Notes on Semantic Texture: Textual Coherence, Frames, and Rhetorical Structures (Esser 2009)
Page 1
Topic: Semantic Texture and the move from formal texture to text-internal, psycholinguistic and cognitive aspects of text constitution.
Key idea: Texts are not just aggregates of sentence-level structures; they are outcomes of problem-solving cognitive procedures in language users. Research shows text lies beyond the sentence, co-constructed with human communication and cognition, and realized through cognitive procedures such as understanding and producing texts.
Two broad models for prior encounters with language:
Corpus-based model: uses large corpora (e.g., the British National Corpus, Burnard) to approximate average language use.
Priming model: explains why some wordings are preferred, based on individual encounters and cognitive loading.
Collaboration between approaches: Lexical collocation (corpus) and priming (psychology) complement each other to explain repeated co-occurrence of words.
Hoey’s synthesis (Hoey 2005): To account for collocation, assume every word is mentally primed for collocational use. Learning a word through speech/writing loads it with co-texts and contexts in which it appears.
Illustrative examples (Hoey 2005: 5–7) contrasting normal vs unusual collocations:
Example context: a travel sentence about Hammerfest; the original (1) versus a reformulation (2).
Normal collocations (from a corpus) vs unusual collocations: the rewritten version (2) is grammatically correct but clumsy due to weaker, less interconnected collocations.
Figure 4.1 contrasts normal vs unusual collocations (e.g., "thirty-hour ride" vs reformulated phrase like "ride there by bus").
Figure 4.2 shows how in the original sentence collocations are interconnected (joints printed in bold), supporting text texture beyond grammar.
Conclusions from Hoey: Even rewritten versions reuse existing collocations, but with fewer and weaker connections; normal collocations provide texture and volume beyond grammatical structure. Frequencies in corpora provide an indirect measure of lexical priming at the level of language users in general (corpus-based) and for individual users (priming).
Semantic level priming: Priming is not limited to word forms; there are abstract semantic associations. Hoey introduces the term "semantic association" to describe priming at the level of meaning (e.g., hour with NUMBER and JOURNEY in typical English usage).
Semantic prosody: An umbrella term for positive/negative connotations associated with particular word-forms (e.g., happen frequently co-occurring with unpleasant terms like accident, danger, harm).
Local priming vs textual priming: Local priming concerns immediate word-form co-occurrences; textual priming describes priming of word forms for typical positions (theme at the beginning of a sentence; end-position theme like reason) within larger textual units (paragraphs, sections, texts).
Practical implication: Priming can be observed via corpus data (local frequencies) and via text production/interpretation expectations. Text production often relies on recurrent patterns (themes, positions) that reflect underlying priming.
Example-driven insight (Hoey 2005: 50, 98): Certain word-forms (e.g., consequence) prime for Theme and also for avoidance of paragraph/text-initial position; plural consequents (consequences) shows different priming behavior (paragraph-initial) due to distributional patterns.
Page 2
4.1 Lexical priming (continued)
Priming is not limited to specific word-forms; it operates at more abstract semantic levels. Semantic associations connect word-forms through patterns that recur across corpora and individual language use.
Textual priming: Some word-forms tend to appear at the beginning of sentences or paragraphs; these positions (theme beginnings, end positions) are primed; this affects how readers anticipate and process text.
Hoey (2005: 130) summarizes: a word-type can be primed to occur or avoid occurring in paragraph-initial, section-initial, or text-initial positions; this helps explain coherence at larger textual spans and motivates corpus-based text-linguistics research.
Theoretical implication: The relation between text and world knowledge is mediated by scenes, frames, and scripts (see 4.2), and both lexical priming and textual priming contribute to the texture of text beyond surface forms.
Note: The following examples illustrate ongoing themes of collocation strength, interconnectedness of collocations, and the way priming can extend to semantically related patterns.
The discussion frames priming as a bridge between corpus-based frequency information and individual cognitive tendencies, thereby supporting models of naturalness and usability of language.
Page 3
4.2 Scenes and frame semantics
Central idea: Not only word-forms but also communication scenarios and world knowledge shape understanding and production of text. Two complementary conceptual spheres:
Mental representations of situations and world knowledge (scene knowledge).
Linguistic encodings of those representations in terms of word-forms and constructions (frames).
Fillmore (1977) introduces scene and frame concepts:
Scene: prototypical situations, experiences, and coherent segments of human belief/actions/experiences; can include visual scenes, interpersonal transactions, institutional structures, etc.
Frame: system of linguistic choices associated with prototypical scenes; these frames are the linguistic encoding that helps realize the scene in text or discourse.
Prototypical scenes are typical and recurring; Fillmore uses commercial transaction as a classic example of a prototypical scene. Its description involves several perspectives (at least four sentences such as John bought a sandwich, Henry sold John the sandwich, John paid Henry, The sandwich cost John).
Key takeaway: A single sentence cannot represent all aspects of a scene; text understanding relies on the decoder’s previous knowledge and expectations, which are shaped by frames and scripts.
Text interpretation as world-building: A reader mentally constructs a small world with properties depending on individual experiences; as the text unfolds, details fill in, and expectations adjust, producing experiences like surprise or suspense.
Script theory (Schank/Abelson, 1977): A script describes appropriate sequences of events in a context, with slots and requirements; scripts handle routine situations and support reference to objects already introduced by the script.
Example: A restaurant script: John went into the restaurant; John ordered a Big Mac; He paid for it; He ate in a park. The coherence arises because these actions fit the restaurant script; substituting a dish like coq au vin would require a different script or would feel odd if the script (restaurant) is not applicable to that dish.
Other scripts discussed include riding a bus, watching/playing a football game, attending a birthday party.
Background knowledge: Even when not explicitly stated, scripts and frames explain coherence by filling open slots with world knowledge. If a reader lacks the appropriate script, coherence becomes difficult.
Profile-base model (Croft/Cruse, 2004): Distinguishes between a profile (the concept symbolized by a word) and a base (the knowledge or conceptual structure presupposed by the profiled concept). A frame is a profile against a base; bases can support multiple profiles, forming a domain. No concept exists in isolation; semantic analysis should focus on the profile-base relations and the relations between bases and domains.
Terminological note: Different authors use terms like frame, base, and script inconsistently; Fillmore sometimes uses frame and base similarly; scripts may be sequences within a frame/base/domain.
Page 4
4.3 Coherence
Coherence is the part of the textual world knowledge that makes a text mutually accessible and relevant when surface structure (cohesion) is insufficient.
Terminology distinctions (Beaugrande/Dressler):
Cohesion: surface phenomena (linguistic forms and connectedness).
Coherence: cognitive models that enable understanding when overt cohesive relations are lacking.
The distinction was not always clear in earlier text-linguistics work (Halliday/Hasan; van Dijk). Be aware that coherence can be realized through non-overt relations and background knowledge.
Different notions of coherence across domains: phonetic coherence (intonation and prosody), prosodic coherence (paragraph- or sentence-initial patterns), syntactic coherence (parallelism), stylistic coherence (consistent style), semantic and pragmatic coherence (broader knowledge for understanding).
The central emphasis here is semantic and pragmatic coherence: non-overt relations between sentences provide the glue for understanding when surface cues are missing.
Examples illustrating cohesion vs coherence:
(17) Wash and core six apples. Use them to cut out the material for your new suit. They tend to add a lot to the color and texture of clothing. Actually, maybe you should use five of them instead of six, since they are quite large. → This sentence sequence is cohesive (pronouns, repeated references) but not coherent because there is no underlying frame (scene) that makes a plausible sequence.
(18) John can open Bill's safe. He knows the combination. → Cohesion is present (pronoun 'he'), but coherence requires background knowledge: the safe script and the idea of a combination; without knowledge of the safe script, the sequence is not fully coherent.
The concept of coherence relations: non-overt relations that connect sentences, inferred by the reader using background knowledge and world knowledge.
Halliday/Hasan’s two planes of conjunctive relation (external/world organization vs internal/text organization) are used to discuss how conjunctions relate to coherence, but Halliday/Hasan’s framework is not sufficient alone; many relations cross the two planes.
Van Dijk/Kintsch and Hobbs (1983–1985) contribute a repertoire of coherence relations: (i) generalizations of surface cohesion into semantic relations, (ii) background relations, (iii) exemplification, (iv) contrast, (v) elaboration, (vi) occasion, (vii) background relations, (viii) expansion relations (parallel, elaboration, exemplification, generalization, contrast, violated expectation), etc.
van Dijk/Kintsch examples (21–27) illustrate a range of coherence relations such as general-particular, whole-part, set-subset, including-included, large-small, outside-inside, possessor-possessed, etc.
Hobbs (1985) expands with a rich set of expansion relations: parallel, elaboration, exemplification, generalization, contrast, violated expectation, background, and a cluster of relations labeled as expansion (parallel and elaboration being part of it). He also introduces the notion of “occasion” (previous event sets up the following event) and “background” (information that provides the geographic or contextual ground against which later events occur).
Example (35) John can open Bill's safe. He knows the combination. is analyzed as elaboration: from the first sentence and world knowledge about can and combinations, one infers that John knows the action to open the safe; from the second sentence and knowledge about the combination, one infers a shared proposition P across clauses; this enables co-reference and coherence.
Page 5
4.4 Rhetorical structures (RST)
Rhetorical Structure Theory (RST) aims to describe semantic relations between text components. It is pre-realizational: it describes meanings and intentions rather than linguistic realization.
A rhetorical structure can serve as both a plan for producing a text and a description for understanding it. It is inductively derived from many short texts; it describes how meanings and intentions are structured and combined.
RST uses the notion of rhetorical units (often clauses or clause complexes). Within a clause complex, the matrix unit is the clause (with embedded constituents such as adverbial clauses or non-restrictive relative clauses treated as separate units).
Key relations: nucleus-satellite and list. Nucleus-satellite relations are typically asymmetric; the nucleus is more central/essential, the satellite supports it. List relations coordinate equally ranked elements.
Examples with (41) illustrate an example where unit 2 is in elaboration to unit 1 (nucleus 1 with satellite 2), while 3-4 provide enablement (satellites) for units 1-2; units 3-4 are in a list; unit 1 is the nucleus for the entire text span.
The theory claims that across a multi-unit text, some portions realize central goals and others are ancillary; the nucleus captures the central function, while satellites and lists expand or support it. Frequency of relations is high for a kernel set; new relations are rare.
Example (41) shows the Kentucky Foreign Language Conference Award announcement with items in (1)-(4) mapped to nucleus, satellite relations (enablement, elaboration) and list relations.
The discussion notes that about twenty relations exist; many texts can be analyzed with a small core set; new relations are possible but rare.
Example (42) (thumbs and hand surgeon) demonstrates a sequence where unit 6 expresses the main idea, while other units supply background information; figure 4.5 shows a diagram of nucleus-satellite relations.
The analysis notes that rhetorical units do not have to align with orthographic sentences; different readers may yield multiple analyses for a text, depending on interpretation of rhetorical relations and background knowledge.
RST can also model discourse variation across registers, as demonstrated by the example (43) showing a different ordering of background and main idea.
Overall, RST provides a formalized way to register discourse organization and to identify semantic relations that underlie cohesion and coherence in discourse.
Page 6
4.5 Macrostructures and superstructures
In addition to local coherence relations, researchers have studied higher-order, macro-level organization of texts.
Van Dijk’s macrorules (deletion, generalization, construction) describe how sequences of propositions (clauses) can be rewritten into more general propositions. Macrorules are part of cognitive competence and recursive application yields macrostructures that capture the gist of discourse.
Construction rule (in the macrostructure framework) operates similarly to frames/scripts: a sequence of propositions (X goes to the airport; X checks in; X waits for boarding) entails a macroproposition like 'X is taking a plane', given appropriate background knowledge.
Conventional text types exhibit schematic macrostructures that organize macropropositions; these schematic structures are termed “superstructures” by van Dijk (1980) and van Dijk/Kintsch (1983).
Van Dijk’s narrative superstructure includes categories: Narrative → Story → Plot → Episode(s) → Setting → Event(s) → Complication → Resolution → Evaluation; the Story may have a Moral. The graphic tree of a narrative can be used as a formal representation of the global organization of a narrative.
Superstructures are genre-driven; classical examples include: narration, argumentation, and examination reports. Each text-type has its own conventional superstructure but they are not all necessarily fully definable by superstructure alone.
Propp (1968/1990) is referenced as a forerunner: Russian folktales can be described by a finite set of semantic functions (e.g., absence, departure, interdiction, provision, etc.). Propp’s categories are semantic and not directly tied to particular sentences or clauses.
A key takeaway: Superstructures provide a framework for understanding how macro-level organization constrains and shapes discourse, especially in narrative, argumentative, and other conventional text-types.
The discussion emphasizes that not all categories must be present in every text; the superstructure is one factor among others in describing text typology.
An example (48) illustrates a narrative example with setting, complication, and other elements; Propp’s influence is noted in the idea that plots can be broken down into universal functional elements rather than unique single instances.
Page 7
4.6 Thematic progression and hyperthemes
Beyond macrostructure and superstructure, there are basic relational structures that recur across text types:
General-to-particular progression and its reverse; these are central to how information flows through a text.
Quirk et al. (1985) observe that texts often move from general to particular (example 49) or from particular to general (example 50). In RST terms, the general-to-particular progression can be interpreted as elaboration (nucleus-satellite) while the particular-to-general progression can be interpreted as evaluation.
The notion of theme-rheme and functional sentence perspective (Mathesius) provides a cognitive/functional basis for organizing information. Theme is the element about which something is stated; rheme is the new information; the order T-R vs R-T reflects the perspective toward the addressee (objective vs subjective order).
The theme-rheme relationship is not strictly tied to surface forms; clues include word order, article usage, pronouns, and ellipsis. Many cohesive devices support sentence-level theme-rheme organization.
Catenative function of theme-rheme is described by Mathesius: the theme of the next sentence is often the rheme of the previous sentence; this is a form of chaining thematic progression across sentences.
Daneš (1970, 1974) discusses two types of linear thematic progression at overt word-forms: (i) simple linear progression, (ii) progression with a continuous theme; a third type, progression with derived theme, involves hyperthemes that span multiple sentences or paragraphs (e.g., a paragraph theme).
Distinctions between sentence theme and paragraph/text theme are important: Jones (1977) argues against extending sentence-level theme models to text themes; a paragraph/theme approach is more appropriate for longer discourse.
Examples to illustrate thematic progression include:
(51) A sentence pair showing a simple linear progression (antibiotics discovery) with theme advancing through sentences.
(52) A progression with a continuous theme showing an ongoing thematic thread across sentences.
(53) Progression with a derived theme: a hypertheme (geographical data of New Jersey) that is introduced by a sentence theme and carried by subsequent sentences.
The concept of a text theme or hypertheme emphasizes that a paragraph-level theme can emerge as a higher-level abstraction beyond a single sentence, integrating information across multiple sentences.
Terminology recap: The term theme is used in multiple senses—sentence theme (a constituent of a sentence) and paragraph/text theme (a higher-level thematic notion). The same term is used in different frameworks (Mathesius, Daneš, Halliday, Hoey) with nuanced interpretations.
Page 8
4.7 Questions and exercises
1) Text comparison exercise (Hoey 2006): A/B texts illustrating differences in narrative vs advertising writing. Tasks:
(i) Decide which version A or B is more felicitous and explain why.
(ii) Explain differences in terms of lexical choices and cohesive devices.
2) Propose a superstructure for a complaint letter (e.g., a defective MP3 player purchased online).
3) Use an instruction manual to discuss the relationship between hyperthemes and orthographic paragraphs.
4) Other exercises encourage applying the theoretical frameworks to practical texts.
Page 8 (continued)
4.8 Further reading
Hoey (2005): Extensive applications of priming concepts in corpus-based text-linguistics.
Hoey (2006): Priming in the context of clumsiness.
Bellert (1970): Classic coherence study.
Schank & Abelson (1977): Scripts, plans, and goals in artificial intelligence.
Esser (1984): Terminology on theme, rheme, and related concepts.
Mann & Thompson (1988) and related works: A revised version of RST with further examples.
Page 9
Theoretical and practical synthesis
The material across sections 4.1–4.8 builds a cohesive framework linking lexical-psycholinguistic priming, frames and scripts, coherence relations, rhetorical structures, macrostructures, and thematic progression.
The central aim is to explain how texts are produced and understood by leveraging:
Prior encounters with language (corpus data, priming effects).
Background knowledge about scenes, frames, and scripts.
Coherence relations that tie sentences together beyond overt cohesion.
Higher-order macrostructures and superstructures that organize discourse types (narratives, advertisements, invitations, etc.).
Thematic progression that shows how information is structured and developed across sentences, paragraphs, and sections.
The practical payoff is a toolbox for analyzing text organization and for designing texts with effective texture, coherence, and communicative impact.
Page 10
4.9 (If present) Summary and cross-references
The material integrates multiple strands of linguistics to explain how language users construct meaningful texts.
Core concepts to remember:
Lexical priming and semantic priming
Scenes and frames (Fillmore) and scripts (Schank/Abelson)
Coherence vs cohesion (Beaugrande/Dressler) and multiple coherence relations (Hobbs; van Dijk; Hobbs; Gricean-like inferences)
Rhetorical Structure Theory (RST): nucleus-satellite, list; interpretive and production applications
Macrostructures and superstructures (van Dijk): macrorules, summarization, and the organization of text-types
Thematic progression (Mathesius; Daneš; Jones): sentence-level theme-rheme and text-level hyperthemes
The cross-cutting message: Effective text relies on a blend of linguistic form, cognitive processing, world knowledge, and conventional discourse structures.
Page 11
Notes on terminology and scope
Key terms:
Cohesion vs Coherence: Surface vs cognitive structure.
Scene vs Frame: Mental representation of situations vs linguistic encoding of those situations.
Script: Predetermined sequences of events in a context.
Profile-Base relation: A mechanism to relate a concept’s surface symbol (profile) to its underlying knowledge structure (base).
Macrostructure vs Superstructure: Global organization of propositions vs conventionalized text-type schemas.
Nucleus-Satellite vs List: Central meaning vs adjuncts; lists coordinate equally important units.
Hypertheme: A paragraph- or text-level theme guiding progression.
The literature emphasizes terminological variation; the core ideas remain: knowledge-driven interpretation and text organization beyond single sentences.
Page 12
Selected illustrative examples (brief recap)
4.1 Lexical priming: Examples (1) and (2) show how a reformulation can be grammatically correct but less natural due to weaker interconnections among collocations.
4.2 Scenes and frame semantics: Sandwich transaction and restaurant script examples illustrate scene-frame alignment and how scripts help explain coherence when specific word-forms trigger expected frames.
4.3 Coherence: (17) and (18) show how cohesion can be insufficient for coherence without an available frame/script in the reader’s knowledge base.
4.4 RST: Example (41) demonstrates a clear nucleus-satellite structure with enablement and list relations.
4.5 Macrostructures: Van Dijk’s narrative superstructure shows how events, settings, complications, resolutions, and moral shape a narrative.
4.6 Thematic progression: Examples (51)–(53) illustrate simple linear progression, progression with a continuous theme, and progression with a derived hypertheme.
Page 13
Practical implications for analysis and pedagogy
When analyzing a text, consider:
What scenes or frames are activated by reader/writer knowledge?
Which scripts are likely invoked given the genre and context?
How do coherence relations link sentences beyond surface cohesion?
What macrostructure or superstructure governs the text-type (narrative vs advertisement)?
How does the progression of theme influence information structure and reader expectations?
Exercises in the chapter (4.7) encourage applying these ideas to real texts: comparing versions, constructing superstructures for complaints, examining hyperthemes in manuals, etc.
The breadth of references provides a toolkit for research and teaching in text linguistics, cognitive linguistics, and discourse analysis.
Page 14
Additional resources and connections
The integration of corpus data (Hoey) with cognitive theories (priming, scripts) offers a robust methodological approach for studying text texture.
RST provides a flexible, pre-realizational framework useful for both production planning and comprehension analysis.
The macrostructure/superstructure framework emphasizes how texts of similar genres share deep, conventional organization patterns, even when surface content differs.
The discussions on theme/rheme and progression provide foundational concepts for functional sentence perspective in applied linguistics and language education.
Page 15
Illustrative figures and trees (conceptual references)
Figure 4.3: Profile-base relations (profile vs base) illustrating how different concepts can share the same base while serving as bases for multiple profiles.
Figure 4.4: Rhetorical Structure Theory example (42) showing nucleus-satellite and list relations.
Figure 4.5: Rhetorical structure of example (42) illustrating background, condition, solutionhood, etc.
Figure 4.6: Rhetorical structure of example (45) showing elaboration with a sequence of sentences.
Figure 4.7: RST analysis of Nash example (46) showing a stacked elaboration with background and restatement.
Figure 4.8: Van Dijk’s narrative superstructure diagram showing setting, complication, resolution, and evaluation.
Figure 4.9 and 4.10: Simple linear progression and progression with a continuous theme; Figure 4.11: Progression with derived theme (hypertheme).
Page 16
Terminology recap and conceptual links
Core concepts interact across pages:
Lexical priming and semantic priming influence texture through word-level expectations and semantic associations.
Scenes, frames, and scripts provide world-knowledge scaffolding for interpretation and production.
Coherence relations (Hobbs, van Dijk, etc.) and RST describe how discourse is organized semantically beyond surface form.
Macrostructures and superstructures capture large-scale organization patterns for text-types.
Thematic progression ( Mathesius, Daneš, Jones ) explains how information is distributed and developed across text units.
The overarching aim is to model how readers construct a coherent interpretation by integrating language form, cognitive processes, and discourse structure.
Page 17
Closing notes on methodology and scope
The field emphasizes interdisciplinary methods: corpus analysis, cognitive modeling, discourse analysis, and rhetorical theory.
Researchers acknowledge terminological variation, but consensus centers on the interplay of form, cognition, and world knowledge in constructing text meaning.
The material invites readers to apply these theories to practical texts, acquiring skills in analysis, critique, and text design.
Page 18
Publication information
Text: Textbooks in English Language and Linguistics (TELL)
Editor: Magnus Huber and Joybrato Mukherjee
Series: S-TEX:51, Peter Lang, 2009
Acknowledgments and dedication pages follow, including copyright notice and printing details.
Page 19
Bibliographic and rights information
Deutsche Nationalbibliothek listing and bibliographic data available online.
Publication details, ISSN, ISBN, and rights statements provided.
Dedication to family members (Anne, Laura, and Matthias).
Note: The above notes paraphrase and organize the material from Esser (2009) on Semantic Texture (Chapter 4), capturing the main ideas, definitions, examples, and implications across the pages provided. LaTeX-formatted numbers are used where relevant for mathematical precision, in keeping with the instruction to present numerical references in LaTeX. If you’d like, I can convert any specific section into a printable PDF or tailor the notes to a particular exam outline or slide deck.