TOEIC Link Speaking Discourse Cohesion and Transition Signal Deployment: The Connector-and-Reference Discipline That Keeps a Spoken Response Coherent Under Time-Pressured Production Conditions

A TOEIC Link Speaking response that demonstrates control of the relevant content but delivers the content as a sequence of disconnected sentences will score below the rubric ceiling for fluency and discourse competence even if the content layer is unimpeachable. The rubric assesses whether the candidate's response constitutes a coherent discourse — a connected, signposted, and internally referential exposition — rather than an enumerated list of independently scored sentences. The candidate who has mastered the content layer but has not mastered the discourse layer will lose score points that vocabulary and grammar drilling cannot recover, because the cohesion failure is registered against a separate rubric criterion than the criteria those drills address.

This is the discourse-cohesion gap failure mode, and it is the most common ceiling-suppressor among candidates whose underlying language proficiency is strong enough to clear the higher score bands on the content criteria. The candidate produces appropriate vocabulary, executes grammatically correct sentences, and selects relevant supporting points, but the points are delivered without the transition signals, reference chains, and discourse markers that a fluent speaker deploys automatically to signal the relationship between the points. The rater registers the response as a list rather than as an exposition, and the rater's holistic impression compresses the response into the score band that the rater associates with list-style delivery rather than the score band the content quality would otherwise unlock.

This article is the discourse-cohesion guide for TOEIC Link Speaking. The guide identifies the cohesion functions that a spoken response has to perform, the transition-signal inventory that a high-band response deploys, the reference-chain protocols that maintain entity coherence across sentences, and the rehearsal sequence that builds the deployment automaticity the timed production conditions demand.

The cohesion functions a spoken response has to perform

A spoken response is not a chain of independent sentences delivered in temporal sequence. It is a structured exposition in which each sentence stands in a specific relation to the sentences that precede it — elaboration, contrast, exemplification, conclusion, qualification — and the relations have to be signaled to the listener through linguistic devices that mark the relations explicitly. Without the explicit marking, the listener has to infer the relations from the content alone, and the inference cost reduces the listener's processing of the content itself. A rater listening at production speed will not invest the inference cost; the rater will register the response as low-cohesion and will assess the response against the band the rater associates with low-cohesion responses.

The cohesion functions fall into four primary categories, and the high-band response demonstrates competence across all four.

Function 1 — sequencing and progression. The response signals the order in which the points are being presented and the progression from one point to the next. The signals — first, second, then, next, finally — establish the response's structure and allow the listener to anticipate the response's trajectory. A response that lacks sequencing signals forces the listener to maintain a model of where the response is in its progression without external markers, and the maintenance cost suppresses the listener's content processing.

Function 2 — relational marking between points. The response signals the logical relation between consecutive points — addition, contrast, causation, exemplification, qualification, conclusion. The signals — additionally, however, because, for example, although, therefore — establish the response's argument structure and allow the listener to integrate each point into the developing exposition. A response that lacks relational signals delivers the points as a list and forces the listener to construct the relations through content inference, which suppresses the response's argument quality in the listener's assessment.

Function 3 — entity coherence across sentences. The response maintains a consistent entity representation across the sentences and uses appropriate reference forms — pronouns, definite descriptions, demonstratives — to refer back to entities the response has introduced. The entity coherence allows the listener to track the response's entities without ambiguity and to integrate the entity-related content across the response. A response that re-introduces entities at each mention or that uses ambiguous reference forms disrupts the entity tracking and forces the listener to invest disambiguation effort that suppresses the response's content processing.

Function 4 — discourse-level framing and closure. The response includes opening and closing moves that frame the response as a coherent unit — an opening that establishes the response's topic and structure, a closing that signals the response's conclusion and recaps the response's main point. The framing moves give the listener cues for response-level processing — when to expect new content, when to integrate the content into a holistic interpretation — and the cues are essential to the listener's holistic assessment.

The transition signal inventory the high-band response deploys

The candidate who has identified the cohesion functions has solved the recognition problem; the candidate has not yet solved the deployment problem. The deployment problem is the problem of having the appropriate transition signal available for production at the moment the cohesion function is required, and the deployment automaticity is built only through rehearsal of a deliberate inventory rather than through generic conversational exposure.

Inventory 1 — sequencing signals. First, second, third, finally for explicit numbered sequencing. To begin with, next, after that, lastly for narrative sequencing. Initially, subsequently, ultimately for formal sequencing. The high-band response deploys at least one sequencing signal per response and chooses the register to match the prompt type — formal sequencing for analytical prompts, narrative sequencing for descriptive prompts.

Inventory 2 — additive signals. Additionally, furthermore, moreover, in addition for adding parallel points. What is more, on top of that, beyond this for emphasizing the addition. The high-band response uses additive signals to mark the second and subsequent points in a parallel structure rather than relying on the listener to infer the parallel relation.

Inventory 3 — contrastive signals. However, on the other hand, conversely, in contrast for direct contrast. Although, despite, even though for concessive contrast embedded within sentences. Nevertheless, nonetheless, still for adversative contrast after a concession. The high-band response uses contrastive signals to mark both the direct contrasts between points and the qualifying contrasts within points.

Inventory 4 — causal signals. Because, since, as for direct causation. Therefore, consequently, as a result, thus for consequence marking. For this reason, this is why for causal explanation. The high-band response uses causal signals to mark the causal structure of the response's argument explicitly rather than implicitly.

Inventory 5 — exemplification signals. For example, for instance, such as, including for direct exemplification. Take the case of, consider the example of for elaborated examples. To illustrate, by way of example for formal exemplification. The high-band response uses exemplification signals to mark the transition from general claims to specific examples.

Inventory 6 — concluding signals. In conclusion, to summarize, overall for response-level closure. All in all, on balance, in sum for balanced conclusions. Therefore, this leads to for conclusion linked back to argument. The high-band response always includes at least one concluding signal that frames the response's closure and recaps the response's main point.

The reference-chain protocols that maintain entity coherence

The transition signal layer addresses the relational marking between sentences. The reference-chain layer addresses the entity coherence across sentences, and the layers are independent — a response can have strong transitions but weak reference chains, or vice versa, and the rater registers the deficiencies independently.

Protocol 1 — pronoun reference with disambiguated antecedents. When the response uses a pronoun, the antecedent has to be unambiguously identifiable from the immediately preceding context. If the preceding context introduces multiple entities of the same type that the pronoun could refer to, the response substitutes a definite description for the pronoun to disambiguate. The protocol prevents reference ambiguity that forces the listener to invest disambiguation effort.

Protocol 2 — definite description with consistent labeling. When the response refers to a previously introduced entity by a definite description, the description uses the same head noun the response used to introduce the entity. The protocol prevents the listener from misinterpreting the description as introducing a new entity and from losing track of the entity's role in the response.

Protocol 3 — demonstrative reference with proximal-distal control. When the response uses this, that, these, those to refer to a previously mentioned entity or proposition, the demonstrative form is selected to match the discourse proximity — this for immediately preceding content, that for content further back in the response. The protocol gives the listener a discourse-proximity cue for the antecedent that reduces the disambiguation cost.

Protocol 4 — lexical-cohesion through synonym and paraphrase chains. When the response refers to a previously discussed concept multiple times across the response, the response varies the lexical realization through synonyms and paraphrases rather than repeating the same form. The protocol prevents lexical monotony that the rater would mark down for vocabulary range while maintaining the conceptual continuity the cohesion criterion demands.

The rehearsal sequence that builds deployment automaticity

The candidate who has identified the cohesion functions and the transition inventory has solved the knowledge problem; the candidate has not yet solved the production problem. The production problem is the problem of producing the appropriate transition signals in real time under the production-time pressure of the Speaking section, where the response window is short and there is no rehearsal opportunity between the prompt presentation and the response onset.

Step 1 — single-signal mini-rehearsals. The candidate runs daily mini-rehearsals in which a single transition signal from each inventory is deployed in three different prompt contexts. The rehearsal builds the signal's availability at production time and the contextual flexibility the timed production demands.

Step 2 — signal-pairing rehearsals. The candidate runs rehearsals in which two transition signals are deployed in the same response — a sequencing signal followed by a contrastive signal, an additive signal followed by an exemplification signal. The rehearsal builds the combinatorial deployment competence the higher-band responses demonstrate.

Step 3 — full-response cohesion rehearsals. The candidate runs full-response rehearsals in which the response deploys at least one signal from each of the four cohesion functions — sequencing, relational, reference, framing — and the response is recorded and self-assessed against a cohesion rubric the candidate maintains across rehearsals.

Step 4 — under-pressure rehearsals. The candidate runs rehearsals under TOEIC Link timing conditions with no extra preparation time, and the rehearsal targets the deployment automaticity that the timed condition demands. The under-pressure rehearsal is the conditioning that converts the deliberate-rehearsal competence into the production-automaticity the test condition assesses.

Candidates who run this four-step rehearsal sequence systematically — daily mini-rehearsals on weekdays and full-response rehearsals on weekends across an eight-to-twelve-week window — typically observe a one-band improvement on the discourse-cohesion criterion in the band where they had been ceiling-suppressed. The improvement is realized through the deployment-automaticity development rather than through content-knowledge acquisition, and the automaticity transfers to the test condition because the test condition mirrors the rehearsal condition.

The related discipline of TOEIC Link Speaking prosodic control and stress placement under time pressure addresses the prosodic layer that signals discourse structure through intonation and stress patterns, and the related discipline of TOEIC Link Speaking pragmatic politeness and face management addresses the politeness layer that adjusts the response register to match the prompt's interaction context. The three disciplines combine to build the full speaking-production competence the rubric assesses.