TOEIC Link Reading — Textual Cohesion and Lexical Chains Tracking: The Discourse-Map Discipline That Compresses Multi-Paragraph Passages Into Rubric-Answerable Structure

The band-21 reading candidate fails on long passages for a reason that is structural rather than vocabulary-related. The candidate has the sentence-level reading skill to understand each sentence in isolation, but the candidate is processing the passage as a sequence of independent sentences and is exhausting working memory before reaching the comprehension questions. The exhausted-working-memory failure mode produces a characteristic pattern in the score profile — the candidate scores well on early-passage questions where memory load is low, scores worse on late-passage questions where load is high, and scores worst on cross-paragraph questions that require integrating information from multiple paragraphs simultaneously. The pattern is diagnostic of the candidate operating without a textual-cohesion tracking discipline.

The band-23 candidate is doing something different. The candidate is tracking the passage's cohesion structure during reading, identifying the lexical chains that link sentences across paragraphs, and constructing a compressed discourse map that fits in working memory regardless of passage length. When the candidate reaches the comprehension questions, the questions are not being answered against the surface text — they are being answered against the compressed map, and the map preserves exactly the structural relationships that the questions are designed to test. The compression is what makes the band-23 score achievable on the long-passage items; without it, the score ceiling is set by working-memory capacity rather than by reading skill. This guide formalizes the four cohesion chain types, the real-time tracking discipline, and the installation drill that produces transfer to test-equivalent reading.

Why working memory is the binding constraint on long-passage scoring

The TOEIC Link reading section assesses reading comprehension across passages that range from single paragraphs to multi-paragraph passages of roughly 300-500 words. The score profile across candidate populations shows that the band-21 ceiling is reached primarily on the longer passages, and within the longer passages, the binding constraint is the candidate's ability to retain cross-paragraph structural information through the duration of the passage.

The cognitive mechanism is the limited capacity of working memory to retain unstructured information. Working memory in a reading context can hold roughly seven independent information units simultaneously, and the units decay rapidly without active maintenance. A 400-word passage contains roughly 30 to 40 sentences, which produces 30 to 40 independent information units if the candidate is processing each sentence as an isolated unit. The processing model exceeds working-memory capacity by a factor of four to six, and the resulting overflow produces the cross-paragraph comprehension failure that the rubric scores as a band-21 outcome.

The textual-cohesion tracking discipline addresses the overflow by compressing the 30-to-40-unit sequence into a structure that fits within working-memory capacity. The compression mechanism is the identification of lexical chains and cohesion ties that link sentences across the passage, which collapses sequences of sentences into chain-units that the working memory can hold as single units. A passage that decomposes into four lexical chains is being tracked as four working-memory units rather than 30, which lifts the candidate from exhausted-memory failure to compressed-map success.

The four cohesion chain types

Cohesion chains in a typical TOEIC Link reading passage fall into four types that have to be tracked separately because they perform different discourse functions and the comprehension questions target each type with characteristic question patterns.

Type 1 — Repetition chain

The repetition chain is a sequence of sentences that share an explicit lexical item — the same noun, verb, or noun phrase appears in multiple sentences across the passage. The repetition chain is the easiest chain type to detect because the shared lexical item is visible at the surface, and the chain tracking does not require the candidate to identify semantic relationships. Repetition chains typically correspond to the passage's primary topic — the entity the passage is about — and the comprehension questions that target repetition chains are typically asking about the entity's attributes, actions, or development across the passage.

Type 2 — Synonym chain

The synonym chain is a sequence of sentences that share semantically equivalent lexical items — synonyms, paraphrases, or hyponyms of a common concept. The synonym chain is harder to detect than the repetition chain because the surface lexical items differ, and the chain tracking requires the candidate to recognize the semantic equivalence. Synonym chains typically correspond to a secondary topic in the passage — an entity, concept, or theme that the passage develops through varied lexical expression. Comprehension questions that target synonym chains are typically asking about the candidate's ability to recognize that the same concept is being discussed under different lexical surfaces, which is why these questions are sensitive to vocabulary breadth even though they are scored as reading-comprehension items rather than vocabulary items.

Type 3 — Reference chain

The reference chain is a sequence of sentences linked by anaphoric or cataphoric reference — pronouns, demonstratives, or definite-article noun phrases whose referent is established earlier or later in the passage. The reference chain is the foundation of intra-passage coherence and is the chain type that the comprehension questions target most frequently. Tracking reference chains requires the candidate to maintain referent identification across paragraph boundaries, which is the working-memory load that the compressed-map discipline most directly addresses. For the specific resolution discipline that supports reference tracking, see the reading anaphora and cataphora resolution strategy guide.

Type 4 — Conjunction chain

The conjunction chain is a sequence of sentences linked by explicit discourse connectives — however, therefore, as a result, on the other hand. The conjunction chain is the easiest chain type to track because the discourse connective makes the link explicit at the surface, and the chain structure is signaled directly to the reader. Conjunction chains typically correspond to the passage's argument structure — the logical relationships among the passage's claims — and the comprehension questions that target conjunction chains are typically asking about the candidate's understanding of causation, contrast, or argument flow rather than about specific facts.

The real-time tracking discipline

The four chain types have to be tracked during the first reading pass, because there is not enough time on the test to do a separate analytical pass after the initial reading. The tracking discipline below has been calibrated to fit inside the standard reading-pass timing and to produce a compressed discourse map by the time the candidate reaches the comprehension questions.

The discipline has three components. The first is the chain anchor identification at paragraph boundaries — at the end of each paragraph, the candidate identifies the one or two lexical items that link the paragraph to the next paragraph (or to a previous paragraph), and stores the anchor as a working-memory unit. The second is the chain type classification — for each anchor, the candidate classifies the chain type (repetition, synonym, reference, conjunction) and uses the classification to predict which comprehension-question types are likely to target the chain. The third is the chain pruning at passage end — at the end of the passage, the candidate prunes chains that did not produce comprehension-question-relevant structure and consolidates the remaining chains into a three-to-five-unit compressed map. The three-to-five-unit target is calibrated to working-memory capacity and produces a map that can be held actively through the question-answering phase. For the question-stem mapping discipline that operates on top of the compressed map, see the reading question stem keyword mapping guide.

The four-week installation drill

Real-time chain tracking is a discipline that has to be installed through deliberate practice, because the chain-detection sensitivity is initially low and the working-memory cost of running the discipline is initially high. The four-week drill below produces transfer to test-equivalent passage processing.

Week one focuses on repetition and conjunction chain detection because these are the surface-visible chain types and produce the fastest detection sensitivity gains. The candidate reads single-paragraph passages, marks repetition and conjunction chains with manual annotation, and verifies the chain identification against the comprehension questions that the chain types predict. The week-one exit criterion is the candidate's ability to detect repetition and conjunction chains without conscious search in single-paragraph passages.

Week two adds synonym chain detection. The candidate reads two-to-three-paragraph passages and detects synonym chains alongside the repetition and conjunction chains established in week one. The week-two drill is more cognitively demanding because synonym detection requires active semantic processing rather than surface matching. The week-two exit criterion is the candidate's ability to detect at least one synonym chain per passage across five consecutive passages.

Week three adds reference chain detection and the full four-chain tracking discipline. The candidate reads full-length passages and runs the complete chain-anchor identification at every paragraph boundary. The week-three drill emphasizes the chain pruning step — the candidate produces an explicit compressed map at the passage end and verifies that the map preserves the relationships that the comprehension questions target. The exit criterion is three consecutive full-length passages with a compressed map that correctly predicts the comprehension question structure.

Week four moves the discipline into timed practice. The candidate runs the full discipline under test-equivalent timing and measures the comprehension-question accuracy on the long-passage items. The week-four exit criterion is a measurable improvement in long-passage accuracy across three consecutive practice sessions, with particular attention to cross-paragraph questions where the compressed map produces the largest score uplift.

How chain tracking interacts with the rest of the reading module

Textual cohesion tracking is the discourse-level sub-skill that complements the sentence-level reading skills the candidate brings to the test. The sentence-level skills determine the candidate's accuracy on within-sentence comprehension questions; the cohesion-tracking sub-skill determines the candidate's accuracy on cross-sentence and cross-paragraph questions. The two sub-skill layers are largely independent, and the band-21-to-23 transition almost always requires the discourse-level layer to be installed because the within-sentence layer has typically already reached its ceiling at the band-21 level.

The candidate who installs cohesion tracking without the surrounding reading sub-skills will see a structural shift on the long-passage items while the short-passage scores remain stable. The candidate who installs cohesion tracking alongside the anaphora-resolution discipline and the question-stem mapping discipline sees the discourse-level layer fully integrate with the reading-comprehension scoring frame, and the long-passage items move from being the binding constraint on the reading score to being a strength that lifts the overall reading score above the sentence-level ceiling. The cohesion-tracking layer is the most leverage-rich installation target in the reading module at the band-21-to-23 boundary.