence creates a new task cost. Fade from full columns to initials, then to mental source labels for shorter tasks. Keep the matrix available for genuinely dense multi-speaker sections. The correct amount of external structure depends on information load.
Teacher Laboratory | Designing fair attitude questions
Ensure the audio provides enough lexical, contextual and prosodic evidence to distinguish the intended stance. A good question tests interpretation, not mind reading. Distractors can differ in degree, but the correct answer should be defensible from the whole turn or trajectory. Poorly written attitude questions teach distrust rather than listening.
Teacher Laboratory | Designing fair source questions
When several speakers discuss the same topic, make the discriminating reason or position clear enough that attentive listening can separate them. Avoid relying solely on voice identity if the educational target is proposition attribution. Source tasks should reward meaning, not only auditory voice recognition.
Teacher Laboratory | Designing negotiation questions
Use genuine state changes. The final answer should emerge from interaction, while earlier options remain plausible because they were truly considered. This creates a meaningful final-state challenge rather than a trick based on obscure wording.
Self-study Laboratory | Conversation log
After one unseen dialogue, write five lines: speakers, initial positions, key disagreement, change if any, final shared state. Then answer the questions. If the five-line model is wrong, review the conversation architecture before individual answer details. This keeps self-study from becoming answer-key chasing.
Self-study Laboratory | Interview question map
Write each interviewer question as a heading and one-sentence answer summary beneath it. Mark where the interviewer challenges, redirects or requests evidence. Afterward, compress the whole interview into a final position and two key supports. This makes long interviews manageable without transcription.
Self-study Laboratory | Speaker-change retest
After repairing a source-binding error, find a different interview or dialogue with new voices and similar interactional demands. If the correction survives, the repair is more likely transferable. Replaying the original speakers is not enough.
Parent Laboratory | Shared listening without testing atmosphere
Listen to a short public interview together and discuss who changed their mind, what each person cared about and which answer felt strongest. The parent does not need technical labels. Questions such as “Did they actually agree?” and “What made her change?” naturally reinforce speaker models while keeping the activity conversational.
Exam Pressure Lab | Opening uncertainty
If the first two speakers are difficult to distinguish, do not over-build notes in panic. Use turn content, names and interviewer cues. A source label can remain provisional until the next stable clue. Overconfidence and panic are both costly; calibrated provisional modelling is safer.
Exam Pressure Lab | One lost turn
When a turn disappears, the learner should avoid replaying it internally. Re-enter at the next clear speaker boundary and reconstruct only if later evidence makes the missing turn relevant. A conversation provides frequent natural reset points; use them.
Exam Pressure Lab | One uncertain stance
If a speaker’s attitude remains unclear, mark the uncertainty and continue. A later turn may clarify or change the stance. Premature classification is especially risky in interviews because follow-up questions often elicit qualification. The model can remain temporarily unresolved.
Exam Pressure Lab | Rich understanding, narrow question
At the end of a long discussion, the learner may understand all positions and still need one simple answer: “Which speaker preferred the earlier date?” Answer the job, not the whole conversation. Rich models support narrow decisions; they do not justify over-answering.
Advanced Exit Criterion 1 | Source stability
The learner keeps propositions attached to the correct speakers across new voices, longer turns and multiple speakers.
Advanced Exit Criterion 2 | Stance trajectory
Initial, changing and final views remain distinct. The learner can identify what caused a change.
Advanced Exit Criterion 3 | Pragmatic calibration
Politeness, hesitation, indirect refusal and irony are interpreted conservatively from converging evidence rather than guessed from tone alone.
Advanced Exit Criterion 4 | Negotiation state
Proposals, alternatives, conditions and final decisions remain separate until commitment is clear.
Advanced Exit Criterion 5 | Interview architecture
The learner can use interviewer questions as structural markers and distinguish challenge, clarification, redirect and invitation.
Advanced Exit Criterion 6 | Note economy
Source maps remain compact enough that writing does not steal the next turn.
Advanced Exit Criterion 7 | Recovery
One difficult or missed turn rarely creates a source cascade.
Advanced Exit Criterion 8 | Transfer
Conversation skills survive unfamiliar speakers, topics, registers and exam question wording.
Advanced Exit Criterion 9 | Answer scope
The learner can produce the minimum required answer without losing source, condition or degree.
Advanced Exit Criterion 10 | Independence
The source-and-stance model self-triggers without teacher reminders in a mixed listening paper.
Advanced closing synthesis | The conversation is a moving relationship map
The mature listener does not chase every voice equally. She builds a compact, continuously updated relationship map. Who owns this claim? Is the response agreement, qualification, correction or refusal? Did the reason change? Did the decision change? Is the interviewer summarising accurately or being corrected? Is the anecdote evidence, explanation or merely context? These questions keep the conversation coherent even when the language is fast, the topic is unfamiliar or the stance is subtle.
The deepest shift is from hearing alternating speakers to tracking changing propositions between people. That shift makes dialogue portable. Once the learner can maintain ownership, relation, status and trajectory, the same system can handle a service exchange, an interview, a classroom discussion, a debate or a panel. Surface variety increases; the operating model remains recognisable.