How English Works | Text-Model Construction: How Readers Turn Sentences Into a World They Can Reason About

Read these sentences:

The lights went out. Amir reached for his phone. A few seconds later, the emergency lamps came on.

A competent reader does not store three isolated statements.

The mind builds a small world:

• there was a loss of normal power
• Amir reacted to the darkness
• the phone was probably being used as a light or information tool
• a backup lighting system then activated
• the events belong to one continuous episode

Some of that is stated. Some is inferred. All of it is organised.

That organisation is what makes reading more than word recognition.

The shortest useful definition

Text-model construction is the process by which a reader integrates sentence meanings with reference, sequence, causality, goals, background knowledge and justified inference to build one coherent representation of what the text is describing or arguing.

AI Extraction Box

• Mechanism: text-model construction
• Input: decoded words + sentence structure + discourse links
• State tracked: entities, events, locations, time, causes, goals, beliefs, claims
• Update rule: each sentence modifies the existing model rather than starting from zero
• Inference rule: add only what is sufficiently supported by text + context + world knowledge
• Failure mode: sentence-level understanding without passage-level integration
• Repair: identify who/what is active, what changed, why it changed and how the new sentence alters the current model

1. Literal recovery is the floor

The reader first needs enough lexical and grammatical control to recover what each clause says.

If “The committee postponed the vote” is misread as “The committee cancelled the vote”, the model begins with the wrong event.

Inference cannot repair a corrupted literal floor reliably.

2. Reference keeps the cast stable

Maya handed Aisha the folder. She opened it immediately.

The reader must resolve she and it before the event model can update correctly.

Batch 3’s Referential Accessibility explains how entities stay available across discourse.

3. Time determines event order

By the time the train arrived, the crowd had doubled.

Readers must place the crowd increase before the train’s arrival.

Batch 4’s Reference-Time Sequencing supplies the timeline machinery underneath this reading task.

4. Cause is not the same thing as sequence

The alarm sounded. The staff evacuated.

Readers will normally connect the alarm to the evacuation because the causal relation is highly plausible.

But careful reading still distinguishes explicit cause from inferred cause.

Text models need relations, but those relations must be evidentially disciplined.

5. Goals make actions intelligible

Amir checked the departure board, then began walking faster.

The text may never say he wants to catch a train. Yet the action becomes coherent if that goal is inferred from the setting.

Good reading often reconstructs goals because goals explain why characters or institutions act.

6. Beliefs can differ from reality

Maya thought the meeting started at three. It had been moved to two.

A strong reader stores two states:

• Maya’s belief: 3 p.m.
• textual reality: 2 p.m.

Collapsing belief and reality produces comprehension errors.

7. Text models operate above individual words

Batch 5’s Word-Sense Resolution explains how context selects the meaning of a word.

Text-model construction goes one scale higher: once word and sentence meanings are recovered, the reader must integrate them into a representation of the passage.

8. Text models operate above sentence cohesion

Batch 7’s Cohesive Architecture explains how sentences link linguistically.

A reader can follow every cohesive link and still miss the larger point. Text-model construction asks what whole situation or argument those links are building.

9. Inference is model completion, not free invention

The warehouse article Inference Is Not Guessing owns the applied comprehension rule.

The systems view is simple: inference fills a necessary or strongly supported gap in the current model. It should make the passage more coherent without adding unsupported detail.

10. Bridging inferences connect events

Maya dropped the glass. She swept the floor.

The reader normally bridges the gap: the glass broke, creating fragments that needed sweeping.

The text does not explicitly state every intermediate event because ordinary world knowledge makes the bridge recoverable.

11. Elaborative inferences are less secure

From the same example, imagining that the glass was expensive, blue, or emotionally important goes beyond what the text licenses.

Strong readers distinguish useful bridging from imaginative elaboration.

12. Global coherence requires hierarchy

Readers must distinguish the central situation from supporting details.

A paragraph may mention three examples, but all three may support one higher-order claim. If the reader stores the examples without the claim, comprehension remains fragmented.

13. Reading is continual prediction and correction

Readers form expectations about what likely comes next.

The road was closed, so Maya…

Several continuations become plausible: turned back, took another route, called ahead.

The next sentence selects one. Good reading updates rather than clinging to the first prediction.

14. Contradictions force model repair

If an earlier sentence suggests the shop is open and a later sentence says it closed permanently last month, the reader must re-examine reference, time or viewpoint.

Comprehension includes detecting when the current model no longer fits incoming evidence.

15. The CivDJ forward pass

decode sentence → resolve reference → place event in time → identify relation to prior state → add justified inference → update entities/goals/claims → test global coherence → continue

16. The CivDJ backward pass

1. List the active people, objects, institutions or claims.
2. Write what changed in the latest sentence.
3. Mark what is explicit and what is inferred.
4. Check time order and causal relation.
5. Identify the current goal or governing claim.
6. Remove inferences that the text does not need or support.

17. Rotate one short passage

The queue stretched beyond the entrance. Two counters were closed. By noon, the manager had called in three additional staff.

• Literal: long queue, two counters closed, extra staff called
• Likely relation: reduced counter capacity contributed to congestion
• Goal inference: manager wanted to reduce the queue
• Unsafe addition: the closed counters were caused by staff illness—possible, not stated

18. Common failure modes

• Sentence islands: each sentence is understood but not integrated.
• Reference loss: reader forgets who or what a pronoun refers to.
• Timeline collapse: events are stored in mention order rather than actual order.
• Inference inflation: plausible additions are treated as facts.
• Detail capture: memorable details displace the governing claim.
• Model persistence: reader keeps an early interpretation after later evidence contradicts it.

19. Repair route

1. Rebuild the literal floor.
2. Restore reference chains.
3. Reconstruct time and cause.
4. Separate explicit facts from inferences.
5. Identify the paragraph or passage-level owner.
6. Update the model when new evidence arrives.

20. Why this matters for students

Comprehension questions often appear to test vocabulary, inference, sequencing, cause or character motivation separately. Underneath, many of them test whether the student has built the right text model.

The durable question is: What world, event chain or argument does this passage make me build—and which parts of that model are stated versus inferred?

21. The EnglishOS reading

The canonical How English Works V1.1 treats English as recoverable coordination. Reading succeeds when the receiver reconstructs not only the words, but enough of the sender’s intended model to reason inside it.

Reading is the controlled construction of a world from language. The better the model, the less the reader depends on guesswork.

Continue Batch 8: Reading

• How English Works | Evidential Relevance
• How English Works | Contextual Integration
• How English Works | Stance Reconstruction

Return to How English Works V1.1.