📘 Lecture 7 — Immediate Processing of Sentence
📖 Overview: This lecture explores how the human mind processes sentences in real-time during reading and listening. It examines parsing strategies like late closure and minimal attachment, contrasting modular versus interactive processing models, and concludes with how working memory and expectations can lead to comprehension errors or incomplete representations.
🗂️ Topics Covered
The lecture covers parsing as immediate decision-making in building syntactic structures, detailing the immediacy principle and its trade-offs. It examines specific parsing strategies: the late closure strategy and minimal attachment strategy. The modular versus interactive models of sentence processing are compared, followed by the constraint-based model as an alternative to the garden path model. Finally, it discusses the role of working memory in comprehension and the phenomenon of incomplete or inaccurate representations, including the Moses illusion.
📝 Lecture Summary
Topic-037: Parsing
Parsing is described as a form of problem solving or decision making where we make decisions about where to place incoming words into the phrase marker we are building. These decisions are made immediately as we encounter a word, a principle called the immediacy principle. This means when we first see or hear a word, we access its meaning from permanent memory, identify its likely referent, and fit it into the syntactic structure of the sentence. The alternative to this is a "wait-and-see" approach, postponing interpretation until it is clearer where the sentence is going.
Considerable evidence supports the immediacy principle, as the number of decisions in understanding even a single sentence can overload our cognitive resources. For example, the sentence "John bought the flower for Susan" is syntactically ambiguous. While immediacy reduces memory load, it can lead to errors in parsing. For instance, the fragment "The florist sent the flowers..." initially looks like a simple declarative sentence, but if it continues as "...was very pleased", it becomes a grammatical sentence with an embedded reduced relative clause (a clause that modifies a noun, not signaled with a relative pronoun like "who").
🔑 Definition — Parsing: A form of problem solving or decision making where we make decisions about where to place incoming words into the phrase marker we are building. 🔑 Definition — Immediacy Principle: When we first see or hear a word, we access its meaning from permanent memory, identify its likely referent, and fit it into the syntactic structure of the sentence. 🔑 Definition — Phrase Marker: The syntactic structure we are building for a sentence. 🔑 Definition — Reduced Relative Clause: A relative clause that is not signaled with a relative pronoun (e.g., "who"). 📌 Example: The sentence fragment "The florist sent the flowers..." is initially parsed as a simple declarative sentence (subject: florist, verb phrase: sent the flowers). However, when continued as "...was very pleased", the parsing must be revised to interpret it as a reduced relative clause, making it equivalent to "The florist who was sent the flowers was very pleased."
Topic-038: Parsing Strategies
The Late Closure Strategy states that, wherever possible, we prefer to attach new items to the current constituent. A primary motivation is that this reduces the burden on working memory during parsing. An example is the sentence "Tom said that Bill had taken the cleaning out yesterday", where the adverb yesterday may be attached to the main clause or the subordinate clause; we tend to prefer the latter. Another example is "Jessie put the book Kathy was reading in the library", where the prepositional phrase in the library could modify either the verb put or reading; we prefer attaching it to the latter.
Further evidence comes from garden path sentences, such as "Since Jay always jogs a mile seems like a very short distance to him". Frazier and Rayner found that fixation times on the last few words were longer, implying readers misinterpreted a mile and had to make adjustments. In a garden path sentence, we interpret a sentence in a particular way only to find out near the end that we misinterpreted it, lending further support to the immediacy principle. The Minimal Attachment Strategy states that we prefer attaching new items into the phrase marker being constructed using the fewest syntactic nodes consistent with the rules of the language.
🔑 Definition — Late Closure Strategy: Wherever possible, we prefer to attach new items to the current constituent. 🔑 Definition — Minimal Attachment Strategy: We prefer attaching new items into the phrase marker being constructed using the fewest syntactic nodes consistent with the rules of the language. 🔑 Definition — Garden Path Sentence: A sentence where we interpret it in a particular way only to find out near the end that we misinterpreted it, giving the subjective impression of being led down a garden path. 📌 Example: In the garden path sentence "Since Jay always jogs a mile seems like a very short distance to him", the ambiguity is artificial because it lacks a comma after jogs. Readers initially interpret a mile as the object of jogs, but must reanalyze it as the subject of seems.
Topic-039: Modular Versus Interactive Models
The modular approach suggests that words activate syntactic processing strategies used to organize words into a phrase marker, with strategies like preferring to attach incoming words to the most recent constituent. Though generally useful, these strategies can lead to errors and subsequent re-analyses. The interactive approach emphasizes that we use all available information, including lexical, discourse, and contextual factors. While the modular approach insists syntactically-based strategies are used first, with lexical and discourse factors coming later, the interactive model asserts we simultaneously use all available information to parse sentences. Current research supports the role of lexical and contextual factors, but the role of discourse factors is less evident.
🔑 Definition — Modular Approach: Syntactic processing strategies are used first to organize words into a phrase marker, with lexical and discourse factors coming in later. 🔑 Definition — Interactive Approach: We simultaneously use all available information (lexical, discourse, and contextual) to parse sentences.
Topic-040: Constraint Based Model
Theories into sentence processing can be viewed from two main accounts: modular or interactive. The Garden Path model proposed by Frazier and Rayner supports the modular account, arguing sentence processing involves analysis of each individual module with little or no feedback, thus inhibiting correction. It argues readers only consider initially one syntactic structure for any given sentence, and meaning is not involved in selection of preliminary syntactical meaning, allowing readers and listeners to be misled by ambiguous sentences. The Constraint based theory proposed by MacDonald supports the interactive account, arguing sentence processing involves immediate incorporation of all available information. All relevant information is available immediately to the parser, and the incoming information is analyzed with all possible constraints or outputs activated and ranked according to activation strength. The syntactic structure receiving the most support from constraints is highly activated and thus chosen. When two constraints are equally activated, ambiguity arises.
💡 Why this matters: These two theories propose conflicting basic ideologies. The Garden Path model argues only one syntactic structure is initially considered and meaning is not involved in selection, whereas the constraint based theory argues all relevant information is used and several syntactical meanings are initially considered before the most appropriate is selected.
🔑 Definition — Garden Path Model (Frazier & Rayner): Readers only consider initially one syntactic structure for any given sentence; meaning is not involved in the selection of preliminary syntactical meaning. 🔑 Definition — Constraint Based Theory (MacDonald): All relevant information is available immediately to the parser; incoming information is analyzed and all possible constraints or outputs are activated and ranked according to the strength of subsequent activation. The syntactic structure receiving the most support from constraints will be highly activated and chosen.
Topic-041: Working Memory and Comprehension
Given the complexity of comprehension, working memory capacity is related to individual differences in comprehension performance. Gernsbacher and Faust found that less skilled comprehenders were less efficient in rejecting inappropriate meanings of ambiguous words. For example, when presented with sentences such as "He dug with the spade", less skilled comprehenders were slower to reject the meaning of spade that pertains to playing cards in favor of the meaning that pertains to gardening.
🔑 Definition — Working Memory: The cognitive system that temporarily holds and manipulates information during complex tasks like comprehension. 📌 Example: When presented with "He dug with the spade", less skilled comprehenders were slower to reject the meaning of spade related to playing cards in favor of the gardening meaning, demonstrating less efficient suppression of inappropriate word meanings.
Topic-042: Incomplete or Inaccurate Representations
The observation that comprehenders may develop incomplete or inaccurate representations of sentences is not new. In one classic example, participants were asked, "How many animals of each sort did Moses put on the ark?" Most people respond "two," instead of noticing it was Noah, not Moses, who gathered the animals. The significance of these errors is twofold. First, in naturalistic situations, people frequently misinterpret what others say for reasons like distraction or noise; these errors perhaps tell us more about comprehension in the natural environment than correct performance. Second, studies of incomplete representations emphasize the influence of expectations in sentence comprehension. The "Moses illusion" illustrates we come to sentence comprehension with preexisting ideas or preferences. When sentences don't match our expectations, we sometimes misinterpret them and correct ourselves, but other times the expectations win out, and the meaning we carry is fundamentally flawed.
🔑 Definition — Incomplete/Inaccurate Representations: Comprehenders may develop flawed interpretations of sentences, often due to expectations or cognitive load. 🔑 Definition — Moses Illusion: A phenomenon where people fail to detect an inaccuracy in a question (e.g., "How many animals did Moses put on the ark?") because the overall gist or expectation overrides the specific detail. 📌 Example: When asked "How many animals of each sort did Moses put on the ark?", most people answer "two" instead of noticing the error (it was Noah, not Moses), demonstrating that expectations can lead to fundamentally flawed comprehension.
⭐ Key Takeaways
The lecture establishes that immediate processing (the immediacy principle) is the default for parsing, as it reduces the cognitive load on working memory, but it can also lead to systematic errors like garden path sentences. Two key parsing strategies—late closure (attaching new items to the current constituent) and minimal attachment (using the fewest syntactic nodes)—guide how we build syntactic structures, though they can sometimes mislead us. The modular Garden Path model and the interactive Constraint Based model represent opposing theories on whether only syntactic information is considered first or all available information (including lexical and contextual) is used simultaneously. Individual differences in working memory capacity affect comprehension efficiency, particularly in suppressing inappropriate word meanings. Finally, expectations can lead to incomplete or inaccurate representations, as illustrated by the Moses illusion, where the overall meaning overrides specific factual inaccuracies.
🧠 Quick Revision Questions
- What is the immediacy principle, and what is its main advantage and disadvantage in sentence processing?
- Define the late closure strategy and provide an example sentence that illustrates it.
- What is the fundamental difference between the Garden Path model and the Constraint Based theory regarding how many syntactic structures are initially considered?
- What did Gernsbacher and Faust's study on working memory find about how less skilled comprehenders handle ambiguous words?
- What is the Moses illusion, and what does it reveal about the role of expectations in sentence comprehension?
📘 Lecture 8 — Comprehending Figurative Language
📖 Overview: This lecture explores how we understand language that means something other than its literal interpretation, focusing on figurative language comprehension. It covers major types like metaphors and indirect speech acts, examines key theoretical models (pragmatic, conceptual metaphor, and class inclusion theories), and provides a comparative analysis of various figurative forms. This matters because figurative language is remarkably common—speakers use one unique metaphor for every 25 words in TV news—making it essential for a complete theory of language comprehension.
🗂️ Topics Covered
The lecture begins by defining figurative language and its prevalence, then categorizes types of figurative language with emphasis on indirect speech acts and metaphors. It examines studies of figurative language comprehension, comparing the pragmatic theory, conceptual theory, and class inclusion theory. Finally, it provides a comprehensive comparative analysis of six types of figurative language: metaphor, simile, personification, hyperbole, and symbolism, with numerous examples for each.
📝 Lecture Summary
Topic-043: Comprehending Figurative Language
Figurative language is language that means one thing literally but is taken to mean something different. It is a ubiquitous aspect of language—Honeck has noted its prevalence in advertising, and studies of TV news programs have found that speakers use one unique metaphor for every 25 words. Generally, in figurative language, the intended meanings of words, sentences, and expressions do not coincide with their literal meanings. When speaking figuratively, speakers mean something other than what they literally say. Therefore, to understand figurative language, an individual must be able to grasp the speaker’s intention in a given context. The most common examples include metaphors (e.g., ‘Love is a journey’), which involve ‘understanding and experiencing one kind of thing in terms of another.’
Topic-044: Types of Figurative Language
Two of these types have been examined most intensively in psycholinguistic research: indirect speech acts and metaphor.
🔑 Definition — Indirect speech acts: To understand these, we need to first understand the concept of speech act. A good deal of research has focused on utterances that function as actions rather than communicating information. When we use phrases such as I promise…, I apologize…, and I congratulate…, the very act of uttering the sentence is a kind of action. These are quite different than utterances in which assertions are made. For example, it makes no sense to respond No, that’s not true to “It’s going to be cold today.”
🔑 Definition — Metaphor: When someone says Jim’s head is full of rocks, we instantly recognize it as a metaphoric statement. The comprehension of metaphoric language poses very interesting problems for a general theory of language comprehension. Metaphors are invariably literally false yet often meaningful, amusing, thought-provoking, or poignant. Metaphors consist of three main parts. Consider: Billboards are warts on the landscape.
Topic-045: Studies of Figurative Language Comprehension
The different types of figurative language enable us to communicate a wider range of meanings than would be possible with literal language alone. Metaphors are primarily used to convey ideas and feelings that are difficult to express, while indirect speech acts are often employed to state requests politely. Evidence to date does NOT support the pragmatic theory that we comprehend figurative language by first considering and then rejecting the literal meaning. Proponents of the conceptual and class inclusion theories have responded to the limitations of the pragmatic theory. The conceptual theory appears best equipped to explain instances where we automatically access figurative meaning. The class inclusion model is most helpful in connecting figurative language study with language comprehension in general and lexical comprehension in particular.
Topic-046: Comparison and Contrast: Pragmatic Theory and Conceptual Metaphor Theory
Although figurative language is an important aspect of everyday usage, psycholinguists have only studied it in detail in recent years. Research has been conducted in the context of three main theories: pragmatic, conceptual metaphor, and class inclusion theories.
🔑 Definition — Pragmatic Theory: Linguistic communication takes place within a context of shared assumptions called conventions. Grice identified four conventions (called “maxims”) governing conversation: we strive to be informative, clear, relevant, and truthful. These conventions provide ground rules for successful conversations. Grice’s point is that these conventions provide a basis for interpreting what others mean because we generally assume they will be observed, unless we have information to the contrary.
💡 Why this matters: The pragmatic theory proposed that we first process literal meaning, then reject it and derive figurative meaning. Current evidence suggests this is not how we process figurative language—we access figurative meanings directly in many contexts.
Topic-047: Class Inclusion Theory
Glucksberg and his colleagues advanced a model stating that metaphors are class inclusion statements. When we see a metaphor, we understand it as analogous to the kinds of class inclusion statements we encounter regularly. To determine whether either interpretation is true, we must retrieve the lexical representations of the appropriate nouns and assess whether the class inclusion relation is applied appropriately.
Topic-048: Comparative Figurative Languages
🔑 Definition — Metaphor: A statement that doesn’t literally make sense but conveys meaning through comparison. Examples include: “Time is a thief,” “The world is my oyster,” “You’re a couch potato,” “Time is money,” “He has a heart of stone,” “America is a melting pot,” “You are my sunshine.” Metaphors only make sense when the similarities between the two things being compared are apparent or the reader understands the connection.
🔑 Definition — Simile: A comparison that uses the words “like” or “as.” Examples: “Busy as a bee,” “Clean as a whistle,” “Brave as a lion,” “The tall girl stood out like a sore thumb,” “It was as easy as shooting fish in a barrel,” “My mouth was as dry as a bone,” “They fought like cats and dogs,” “Watching that movie was like watching grass grow.”
🔑 Definition — Personification: Gives human characteristics to inanimate objects, animals, or ideas. Often used in poetry, fiction, and children’s rhymes. Examples: “Opportunity knocked at his door,” “The sun greeted me this morning,” “The sky was full of dancing stars,” “The vines wove their delicate fingers together,” “The radio suddenly stopped singing and stared at me,” “The sun played hide and seek with the clouds.”
🔑 Definition — Hyperbole: An outrageous exaggeration that emphasizes a point. It tends toward the ridiculous or funny and adds color and depth to a character. Examples: “You snore louder than a freight train!”, “It’s a slow burg. I spent a couple of weeks there one day,” “She’s so dumb; she thinks Taco Bell is a Mexican phone company,” “I had to walk 15 miles to school in the snow, uphill, in bare feet,” “You could’ve knocked me over with a feather.”
🔑 Definition — Symbolism: Occurs when a word has its own meaning but is used to represent something entirely different. Examples: Using the American flag to represent patriotism; a red rose to symbolize love; an apple pie to represent traditional American lifestyle; a chalkboard to represent education; the color black to symbolize evil or death; an owl to represent wisdom.
⭐ Key Takeaways
Figurative language is pervasive in everyday communication, with metaphors occurring remarkably frequently—about once every 25 words in TV news. The three main theoretical frameworks for understanding figurative language comprehension are the pragmatic theory (based on Grice’s maxims of conversation), the conceptual theory (which explains automatic figurative meaning access), and the class inclusion theory (which treats metaphors as class inclusion statements). Evidence does NOT support the idea that we process literal meaning first and then reject it; instead, we access figurative meanings directly. The six major types of figurative language—metaphor, simile, personification, hyperbole, symbolism, and indirect speech acts—each serve distinct communicative functions, from conveying difficult ideas to making polite requests.
🧠 Quick Revision Questions
- According to the lecture, how frequently do speakers use metaphors in television news programs?
- What are Grice’s four conversational maxims, and how do they relate to understanding figurative language (according to the pragmatic theory)?
- Why does the lecture state that evidence does NOT support the pragmatic theory’s claim about how we comprehend figurative language?
- According to Glucksberg’s class inclusion theory, how do we process metaphorical statements like “Billboards are warts on the landscape”?
- What distinguishes a simile from a metaphor, and provide one original example of each?
📘 Lecture 9 — Memory for Sentences
📖 Overview: This lecture examines how we process and retain sentences, focusing on the distinction between memory for meaning versus exact wording. It explores the time course of retention, the role of pragmatic factors, and how inferences and false memories shape our recall. Understanding these mechanisms is crucial for comprehending how language is stored in long-term memory.
🗂️ Topics Covered
The lecture begins by investigating whether we retain the exact wording of sentences or just their meaning, supported by early research showing that meaning is primarily stored. It then examines the time course of retention and the influence of pragmatic factors that can cause us to remember exact forms. The lecture also covers how inferences drawn from general world knowledge affect sentence memory, leading to false recognition errors as demonstrated by the DRM paradigm. Finally, it introduces propositions as idea units and discusses their role in text comprehension and retention, along with the impact of sentence and clause types on readability.
📝 Lecture Summary
Topic-049: Memory for Meaning Versus Surface Form
A fundamental question in sentence memory research is whether we retain the exact (verbatim) wording of a sentence or simply its meaning. Early research, such as studies by Fillenbaum, suggested that only meaning was retained. In these experiments, participants were presented with a long list of unrelated sentences and later given a multiple-choice test. The results indicated that people remembered the gist of the sentence rather than its precise wording.
Topic-050: Time Course of Retention and Pragmatic Factor
Time course of retention: A classic study by Sachs examined the time parameters within which we extract meaning from syntactic structure. This research supported the idea that we ordinarily use the syntactic structure of a sentence to extract its underlying meaning. The tendency to store only the meaning of a sentence in permanent memory is not limited to spoken languages.
Pragmatic factors: In some situations, however, we remember the exact form of what was said. This occurs when a statement is puzzling, confusing, or irritating, prompting us to mull it over. Studies have examined how pragmatic factors interact with semantic and syntactic considerations in sentence memory.
💡 Why this matters: This shows that memory is not a simple recording but is influenced by the relevance and salience of the utterance.
Topic-051: Inferences and Sentence Memory
The idea that greater elaboration of processing leads to better retention has received substantial support. Elaboration is a process by which incoming information is related to information already stored in permanent memory, enriching the memory representation of the new material. A specific form of elaborative processing is the drawing of inferences.
Inferences and false recognition errors: The general experimental procedure used by Bransford presented people with long lists of sentences and later probed their tendency to make false recognition errors. These are errors where people believe they saw or heard something that was actually not presented. A long list is necessary to encourage participants to attend to the meaning, not just the form, of the sentences. In one study, Johnson, Bransford, and Solomon examined people’s comprehension and retention.
🔑 Definition — Inferences: Conclusions derived from a sentence based on general world knowledge, which can enrich memory but also lead to false memories.
Topic-052: False Recognition Errors
The Deese–Roediger–McDermott (DRM) paradigm is a procedure used to study false memory in humans. Pioneered by James Deese in 1959 and popularized by Henry L. Roediger III and Kathleen McDermott in 1995, the procedure involves the oral presentation of a list of related words (e.g., bed, rest, awake, tired, dream, wake, snooze, blanket, doze, slumber, snore, nap, peace, yawn, drowsy). Participants are then asked to remember as many words as possible.
🔑 Definition — False memory: Memory for an event that never occurred.
📌 Example: In the DRM paradigm, subjects typically recall a related but non-presented word (e.g., sleep), known as a 'lure', with the same frequency as presented words. After the test, about half of all participants report that they are sure they remember hearing the non-presented word, indicating a false memory.
Topic-053: Propositions and Sentence Memory
We generally store the gist of what another person has said rather than the exact form. An exception is statements that are pragmatically striking, such as those requiring a response or flouting normal discourse conventions. In these cases, we draw some inference based on what was said and store this enriched meaning along with the surface form. Other inferences are based on general world knowledge, not purely linguistic knowledge. After a period of time, these inferences become increasingly indistinguishable from the exact sentences we were exposed to. This suggests that a linguistically based representational system (such as deep structure in transformational grammar) is a poor model for sentence memory.
Topic-054: Analysis of Propositions, Sentences and Clause Type
A proposition is an idea unit and a statement that expresses a factual claim. It is the basic unit involved in the understanding and retention of text. Propositions correspond roughly to verbs, adjectives, adverbs, prepositions, and subordinating conjunctions (not nouns or pronouns).
🔑 Definition — Proposition density: An important factor in reading comprehension because of a proposition’s role in text comprehension and retention. The combination of text comprehension and demands on working memory suggests that proposition density might be useful in the selection of college textbooks.
📐 Readability formulas: Widely adopted formulas include Flesch-Kincaid Grade Level or Reading Ease (based on word and sentence length), Degrees of Reading Power and Lexiles (based on readers’ performance on a cloze procedure), and Coh-Metrix (based on various language-discourse levels). These formulas do not estimate proposition density. Sentence and clause types and nonfinite verbals are important because they are directly related to complex, embedded syntax.
⭐ Key Takeaways
You must remember that we generally retain the meaning or gist of sentences rather than their exact wording, a finding supported by early studies like Fillenbaum’s. However, pragmatic factors such as confusion or irritation can cause us to remember exact forms. The drawing of inferences, based on general world knowledge, enriches memory but also leads to false recognition errors, as demonstrated powerfully by the DRM paradigm. A proposition, defined as an idea unit, is key to text comprehension and retention, and its density is a critical factor that is not captured by standard readability formulas. Finally, sentence and clause types that create complex, embedded syntax place demands on working memory.
🧠 Quick Revision Questions
- What did Fillenbaum’s early research conclude about memory for sentences?
- According to Sachs, what is the role of syntactic structure in memory retention?
- Define a false recognition error and provide an example from the DRM paradigm.
- What is a proposition, and why is proposition density important for reading comprehension?
- How do pragmatic factors influence whether we remember the exact form of a sentence versus its meaning?
📘 Lecture 10 — Comprehension of Discourse
📖 Overview: This lecture explores how humans comprehend connected discourse beyond individual sentence meanings. It examines the structural and functional approaches to discourse, the critical roles of cohesion and coherence, and the cognitive strategies—like direct matching, bridging, and reinstatement—that listeners/readers use to build a coherent mental representation of text.
🗂️ Topics Covered
The lecture begins by distinguishing local and global discourse structure and the importance of arrangement over individual sentence meanings. It then covers theories of discourse structure from structural and functional perspectives. The central concept of cohesion is introduced, including reference as a key category. Strategies for establishing coherence are detailed, focusing on given/new information, direct matching, bridging inferences, and reinstating old information. The lecture concludes by examining the role of working memory and individual differences (reading span, background knowledge) in discourse comprehension.
📝 Lecture Summary
Topic-055: Local and Global Discourse Structure
Comprehension of connected discourse depends less on the meanings of individual sentences than on their arrangement. It is possible for a group of meaningful sentences to be thrown together in a way that makes no sense at all, highlighting the importance of structure at both local (sentence-to-sentence) and global (overall text) levels.
Topic-056: Discourse Structure: Theory, Practice, and Use
The search for rules or conventions has dominated both structural and functional approaches to discourse. In structural approaches, the aim is to discover rules which, if followed, result in an acceptable or well-formed text. In approaches viewing discourse in terms of communicative behavior, the aim is to discover the social conventions which determine which utterances may occur and what they may be combined with. The main concern here is the acceptability of discourse.
Topic-057: Cohesion
A central concept is the notion of cohesion. Halliday and Hasan (1976) define cohesion as referring to "the range of possibilities that exist for linking something with what has gone before."
🔑 Definition — Cohesion: The range of possibilities that exist for linking something with what has gone before in a text.
🔑 Definition — Reference: A semantic relation whereby information needed for the interpretation of one item is found elsewhere in the text. Reference deals with links between words (or phrases) and other words (or phrases) in discourse. We often use pronouns such as she, he, it, his, her, and their to refer to earlier items.
Topic-058: Strategies Used to Establish Coherence
Given information refers to information an author or speaker assumes the reader or listener already knows, whereas new information is information the comprehender is assumed not to know. Most sentences contain both.
📌 Example: Sentence (1) assumes readers know the bank was robbed (given) but not who did it (new). Sentence (2) assumes readers know Steve robbed something (given) but not what. (1) It was Steve who robbed the bank. (2) It was the bank that Steve robbed.
Direct matching is the simplest case where given information in the target sentence directly matches an antecedent in the context sentence. 📌 Example: (3) We got some beer out of the trunk. (4) The beer was warm. In comprehending (4), we divide it into given (the beer, marked by the) and new (was warm), search memory for a previous reference to beer in (3), and attach the new information.
Bridging occurs when there is no direct antecedent, but we can still tie sentences together by making an inference. 📌 Example: (5) Last Christmas, Eugene went to a lot of parties. (6) This Christmas he got very drunk again. A bridging inference (that Eugene got drunk at last year's parties) is needed to make sense of again. In contrast, a direct antecedent pair like (7) "got absolutely smashed" and (8) "got very drunk again" requires no such bridge. Haviland and Clark (1974) showed that target sentences requiring bridges take longer to comprehend.
Reinstating old information occurs when a sentence refers to something already introduced but no longer in the foreground of the discourse. The comprehender must reinstate the information to match it with the target. 📌 Example: Passage (i) (I am trying to find a black dog... Yesterday that dog bit a little girl. She...) is easier to comprehend than passage (ii) (Yesterday a black dog bit a little girl... She...) because the antecedent for she (the little girl) is foregrounded in (i) but backgrounded by the dog in (ii). Studies have shown reinstatements increase comprehension time.
Topic-059: Identifying New Topics of Discourse
Three cases are discussed: (1) Direct match: given information in the target has an antecedent immediately preceding it—the given/new strategy works without problem. (2) Bridging: when we cannot find an antecedent readily, we form a bridge if we believe the author intends a relationship but hasn't spelled it out. (3) Reinstatement: we search permanent memory for antecedents when we think our failure to find an antecedent was caused by the author's carelessness.
Topic-060: Role of Working Memory
Individual experiences and abilities vary. Because discourse comprehension involves working memory, individual differences in working memory capacity influence comprehension.
Daneman and Carpenter developed a complex reading span task to examine this trade-off. Participants read aloud a series of sentences (processing function) and then recalled the final word in each sentence (storage function). The task began with two sentences and progressed until a person could not recall the final words. Reading spans varied from two to five. The researchers then found a significant correlation between reading span and reading comprehension.
Working memory capacity is not the only influence. Another is background knowledge of the subject matter. When we encounter unfamiliar passages, it is more difficult to draw appropriate inferences. When we have relevant information in permanent memory, it is easier to draw inferences.
💡 Why this matters: These findings show that discourse comprehension is not just about language ability but also about cognitive resources (working memory) and prior knowledge, explaining why comprehension varies so much across individuals and contexts.
⭐ Key Takeaways
Discourse comprehension depends on the arrangement of sentences, not just their individual meanings. The concept of cohesion, especially reference, is fundamental to linking sentences together. Readers use the given/new strategy to build coherence, and when direct matching fails, they resort to bridging inferences or reinstating old information—both of which increase processing time. Individual differences in working memory capacity (measured by reading span) significantly predict comprehension ability, and background knowledge also plays a crucial role in drawing inferences. The distinction between foreground and background information is critical for understanding why some texts are harder to process than others.
🧠 Quick Revision Questions
- What is the difference between structural and functional approaches to discourse?
- Define cohesion according to Halliday and Hasan (1976), and explain the category of reference.
- How does the comprehension process differ between direct matching and bridging? Provide an example of each.
- Explain the reinstatement strategy and why it increases comprehension time.
- What was the design and key finding of Daneman and Carpenter's reading span task?
📘 Lecture 11 — Memory for Discourse
📖 Overview: This lecture examines how discourse is represented and stored in human memory. It explores three distinct levels of representation—surface form, propositional content, and situational models—and explains how working memory capacity influences discourse comprehension. Understanding these memory systems is crucial for grasping how readers process and retain complex linguistic information.
🗂️ Topics Covered
The lecture covers working memory's role in discourse processing, surface representations of verbatim text that are quickly purged after comprehension, propositional representations where meaning is stored as networks of propositions influenced by propositional density, inferences that integrate implicit propositions with explicit ones, situational models representing the described state of affairs including spatial and causal models, and simultaneous investigation of all three levels of representation in a single experiment.
📝 Lecture Summary
Topic-061: Memory for Discourse
Preserved working memory is pivotal for language processing effectiveness, being an essential tool for resolving structural and lexical ambiguity during discourse processing and comprehension. Based on this premise, several investigators have suggested that changes in the capacity of working memory will have an impact on linguistic processing, leading to problems in comprehension of larger units like discourse.
💡 Why this matters: Working memory limitations directly explain why some individuals struggle with understanding longer texts or complex sentences.
Topic-062: Surface Representations
One early study that suggested that surface representations of discourse are very short-lived presented individuals with a long oral passage that was interrupted at irregular intervals. Individuals were asked at each interruption to write down in verbatim form as much of the preceding discourse as they could. Two versions of the passage were created. Consider sentences (1) and (2): (1) The confidence of Kofach was not unfounded. To stack the meeting for McDonald, the union had even brought in outsiders. (2) Kofach had been persuaded by the international to stack the meeting for McDonald. The union had even brought in outsiders.
Although the final clauses in (1) and (2) were identical, the material immediately preceding either came from the same sentence (1) or the earlier sentence (2). It was found that the percentage of correct recall of the next-to-last clause was far better when it was part of the current sentence than when it was part of an earlier sentence. These and similar results have been taken as evidence that the surface or verbatim form of a sentence is stored in working memory only until its meaning is understood, then purged to make room for the next sentence. There is an exception to this rule, however. Subsequent results indicate that we sometimes remember the exact wording over a long time period.
🔑 Definition — Surface representation: The verbatim or exact wording of a sentence stored temporarily in working memory until meaning is extracted.
Topic-063: Propositional Representations
Evidence for the psychological reality of propositions comes from Kintsch and Keenan, who showed that the number of propositions influences the time required to read a passage when preparing to recall it. For example, the following two sentences have about the same number of words: (3) Cleopatra’s downfall lay in her foolish trust in the fickle political figures of the Roman world. (4) Romulus, the legendary founder of Rome, took the women of the Sabine by force.
However, sentence (3) is more complex propositionally than (4), which contains four propositions. Kintsch and Keenan found that a proposition added about 1.5 seconds to the reading time. Later studies provide somewhat lower estimates of the time needed to encode a single proposition but support the general conclusion that the number of propositions is related to reading time. Further work explored the notion that discourse is stored as a network of propositions.
📐 Formula: Reading time increase = 1.5 seconds per additional proposition (Kintsch & Keenan estimate) 📌 Example: Sentence (4) "Romulus, the legendary founder of Rome, took the women of the Sabine by force" contains only 4 propositions, while sentence (3) about Cleopatra contains more propositions despite similar word count. Therefore sentence (3) requires approximately (n-4) × 1.5 seconds longer reading time, where n = number of propositions in sentence (3).
Topic-064: Inferences and Propositional Representations
From a communication standpoint, an inference is a proposition in the underlying discourse structure that is intended but not explicitly expressed by the author and thus must be drawn by the reader. Furthermore, evidence indicates that when we draw inferences from a text, we store the implicit propositions right alongside the explicit propositions we have derived from the text itself.
🔑 Definition — Inference: A proposition that is implied but not explicitly stated in the text, which the reader must generate and then store alongside explicit propositions.
Topic-065: Situational Models
Situational models represent the state of affairs that a text refers to. That is, the assumption is that as we comprehend the propositions of a text, we construct a mental or situational model of the world as described by the text. One possibility is a spatial layout, other is situational models.
Kinds of situational models: All of the preceding examples of situational models are spatial models, but there are other kinds of situational models. Zwann and Radvansky identified a number of different types of models other than spatial models. One is a causal model, in which the parts of a text are connected by causal relations. The highest level of representation is the situation model which represents ideas of the text in an abstract format (e.g., the people, the actions, the setting, the events, and the inferences). Information in the situation model interacts with preexisting knowledge (i.e., episodic and semantic memory) of the reader, so that situation models can be updated online as one continues reading/listening to the text.
🔑 Definition — Situational model: A mental representation of the world described by a text, including people, actions, setting, events, and inferences, which interacts with prior knowledge and can be updated during reading.
Topic-066: Simultaneous Investigations of All Three Levels
As we have seen, we form surface, propositional, and situational representations during the course of comprehending discourse. Most of the studies we have discussed to this point have attempted to isolate one of these levels or to distinguish between different levels. It is also helpful, however, to set up a study that attempts to investigate how each of the levels operates in the same experiment. Fletcher and Chrysler have reported such a study.
⭐ Key Takeaways
Memory for discourse operates at three distinct levels: surface form (verbatim wording stored briefly until meaning is extracted), propositional representation (meaning stored as propositional networks where more propositions increase reading time by approximately 1.5 seconds each), and situational models (mental representations of the described world that can be spatial or causal). Inferences create implicit propositions stored alongside explicit ones, and working memory capacity directly impacts comprehension of larger discourse units. The three levels can be studied simultaneously in a single experimental design.
🧠 Quick Revision Questions
- According to the interrupted passage study, how long does the surface/verbatim form of a sentence typically remain in memory?
- What was the estimated reading time increase per additional proposition found by Kintsch and Keenan?
- What is an inference in the context of discourse processing, and where are inferred propositions stored?
- What is the difference between spatial situational models and causal situational models?
- Who conducted the study that investigated all three levels of discourse representation (surface, propositional, and situational) simultaneously?