The Construction–Integration model is Walter Kintsch’s influential account of how readers develop coherent understanding from words, propositions and prior knowledge. The 1988 model distinguishes a broad construction phase, which generates meanings and possible connections, from an integration phase, in which contextually compatible elements support a stable interpretation. It explains why individual vocabulary recognition is only one step toward comprehension.
Did you know that a familiar word can make several meanings available before the sentence tells us which one matters? A reader seeing bank may consider both a financial institution and a river edge. The surrounding propositions help select the relevant sense. Vocabulary therefore works as a technology of meaning selection: readers must combine lexical knowledge with grammar, context and evidence rather than merely recite definitions.
This guide explains Kintsch’s 1988 original research, how a textbase differs from a situation model, what activation and integration contribute, and why inference can both help and mislead. It uses an original classroom case involving a library book and extends the framework to Primary and Secondary English, Mathematics, Science, History, General Paper, translation and professional communication.
Editorial note: All classroom situations and explanatory examples are original and are not presented as results of a new experiment. The model is a scientific account with alternative theories and methodological limits, not a diagnostic instrument for individual learners.
Contents: from possible word meanings to a coherent interpretation
Open 50 chapters and teaching checks
- 1. The model asks how words become a coherent understanding
- 2. Walter Kintsch proposed an influential 1988 formulation
- 3. Construction is deliberately broader than the final answer
- 4. Integration tests how pieces fit together
- 5. Bottom-up has a specific meaning in the 1988 account
- 6. Words contribute meaning but are not enough
- 7. A proposition represents an expressed relationship
- 8. Surface wording and proposition are different
- 9. A textbase preserves the expressed meaning structure
- 10. A situation model represents the described world
- 11. Textbase and situation model serve different tests
- 12. Knowledge supplies candidates for inference
- 13. The same form can activate several senses
- 14. Context can also be insufficient
- 15. Semantic associations are useful but not equal
- 16. A bridging inference can connect adjacent statements
- 17. Elaboration goes beyond what was explicitly stated
- 18. Coherence is the target of integration
- 19. Activation can spread through related concepts
- 20. A highly related idea can become accessible
- 21. Irrelevant candidates can lose influence
- 22. Competition between interpretations is normal
- 23. Integration can be modelled across processing cycles
- 24. Local coherence connects nearby propositions
- 25. Global coherence connects distant passages
- 26. A topic sentence is a useful but incomplete cue
- 27. A contradiction can prevent smooth integration
- 28. Cognitive models do not certify factual truth
- 29. The 1988 model included arithmetic word problems
- 30. A problem model turns the situation into mathematics
- 31. Arithmetic keywords are not enough
- 32. An original three-sentence experiment example
- 33. A role reversal changes the proposition
- 34. A negation should survive paraphrase
- 35. Primary English: connect pronouns correctly
- 36. Primary English: link cause with evidence
- 37. Secondary English: identify the author’s larger claim
- 38. Science: a correct word can carry a wrong explanation
- 39. History: distinguish source from inference
- 40. General Paper: coherence can disguise a bad argument
- 41. Translation: preserve semantic roles and scope
- 42. Professional notices: a coherent guess is not approval
- 43. The model differs from the event-indexing framework
- 44. The model differs from the Structure Building Framework
- 45. The model differs from memory-based resonance
- 46. The Landscape Model makes time-course predictions differently
- 47. A computational model is not a pupil’s mind scan
- 48. The theory has boundaries and alternatives
- 49. An eight-column research notebook can test claims
- 50. The closing skill is accurate independent use
Quick Test: five different parts of comprehension
| Element | Question to ask | What must not be confused |
|---|---|---|
| Word form | Which printed or spoken lexical item is present? | Form recognition is not full understanding |
| Candidate senses | Which meanings are initially possible? | Every activated meaning is not necessarily intended |
| Proposition | Who did what, under which conditions? | A bag of nouns is not a sentence meaning |
| Textbase | What relationships were actually expressed? | Exact proposition structure is not the same as all imagined details |
| Situation model | What coherent state of affairs is described? | A plausible inferred event is not automatically a textual fact |
This table describes educational analogies to the model’s conceptual distinctions. The scientific research uses computational representations and controlled behavioural observations; this classroom guide does not record anyone’s actual activation levels.
1. The model asks how words become a coherent understanding
The Construction–Integration model, often shortened to CI, explains how a reader can move from individual word meanings and sentences toward a coherent representation of a passage. Its central distinction is between constructing candidate meanings and connections, then integrating mutually supportive information into a more stable interpretation.
The model is not simply advice to make a mind map after reading. It is a computational and psychological theory of discourse comprehension. Its predictions concern which propositions become activated, how knowledge participates and why some candidate interpretations become more accessible than others. For a student, the practical question is whether the final understanding faithfully represents the actual text.
2. Walter Kintsch proposed an influential 1988 formulation
Walter Kintsch published The Role of Knowledge in Discourse Comprehension: A Construction–Integration Model in Psychological Review in 1988. The paper described an initial relatively bottom-up construction process that activates meanings, builds propositions and generates elaborations without first imposing one complete contextually selected interpretation.
The resulting network is then integrated through activation among related elements. The model was offered as a theoretical and computational explanation of language comprehension, not as a claim that readers consciously perform a two-step classroom ritual every time they encounter a word. Its original paper also discussed arithmetic word-problem comprehension.
3. Construction is deliberately broader than the final answer
In the construction phase, a word may activate multiple possible senses and associated knowledge. The first set can include candidates that will later prove irrelevant to the passage. Constructing alternatives is useful because the reader cannot always know the intended interpretation from an isolated word.
An important consequence is that early activation is not the same as final belief. Someone encountering bank might briefly consider a river edge and a financial institution. The sentence that follows provides evidence for choosing among those possibilities. A student should not treat every early thought as a conclusion the author endorsed.
4. Integration tests how pieces fit together
The integration phase allows related propositions and contextual cues to support one another. Connections that fit the developing discourse tend to become more coherent within the model, while inconsistent or poorly connected candidates lose relative influence.
This is a theory of activation and constraint satisfaction, not a guarantee that human readers always reach the objectively correct interpretation. Misunderstandings can occur when background knowledge is wrong or the text is ambiguous. Educational practice still needs explicit paraphrase, reference checks and, where relevant, factual verification.
5. Bottom-up has a specific meaning in the 1988 account
Kintsch’s original proposal contrasted its initial broad construction with models relying heavily on advance contextual prediction. In that formulation, meanings and possible inferences could first be activated without being tightly restricted to the interpretation ultimately favoured by the discourse.
This does not mean the brain is permanently blind to context or that no other reading model incorporates prediction. The broader literature includes interactive and expectancy-based theories. Describing the CI model accurately requires distinguishing a particular theoretical architecture from a universal claim about every human thought.
6. Words contribute meaning but are not enough
A vocabulary item provides potential lexical information. Yet a passage also contains relations among agents, actions, objects, conditions and time. A list of dictionary definitions does not automatically produce those relationships.
Consider ‘The scientist heated the sample’ and ‘The sample heated the scientist’. The same content words appear, but the propositions differ because grammatical roles change. The model’s focus on coherent representations encourages learners to identify who did what rather than memorise word meanings in isolation.
7. A proposition represents an expressed relationship
In discourse modelling, a proposition is a unit of expressed meaning, commonly representing a predicate and its participants or other semantic relations. A sentence may express several connected propositions rather than one indivisible idea.
The sentence ‘Mira borrowed a book because she needed evidence’ includes a borrowing event and a reason for it. A student who remembers only Mira and book has lost the causal or explanatory link. Proposition-based reading helps retain the relationship that matters.
8. Surface wording and proposition are different
Two sentences can express a similar event using different grammatical forms. ‘The nurse examined the patient’ and ‘The patient was examined by the nurse’ preserve the agent and affected person while changing the surface arrangement.
A comprehension model can represent their common proposition without claiming their exact wording is identical. That distinction is useful in paraphrasing and translation. It must not be abused where exact quotation, technical terminology or the force of a legal clause matters.
9. A textbase preserves the expressed meaning structure
The textbase is commonly described as the representation of propositions conveyed by the written or spoken text, including their connections at local and global levels. It differs from merely retaining the original word sequence.
Knowing the textbase lets a reader paraphrase the author’s explicit relationships without copying every sentence. But a textbase does not guarantee complete understanding of a scientific or historical situation. The model also considers how knowledge contributes to a richer situation model.
10. A situation model represents the described world
A situation model includes information about the entities, events and relationships that the text describes, supplemented by relevant prior knowledge. It can help readers reason about what happened beyond a verbatim sentence list.
If a passage says a character carried a book from the library to a classroom, a reader can update where the book is. This is more than lexical recognition. Yet any inferred details, such as which bus route was used, should remain uncertain unless the passage or another source supplies them.
11. Textbase and situation model serve different tests
A reader may accurately repeat the proposition ‘The books were moved because the roof leaked’ but misunderstand how water damage occurs. Another reader may understand the mechanism while not remembering the exact phrase.
Researchers can design questions about expressed propositions and situation-level inferences to distinguish aspects of comprehension. Teachers should similarly ask students to state what the passage actually said and what background knowledge they used to explain it. Neither outcome is a substitute for the other.
12. Knowledge supplies candidates for inference
Long-term knowledge helps people recognise ordinary events: keys can open locks, books may be borrowed and heat can change matter. That knowledge can help connect separate propositions into a meaningful situation.
Kintsch’s account describes activation of relevant knowledge during construction and its selection through integration. The same process can mislead if a learner assumes a familiar event occurred when the writer never stated it. Background knowledge is an interpretive resource, not independent proof that a new claim is true.
13. The same form can activate several senses
Polysemous words can evoke related or distinct meanings. Charge may concern a fee, electricity, an allegation or an action of moving forward. Early lexical activation can be broad before the sentence narrows the intended sense.
Consider ‘The technician measured the charge’. Scientific context makes an electrical interpretation plausible, while ‘The hotel added a charge’ favours a fee. A student should verify the domain rather than choose the first familiar definition. Contextual integration is where vocabulary becomes precise.
14. Context can also be insufficient
The sentence ‘The manager reviewed the charge’ may remain ambiguous without further information. A financial fee and a formal accusation could both be possible in some settings.
A coherent reading can therefore contain unresolved alternatives. A model describing integration should not be treated as an automatic ambiguity eraser. In education, honest interpretation includes saying which information is missing and asking a targeted follow-up question instead of inventing certainty.
15. Semantic associations are useful but not equal
The words rain and umbrella are associated through a familiar activity, but the objects are not synonyms. Constructing a network can activate the association without establishing that someone opened an umbrella in a particular event.
A well-integrated interpretation respects the grammatical and causal evidence actually available. This helps students distinguish semantic relatedness from proposition truth. It also guards against incorrect substitutions in essays, where strongly associated words can have different functions.
16. A bridging inference can connect adjacent statements
Imagine the text says ‘Mira put the keys on the counter. The door remained locked.’ A reader may consider how the keys and door are related and infer a possible missing connection, depending on surrounding context.
Some bridging inferences are necessary to make a passage coherent, but they still need support. The CI account offers a framework for how relevant knowledge becomes active. A teacher should distinguish a logically required link from an attractive but unconfirmed addition to the narrative.
17. Elaboration goes beyond what was explicitly stated
An elaboration adds knowledge or a plausible detail related to the current material. A reader may imagine a noisy school library because they have visited one, although the passage only mentions a quiet reading room.
Such elaborations can help memory or imagery in some tasks, but they can also contaminate a factual summary. The useful skill is to label direct statements, justified inferences and imaginative additions separately. Construction can generate possibilities; disciplined integration should not turn all possibilities into facts.
18. Coherence is the target of integration
A coherent representation allows related propositions to fit into a mutually understandable whole. The referent of a pronoun, the cause of an event and the meaning of an ambiguous word should support one another rather than contradict the passage.
Coherence is not the same as truth. Fiction can be internally coherent while describing impossible technology. An inaccurate scientific explanation can also feel coherent because its premises resemble a familiar story. A reader must evaluate the representation against both text and evidence.
19. Activation can spread through related concepts
The CI architecture uses a network of connected elements and an integration process often described in terms of spreading activation or activation dynamics. Related elements can support one another as the representation settles.
Students should not interpret a diagram of nodes as a photograph of literal neurons or a record of consciously flashing ideas. The network is a computational representation. Its usefulness lies in predicting patterns of comprehension and retrieval under defined theoretical assumptions.
20. A highly related idea can become accessible
If a paragraph concerns a library, concepts such as books, borrowing and shelves may be activated because they connect with the described situation. A later sentence about finding a reference can draw upon those relations.
Activation alone does not guarantee that the reader’s choice is relevant. A shelf can hold many objects, and a library can host events unrelated to borrowing. The actual discourse must select which associations contribute to the interpretation. Strong vocabulary means using context to filter possibilities.
21. Irrelevant candidates can lose influence
Integration can favour candidates that fit the full passage over candidates that conflict with other propositions. If the story says the bank flooded after the river rose, the river-edge sense gains support while an unrelated financial reading becomes less useful.
This should not be paraphrased as a rule that all unchosen knowledge disappears permanently. A person’s financial knowledge remains available in memory for later contexts. The process concerns relative activation in the current discourse representation, not deletion of a dictionary entry.
22. Competition between interpretations is normal
Early reading may leave several candidate meanings active. As the sentence unfolds, grammatical agreement, lexical partners and domain knowledge can narrow the available interpretations.
A reader who briefly entertains another sense has not failed. The educational problem arises when the wrong candidate survives despite contrary evidence. Ask students to identify the word that disambiguated the expression and explain why the final reading fits the entire sentence.
23. Integration can be modelled across processing cycles
Computational descriptions of comprehension often treat successive text segments as processing cycles. Newly presented material activates associated concepts while earlier information may remain influential or become less accessible.
The exact equations and parameter values are theoretical choices, not universal human constants. A classroom flowchart can represent the logic without pretending to simulate a particular reader’s brain. A strong scientific explanation makes the computational model’s assumptions explicit.
24. Local coherence connects nearby propositions
Local coherence concerns how successive sentences or clauses link. A pronoun should have a plausible antecedent, and a new action should connect with the developing scene or be introduced clearly.
Consider ‘The glass fell. It shattered.’ The pronoun refers to the glass, and the sequence suggests a possible cause. In a Science or History task, the causal mechanism needs proper support. A locally coherent text may still have an unsupported global argument.
25. Global coherence connects distant passages
Global coherence concerns the larger theme, goal or explanatory structure across a passage. A paragraph about borrowing a book can relate to a later paragraph about preparing a school presentation.
A pupil who only understands adjacent sentences may miss the reason the author included the earlier event. An integrated situation model helps retain the wider purpose. A good summary preserves the overarching claim while avoiding invented details that the passage did not state.
26. A topic sentence is a useful but incomplete cue
Topic sentences can announce a paragraph’s central claim and help readers organise later details. The presence of a clear topic reduces some inferential work, but it does not automatically establish every relation in the paragraph.
A student should identify how examples, causes and qualifications support the claim. The CI framework concerns coherence among propositions rather than a mechanical template in which the first sentence must always contain the complete meaning. Some genres deliberately reveal their theme later.
27. A contradiction can prevent smooth integration
A text may first state that Mira left the book in the classroom, then claim she was holding it at the library without describing any intervening movement. The reader may detect a conflict in object location.
A coherent model would need additional information or a correction. The right response is not to invent a journey with complete confidence. Instead, mark the inconsistency, check the source and consider whether a later explanation resolves it. Vocabulary is a tool for managing uncertainty.
28. Cognitive models do not certify factual truth
A false claim can connect smoothly with familiar knowledge and be easy to integrate. A surprising true result can demand more processing because it conflicts with the reader’s previous expectations.
Therefore, one cannot use the ease of construction–integration as a lie detector or a substitute for experimental evidence. In school, distinguishing comprehension from critical evaluation is essential. A pupil may understand a false argument perfectly and still need to challenge it.
29. The 1988 model included arithmetic word problems
Kintsch’s original paper discussed a simulation of arithmetic word-problem comprehension. A verbal problem requires more than recognising numbers and operation words; it requires building an appropriate representation of entities and relationships.
This gives the model particular educational relevance. A child can understand that a character has fewer coins after giving some away yet still choose an incorrect arithmetic operation if the quantities were misassigned. Text comprehension and solving the formal problem are connected but distinct stages.
30. A problem model turns the situation into mathematics
In discussions of mathematical word-problem understanding, researchers distinguish the textbase, the situation model and a problem model that maps the described relationships into formal quantities or equations.
For example, a story may state that Mira has twelve books, gives away four and receives two. The situation model tracks possession changes; the mathematical model applies subtraction and addition with the correct referents. A student can have one representation right while making a calculation error in another.
31. Arithmetic keywords are not enough
Teaching pupils to select an operation whenever they see more or less can fail when the relationship is comparative or the question asks for an original rather than current amount.
The CI approach motivates reconstructing the actual proposition before calculating. A pupil should state who owns which quantity, what changed and which unknown is requested. This is more reliable than a list of memorised word–operation pairings.
32. An original three-sentence experiment example
Consider a fictional story: ‘Mira borrowed a book. She returned it on Friday. The librarian added it to the shelf.’ A reader can identify the same object across all three clauses and represent its location and ownership over time.
The exercise makes integration explicit. The pronoun it and later noun phrase connect to the book rather than introducing new objects. A student can test the model by answering where the book was after each event without inventing an unstated action.
33. A role reversal changes the proposition
The sentence ‘The teacher praised the student’ describes a different event from ‘The student praised the teacher’. An integration process that relies only on related words could confuse them.
A teacher can have pupils assign agent and patient roles for both. The role map protects accurate meaning even when the story’s likely social relationship suggests one direction. Vocabulary technology depends on grammar as well as semantic association.
34. A negation should survive paraphrase
‘The experiment succeeded’ and ‘The experiment did not succeed’ share most lexical content while expressing opposite status claims. A coherent representation must preserve the negative operator where it belongs.
A student who mentally drops not may still produce a fluent explanation that contradicts the text. The repair is to highlight the full proposition and verify its condition. Integration should never mean smoothing away a word that changes truth conditions.
35. Primary English: connect pronouns correctly
A Primary pupil might read ‘Mira gave Ben a notebook. He placed it in his bag.’ The tutor asks who he refers to and what it denotes, using the surrounding story and grammar.
Drawing arrows between the pronouns and their referents can make the propositions coherent. The student then writes a new sentence with unambiguous pronouns. The classroom task illustrates integration without claiming to recreate the computational theory or measure neural activation.
36. Primary English: link cause with evidence
A simple passage can say that rain began and Mira closed the window. The pupil may infer she wanted to keep water out, but the text may not explicitly provide that motive.
Ask which sentence states the action and which sentence would be needed to confirm the goal. This teaches pupils to integrate meaning while keeping inference separate from fact. The distinction supports both comprehension and accurate composition.
37. Secondary English: identify the author’s larger claim
A Secondary essay paragraph may introduce an example, explain a consequence and end with a qualification. A student who remembers only the vivid example may miss the actual argument.
Ask for the topic proposition, supporting evidence and limits of the claim. A coherent summary should preserve all three. Vocabulary such as however, therefore and unless helps because these connectives specify relations between propositions rather than merely decorating sentences.
38. Science: a correct word can carry a wrong explanation
A pupil might know that evaporation involves liquid water changing into vapour but wrongly claim that the process always requires boiling. The vocabulary form is familiar while the integrated scientific model is inaccurate.
Teach the relevant mechanism, conditions and examples. A textbase check establishes what the passage states; a Science check establishes whether the physical account is correct. Construction–integration does not replace experiment or subject-specific authority.
39. History: distinguish source from inference
A historical report may quote a witness, then include a historian’s interpretation. Readers must keep the speaker, event and supporting documents distinct in their representation.
A smooth narrative does not establish which claim has been corroborated. A student should mark direct testimony, reported action and later analytical inference. The ability to integrate these without conflating their evidential status is a strong form of historical literacy.
40. General Paper: coherence can disguise a bad argument
An essay can use familiar concepts and natural connectives to produce a persuasive argument whose evidence is inadequate. A reader may integrate its propositions easily yet have no independent basis for accepting its conclusion.
A GP student should identify premises, conclusions and assumptions, then evaluate the sources. The distinction between understanding an argument and endorsing its truth is crucial. Vocabulary should help express both the claim and the grounds for confidence.
41. Translation: preserve semantic roles and scope
A translation can use natural target-language words yet shift who performed an action, which condition applied or whether a statement expressed a possibility rather than certainty.
An integrated source representation helps the translator identify the relationships to preserve. The target sentence should be checked against the source textbase and situation model. A plausible paraphrase is not accurate if it adds an actor or erases a negative.
42. Professional notices: a coherent guess is not approval
A notice can state that a request was received, reviewed or approved. The three words belong to the same workflow and may activate similar knowledge, but they describe distinct procedural statuses.
A worker must verify the current authoritative record before acting. An integrated understanding of the workflow does not grant permission or turn a pending decision into an approval. Vocabulary is most valuable when its distinctions guide the correct action.
43. The model differs from the event-indexing framework
The event-indexing model describes how narrative events connect across time, place, participant, causality and intention. CI offers a broader account of how candidate propositions and knowledge are activated and integrated into a coherent discourse representation.
The frameworks are related but not interchangeable. Event indexing provides dimensions for tracking a situation; construction–integration proposes a processing architecture for building coherent meaning. Comparing them requires stating which scientific question each model addresses.
44. The model differs from the Structure Building Framework
Gernsbacher’s Structure Building Framework emphasises laying foundations, mapping coherent information onto structures and shifting to new substructures when incoming material does not fit. It also discusses enhancement and suppression mechanisms.
CI instead foregrounds a broad construction of candidate elements and their integration through network activation. Both address coherence and comprehension, but their theoretical units and mechanisms differ. A student should not assume similarly named stages are identical mental operations.
45. The model differs from memory-based resonance
The resonance model describes how concepts in the current sentence can reactivate related information from earlier discourse through a relatively passive memory process. It focuses on what becomes available when new cues overlap with prior representations.
CI also involves activation, but its account emphasises construction and integration into a coherent representation. Reactivation is not the same as deciding which interpretation is ultimately consistent. An earlier detail can become accessible without being accepted as the correct explanation.
46. The Landscape Model makes time-course predictions differently
The Landscape Model represents changing activation of concepts over successive reading cycles and connects online comprehension with the memory representation remaining after reading. It incorporates memory-based activation and coherence-seeking processes.
Kintsch’s CI model provides a related but distinct framework for constructing candidate propositions and integrating them. A research article should name the actual model and measured outcome rather than treating every activation diagram as one universal theory of reading.
47. A computational model is not a pupil’s mind scan
A network simulation can show how specified input and connection strengths yield one pattern of activation. That result is a prediction of the model under its assumptions rather than an observation of an individual’s private thoughts.
Evidence for the model comes from comparisons with behavioural research and explanatory success across tasks. Teachers should not promise that a classroom worksheet literally measures the brain’s integration process. The worksheet can still teach students to check proposition accuracy and coherence.
48. The theory has boundaries and alternatives
Not every aspect of human comprehension can be reduced to one processing stage, and other models emphasise prediction, memory resonance, situation-model updating or structural shifting. The empirical literature compares these ideas and refines their assumptions.
A strong guide therefore distinguishes the original 1988 formulation, later theoretical developments and practical classroom analogies. It is misleading to promote CI as a solved account of every reading error, every language or every reader. Scientific literacy includes knowing where a model stops.
49. An eight-column research notebook can test claims
For a classroom inquiry, record the exact sentence, important words, possible senses, expressed propositions, relevant prior knowledge, inferred links, final interpretation and any unsupported additions.
The columns reveal whether the pupil merely generated a plausible association or correctly integrated the passage. A formal experiment would require controlled stimuli, participants and appropriate measures, but the classroom task can still improve precision. A good interpretation explains what the text supports and what remains uncertain.
50. The closing skill is accurate independent use
After analysing a paragraph, ask the learner to state the exact claim, identify its grammatical roles and paraphrase the conclusion without omitting negation or conditions. Then give a fresh paragraph containing a new ambiguous word.
The transfer test matters more than remembering the model’s name. Construction–Integration research teaches that comprehension emerges from relations among words and knowledge. Vocabulary becomes technology when students build meaning that is coherent, faithful to the source and open to correction.
Full case: three sentences, one connected event
Consider the invented story: ‘Mira borrowed a book from the library. She carried it into the classroom. Ben returned the book on Friday.’ A coherent account connects she with Mira and it with the same book. It also tracks where the book moved and who returned it. The words are familiar, but their references must be integrated.
A pupil who says Mira returned the book has recognised the event but assigned the wrong actor. Ask the pupil to identify each proposition and underline the grammatical evidence for the participant, then paraphrase the whole story. The final interpretation should preserve Ben’s return rather than a more familiar inferred action.
Eight-stage construction–integration classroom workbook
- Read the exact sentence and identify every potentially ambiguous content word.
- List candidate meanings without immediately treating every one as correct.
- Mark grammatical subjects, objects, negatives and conditions.
- Write the propositions conveyed by the actual sentences.
- Connect pronouns and repeated references to the appropriate entities.
- Add necessary bridging inferences while labelling any assumptions.
- Reject unsupported alternatives and test coherence with the full passage.
- Paraphrase a fresh passage independently, preserving the author’s claim and limits.
Frequently asked questions
1. What is the Construction–Integration model?
A theory explaining how readers construct possible meanings and integrate compatible propositions into coherent discourse representations.
2. Who proposed the influential version?
Walter Kintsch in his 1988 Psychological Review article.
3. What happens during construction?
Meanings, propositions and possible inferences become available relatively broadly in the original model.
4. What happens during integration?
Mutually supportive elements become more coherent through activation within the constructed network.
5. What is a proposition?
A represented meaning relationship, such as an action and its participants, rather than merely isolated words.
6. What is a textbase?
A representation of propositions expressed by a text and their relationships.
7. What is a situation model?
An integrated representation of the described state of affairs that may incorporate relevant prior knowledge.
8. Does a coherent interpretation prove a statement true?
No. Factual claims require independent evidence.
9. Is this the same as the event-indexing model?
No. Event indexing emphasises dimensions of narrative continuity; CI emphasises a processing architecture.
10. What is the classroom goal?
Accurate word senses and grammatical relationships combined into a faithful, justified interpretation.
Original research and connected vocabulary guides
- Kintsch (1988): The Role of Knowledge in Discourse Comprehension — A Construction–Integration Model — original model, Psychological Review, 95(2), 163–182.
- ERIC record of Kintsch (1988) — original article summary and study scope.
- Publisher DOI for Kintsch (1988) — stable bibliographic identifier.
- Van den Broek, Rapp and Kendeou (2005): Integrating Memory-Based and Constructionist Processes — relationship between coherence-seeking and automatic memory activation.
- Gernsbacher (1995): The Structure Building Framework — a distinct comprehension architecture.
- Tzeng, van den Broek, Kendeou and Lee (2005): The Computational Implementation of the Landscape Model — another account of online activation and resulting memory.
- Vocabulary | What Is Situation-Model Updating? — what it means for the described situation to change.
- Vocabulary | What Is the Event-Indexing Model? — dimensions of narrative continuity.
- Vocabulary | What Is Good-Enough Sentence Processing? — plausible interpretations can still be wrong.
- Vocabulary Learning Hub — connect precise words with effective thinking.
The Construction–Integration model shows why meaningful reading is more than naming vocabulary words quickly. Readers must select among possible meanings, construct accurate propositions and integrate those relationships with context and justified knowledge. The practical success is a coherent interpretation that remains faithful to the actual words and open to correction.