week 2
# Ultra-Detailed Lecture Notes: Language and Cognition (PSYC0004) – Week 2: The Science of Reading
Professor: Jenni Rodd
Topic: Reading, from visual word recognition to word meaning access, with an emphasis on connectionist models and learning mechanisms.
---
## 1. Introduction: Reading as a Critical Learned Skill
### 1.1 Key Properties of Reading
- Reading is a relatively recent cultural invention. Biological evolution did not equip humans with dedicated reading brain regions; our brains evolved for spoken language.
- Not a universal human experience. While virtually all hearing individuals acquire spoken language, many sighted people never learn to read. Globally, millions lack literacy.
- Prevalence of low literacy: Even in developed countries, approximately 1 in 5 teenagers (~20%) have low literacy skills (World Literacy Foundation, 2015).
- Extremely valuable skill: Illiteracy costs the global economy over 1 trillion US dollars per year in direct costs (World Literacy Foundation, 2015). Indirect costs include reduced access to health information (e.g., during COVID-19, high-quality information was disseminated primarily through text).
- Foundation for all education: Reading ability at the end of primary school is one of the best predictors of GCSE maths attainment. You need reading to access the entire curriculum.
### 1.2 The Reading Wars
- Reading is a taught skill; most children require explicit instruction.
- There is surprisingly little consensus among educators about the best method to teach reading.
- This debate is termed “The Reading Wars.” It pits phonics-based instruction against whole-language/sight-reading approaches.
- The professor stressed that reading science (cognitive psychology and psycholinguistics) provides the critical evidence base for these educational policy decisions.
- Media example: A newspaper article (shown in lecture) debating whether audiobooks can replace reading – directly relevant to understanding spoken vs. written word processing.
---
## 2. Spoken Word Segmentation (Brief Recap from Week 1)
Before diving into reading, the lecture concluded the topic of speech segmentation:
- The Segmentation Problem: Unlike written English, spoken language lacks consistent gaps between words. Pausing between words would be inefficient.
- Demonstration: Listening to an unfamiliar language sounds like continuous babble. Our brains impose structure on this continuous stream, chunking it into words.
- No single solution: The brain uses multiple probabilistic cues to segment speech:
- Possible Word Constraint: All perceived phonemes must belong to at least one possible word (e.g., /fəʊniːm/ is not parsed as "phoney" + isolated /m/ because English lacks single-consonant words).
- Acoustic Cues: Subtle pronunciation differences mark word boundaries. Example: vowel lengthening in ice cream vs. I scream (the /aɪ/ is longer when it's a standalone word).
- Lexical Knowledge: Recognising a long word (e.g., "alligator") signals the start of the next word, unless a morphological suffix follows (e.g., "alligators").
- Language-Specific Cues: In English, stress tends to fall on the first syllable of words (e.g., apple, table, curtain). Listeners use a stressed syllable as a heuristic for a word onset, even though exceptions exist (e.g., trombone).
- Link to computer speech recognition: Modern systems succeed not by solving segmentation explicitly, but by using powerful models that integrate multiple probabilistic cues, much like the human brain.
---
## 3. Visual Word Recognition: Basic Facts
### 3.1 What Does It Mean to Recognise a Word?
- Identifying constituent letters without confusing visually similar ones (e.g., p vs. b in pig/big).
- Processing letter order correctly (e.g., form vs. from).
- Uniquely identifying the word in the mental lexicon.
### 3.2 Why Reading Might Be Easier Than Listening
- Reader control over timing: You can pause, re-read, or slow down when material is difficult. (In spoken language, rewinding is clunky, and in live conversation impossible.)
- Clear segmentation: In English and many alphabetic languages, spaces between words provide explicit boundary cues. (Not all languages do this: e.g., Mandarin, Japanese often lack inter-word spaces.)
- Stimulus clarity: Printed text is typically high-contrast and uniform; much less variable than speech.
### 3.3 Why Reading Might Be Harder Than Listening
- Different sensory system: Visual processing can be more demanding; eye fatigue is common.
- Attention demands: Reading requires active, voluntary eye movements; spoken language passively arrives at the ears.
- Evolutionary primacy: Spoken language is biologically older, and children begin processing it from infancy. Reading is a secondary skill layered onto existing brain circuits.
### 3.4 Speed and Efficiency
- Visual word recognition is fast. Within 500-600 ms of seeing a word, people can:
- Begin to read it aloud (naming task).
- Decide if a letter string is a real word (lexical decision; e.g., shark vs. sharm).
- Make a semantic categorisation (e.g., is it living?).
- Tachistoscopic presentation: Words displayed for less than 100 ms can be identified.
- Fluent reading speed: About 4 words per second, faster than typical listening speed for audiobooks.
### 3.5 Eye Movements in Reading
- Why eye movements matter: The fovea (central retina) provides high-acuity visual detail over only 4-5 letters; the periphery is low resolution. Therefore, precise eye movements are essential.
- The illusion of smooth reading: Subjectively, our eyes feel like they glide across a page. Actually, they make rapid jumps (**saccades**) interspersed with fixations.
- Characteristics of fixations and saccades:
- Fixations last on average 200-250 ms.
- Perceptual span: About 15 characters to the right (in English), but only 3-4 to the left. Reversed for right-to-left languages (e.g., Hebrew).
- Saccade length: ~8 letter spaces.
- Short words (2-3 letters) are fixated only ~25% of the time; they are often skipped or guessed.
- Long words (8+ letters) are nearly always fixated, sometimes more than once.
- Regressions: 10-15% of saccades go backward when text is difficult, surprising, or when initial visual information was insufficient.
- Preview benefit: Parafoveal information about upcoming word shape and initial letters helps programme the next saccade.
- Skilled reader control: Readers slow down for low-frequency words, surprising words, or complex syntax. Eye movement control is a hallmark of skilled reading and varies across languages with different visual characteristics (e.g., Chinese, Arabic).
### 3.6 Factors Affecting Visual Word Recognition (MegaStudies)
- Numerous variables influence word reading speed/accuracy. Key factors demonstrated in MegaStudies (e.g., Balota et al., 2004; large stimulus sets + multiple regression):
- Frequency: High-frequency words (e.g., house) are processed faster than low-frequency words (e.g., shark).
- Length: Short words (e.g., rat) recognised faster than long words (e.g., hippopotamus).
- Age of Acquisition (AoA): Words learned early in childhood (e.g., giant) show a processing advantage even if they are now less frequent than later-acquired words (e.g., data).
- Semantic Ambiguity: Ambiguous words (e.g., bark) can be slower to process.
- Neighbourhood Effects: Words with many orthographic neighbours (e.g., sink – link, silk, sing) behave differently from words with few neighbours (e.g., yacht), depending on task.
- Concreteness: Concrete words (e.g., book) are faster than abstract words (e.g., hope).
- Important: These factors are often correlated (e.g., frequent words tend to be short, concrete, early-acquired), hence the need for massive datasets to tease them apart.
---
## 4. Models of Visual Word Recognition
### 4.1 Serial vs. Parallel Processing Debate (1970s-80s)
- Core question: How does the brain match a visual input onto the correct entry in the mental lexicon?
- Serial Search Model (Forster, 1976): Proposed that recognition proceeds like looking up a word in a dictionary: check each entry one by one until a match is found. This was heavily influenced by the "brain as digital computer" metaphor.
- Parallel Processing Models: Proposed that all word entries are checked simultaneously.
### 4.2 Morton’s Logogen Model (1969)
- An early parallel model.
- Logogen = "word detector." Each word has a dedicated recognition unit.
- When a word is presented, all logogens receive input in parallel. The one that reaches a threshold "fires", signalling recognition.
- This idea laid groundwork for connectionist models.
### 4.3 Connectionist Models (PDP / Artificial Neural Networks)
- General properties:
- A class of computational model (not just box-and-arrow diagrams, but implemented software).
- Alternative names: parallel distributed processing (PDP) models, neural network models, artificial neural networks (ANNs).
- Inspiration: The brain as a highly interconnected network of simple neurons.
- Massively parallel processing: Complex behaviour emerges from interactions of many simple units.
- Learning: Many later models adjust connection strengths based on experience (not discussed in detail for IAC).
- They move away from the "brain as digital computer" metaphor.
### 4.4 The Interactive Activation and Competition (IAC) Model
- Developers: McClelland & Rumelhart (1981), Rumelhart & McClelland (1982). Foundational publication in the "Connectionist Bible."
- Architecture: Three levels of units:
1. Visual features (input): Detect simple line orientations, etc.
2. Letters: Detect specific letters in specific positions.
3. Words (output): Detect whole words.
- Connections:
- Excitatory (positive): Consistent features feed activation forward and backward (e.g., a vertical line activates the letter T).
- Inhibitory (negative): Units that are inconsistent with a feature reduce its activation within and between layers.
- Between-layer and within-layer inhibition: Within the letter layer, competing letters suppress each other (“winner takes all”).
- Slot-Based Coding:
- The model uses separate banks of letter detectors for each position in a word.
- E.g., there is a specific C-in-position-1 unit, C-in-position-2 unit, etc.
- This is a critical simplification (see limitations below).
- Word Superiority Effect:
- One key phenomenon the IAC was designed to explain.
- Finding: People are faster and more accurate at identifying a letter when it appears within a real word (e.g., R in CARD) compared to a non-word (e.g., R in CQRD).
- Explanation: Top-down feedback from the activated word unit (*CARD*) reinforces the activation of its constituent letters, including the target letter. This parallels the Ganong effect in speech (ambiguous phoneme perceived to form a real word).
- What the IAC Does Well:
- Demonstrated that parallel processing for word recognition is computationally feasible.
- Showed how cognitive processes can be highly interactive, with top-down knowledge (word-level) influencing lower-level processing (letter-level). Some researchers still debate the extent of this interactivity.
- Limitations & Criticisms:
- Slot-based coding of letter position is wrong: The model is completely intolerant to letter transpositions.
- Example: The model sees CLAM and CALM as sharing only 50% of input (letters 1 and 4). It gains no benefit from "right letter, wrong position."
- Real readers are quite tolerant to transpositions. The famous "Cambridge University" email/meme demonstrates this anecdotally:
> Aoccdrnig to a rscheearch at Cmabrigde Uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are...
- Experimental evidence: Priming studies show SERVICE is more easily recognised after seeing SEVRICE (transposed letters) than after SEDLICE (letter substitution). The IAC predicts equal performance, which is incorrect.
- No phonological (sound) representations: The model ignores the role of phonology in reading.
- No word meaning or syntax representations: It stops at word form recognition.
- Key Takeaway (Box’s Aphorism): “*All models are wrong, but some are interesting... The only question of interest is ‘Is the model illuminating and useful?’*” The IAC, while wrong in its letter position coding, revolutionised thinking about parallel, interactive processing.
---
## 5. The Role of Sound in Visual Word Recognition
### 5.1 Phonics and Cracking the Alphabetic Code
- English is an alphabetic language: Visual symbols (letters/graphemes) represent individual sounds (phonemes).
- The child’s task: A pre-literate 4-year-old must learn the direct mapping from print (orthography) to meaning, which is largely arbitrary and hard.
- A helpful shortcut: The child already has a well-established mapping from sound (phonology) to meaning from spoken language.
- Phonics instruction teaches the systematic mapping from print to sound. This is easier than the direct print-to-meaning route because the correspondences, while imperfect in English, are rule-based.
- Why phonics supports reading:
- Child sees C-A-T, sounds it out /k/ /æ/ /t/, and this spoken form activates the meaning (furry four-legged animal).
- This provides a self-teaching mechanism: successful decoding via sound strengthens the direct print-to-meaning connections over time.
- Phonics is necessary but not sufficient: It helps decode words but does not build comprehension (narrative, inference, etc.). An analogy: good eye movement control is necessary for skilled reading but doesn't make one a reader.
### 5.2 The Reading Wars Revisited
- Historical approach: Many adults were taught via "sight reading" (whole-word recognition), not explicit phonics. This worked for some but left many (20-30%) struggling.
- Current controversy: Phonics can be perceived by some parents/authors (e.g., Michael Rosen) as dull, mechanical, and capable of killing a love of reading.
- Professor’s rebuttal: A child will never love reading if they cannot decode. Explicit, systematic phonics is essential for all children to build foundational skills.
### 5.3 Dual-Route Model of Skilled Adult Reading
- For alphabetic languages, skilled adults continue to use two routes:
- Indirect route (print -> sound -> meaning): Vital for reading novel words (e.g., wug), words you've only heard and never seen written, or very rare words. Evidence: we can all "sound out" a non-word, proving we possess phonics knowledge.
- Direct route (print -> meaning): Faster, essential for irregular words where sound rules fail (e.g., yacht).
- Both routes likely operate in parallel throughout adulthood, with relative contribution depending on language regularity and reader skill.
---
## 6. Word Meaning Access
### 6.1 The Problem of Ambiguity
- Ubiquitous ambiguity: Most words are ambiguous. Example: trunk (elephant’s nose, luggage, tree part, etc.).
- Homonymy vs. Polysemy:
- Homonyms: Words with unrelated meanings, often due to historical accident (e.g., bark – dog/tree).
- Polysemy: Words with multiple related senses, far more common. Over time, a word’s meaning mutates and extends (e.g., run: athlete, paint dripping, river, computer program). A word may have 40+ subtly different senses.
- Critical for comprehension: To understand sentences, listeners/readers must select the contextually appropriate meaning. Example from a 2016 SATs reading test (age 10/11):
"Dawn was casting spun gold threads across a rosy sky over South Borneo game reserve."
Nearly every word in this sentence is ambiguous and relies on metaphorical or low-frequency senses. Game (sports vs. wild animals) is especially difficult for children.
### 6.2 Cognitive Mechanisms for Meaning Selection
- General consensus: When encountering an ambiguous word, we automatically activate multiple meanings in parallel, then rapidly select the one that fits the context.
- Occasional errors: Misinterpretations do happen, but they are rare enough that we notice them when they occur.
### 6.3 Cues for Word-Meaning Selection
- Selection relies on the integration of multiple statistical cues:
1. Sentence Context: The most powerful cue. E.g., bark is interpreted differently in "bark of the tree" vs. "bark of the dog".
2. Long-term experience (Dominance): The relative frequency of different meanings. A high-frequency meaning (e.g., game = sport) is easier to access than a low-frequency meaning (e.g., game = wild animals). This can create interference when context supports the subordinate meaning.
3. Recent experience: Prior exposure to a particular meaning temporarily boosts its accessibility.
4. Knowledge about the speaker/writer: E.g., accent (American vs. British English; coach has different dominant meanings), or age of speaker (e.g., tweet).
### 6.4 Key Experimental Studies on Meaning Access
#### Study 1: Duffy et al. (1988) – Context and Dominance
- Method: Eye-tracking while reading sentences containing ambiguous words.
- Materials:
- Balanced ambiguous words: Two equally frequent meanings (e.g., bark).
- Biased ambiguous words: One strong dominant meaning, one subordinate (e.g., port – harbour vs. fortified wine).
- Sentence types:
- Early disambiguation: Context precedes the target word. ("When she finally served it to her guests, the PORT was a great success.")
- Late disambiguation: Context follows the target word. ("Last night the PORT was a great success when she finally served it to her guests.")
- Key Results (very complex):
- In early disambiguation, reading times on the ambiguous word were slower for biased words (because readers had to select between a context-supported low-frequency meaning and a highly active dominant meaning).
- In late disambiguation, reading times on the ambiguous word were slower for balanced words (both meanings activated, need to wait for disambiguating info). At the disambiguating region, reading was slower for biased words when the subordinate meaning had been selected, requiring reinterpretation.
- Conclusion: Supports the Reordered Access Model – meaning access is modulated by both dominance and sentential context, which interact in complex ways.
#### Study 2: Rodd et al. (2016) – Long-Term & Recent Experience in Rowers
- Background: Meaning preferences differ across individuals and change over time. People who move to London update their "tube" meaning dominance.
- Method: Studied recreational rowers, who acquire niche meanings for common words:
- Square (oar blade vertical), Feather (oar blade parallel to water), Catch (moment oar enters water).
- Results: Accessibility of rowing-specific meanings depended on:
- Recent experience: More accessible on days they had rowed.
- Long-term experience: More accessible with years of rowing practice.
- Implication: The mental lexicon is highly flexible; long-term dominance effects are built from cumulative short-term encounters.
#### Study 3: Rodd et al. (2013) – Word Meaning Priming (Auditory)
- Method: Three-stage experiment (Prime – Filler – Test).
1. Prime phase (~6 min): Listen to sentences that bias a low-frequency meaning (e.g., The seal came up onto the bank of the river, where bank = riverbank). Perform a semantic relatedness task to ensure attention.
2. Filler task (~8 min): Digit span or colouring to distract.
3. Test phase (~18 min): Word association task. Hear an ambiguous word (e.g., seal) and say the first word that comes to mind.
- Results: Participants produced associations consistent with the primed meaning on ~35% of trials, a significant increase from the baseline (no-prime) control condition.
- Implication: A single recent encounter with a word meaning has a measurable, though short-lived, boost on its accessibility. This form of learning improves communicative efficiency.
### 6.5 Summary of Learning in Word Meaning Access
- Meaning preferences for familiar words are highly flexible even in adults.
- Short-term priming effects decay quickly.
- Repeated encounters produce robust, long-term cumulative shifts in dominance.
- This lexical flexibility is a hallmark of skilled language use and likely applies to other domains (e.g., syntactic ambiguity resolution).
- Lexical Upkeep: An active, continuously updated lexicon is essential; language comprehension skills will decline if the lexicon is not maintained.
---
## 7. Conclusions & Future Directions
- The Science of Reading provides detailed insights into word recognition and comprehension.
- Sophisticated computational models (like IAC and later variants) offer mechanistic explanations.
- This basic research underpins applied work, notably the phonics revolution in UK schools, which has significantly improved reading outcomes.
- There is a continued need to bridge cognitive psychology and education, and to understand individual differences in reading acquisition journeys.
---
## Multiple Choice Questions
Instructions: Select the best answer for each question. Answers are provided directly below each question.
---
1. What is a primary reason that reading is considered a “taught skill” rather than a naturally acquired one?
A. It develops in all humans with intact visual systems.
B. It is a relatively recent cultural invention that most children need explicit instruction to learn.
C. It is learned effortlessly from the environment, similar to spoken language.
D. It solely depends on biological maturation of the visual cortex.
Answer: B
2. Approximately what percentage of teenagers in developed countries are estimated to have low literacy skills?
A. 5%
B. 10%
C. 20%
D. 50%
Answer: C
3. According to the lecture, illiteracy costs the global economy annually more than:
A. 1 million US dollars
B. 1 billion US dollars
C. 1 trillion US dollars
D. 10 trillion US dollars
Answer: C
4. What is one of the best predictors of GCSE maths attainment mentioned in the lecture?
A. Early numeracy skills
B. Reading skills at the end of primary school
C. Parental income
D. Performance on IQ tests
Answer: B
5. The term “Reading Wars” refers to disagreements about:
A. Whether reading is superior to listening for learning.
B. The best age at which to start formal reading instruction.
C. How to optimally teach children to read (e.g., phonics vs. whole-language).
D. The relative importance of fiction vs. non-fiction in curriculum.
Answer: C
6. In spoken language, why do we not insert gaps between words?
A. Because our articulators cannot stop rapidly.
B. Because it would be deeply inefficient and slow down speaking rate.
C. Because all languages lack clear word boundaries.
D. Because it is impossible to detect word boundaries acoustically.
Answer: B
7. The “possible word constraint” in speech segmentation refers to the idea that:
A. Listeners guess words by their possible meanings.
B. All phonemes heard must belong to at least one possible word in the language.
C. Words are only segmented if they are highly predictable.
D. Listeners can only segment words from a limited possible set.
Answer: B
8. The difference between “ice cream” and “I scream” is cued in speech partly by:
A. A complete pause between words.
B. Vowel lengthening on the first word when it is a separate word.
C. A change in speaker’s pitch register.
D. Wholly different phonemes.
Answer: B
9. In English, listeners often use which suprasegmental cue as a heuristic for a word onset?
A. Syllable-final lengthening.
B. A stressed syllable.
C. A change in speaker identity.
D. A silent pause of at least 50ms.
Answer: B
10. Which task requires participants to decide whether a string of letters is a real word or not?
A. Semantic categorization task
B. Naming task
C. Lexical decision task
D. Stroop task
Answer: C
11. In tachistoscopic presentation, words can be identified with exposure durations of less than:
A. 10 ms
B. 100 ms
C. 500 ms
D. 1000 ms
Answer: B
12. In skilled reading, the high-acuity foveal region typically encompasses how many letters?
A. 1-2 letters
B. 4-5 letters
C. 10-12 letters
D. An entire sentence
Answer: B
13. How long does an average fixation last during fluent reading?
A. 50-100 ms
B. 200-250 ms
C. 500-750 ms
D. 1-2 seconds
Answer: B
14. In English reading, the perceptual span extends further to the ________ of fixation.
A. Left
B. Right
C. Top
D. Bottom
Answer: B
15. Short words (2-3 letters) in text are fixated approximately what percentage of the time?
A. 5%
B. 25%
C. 75%
D. 100%
Answer: B
16. Backward eye movements (regressions) occur on roughly what proportion of saccades?
A. 1-2%
B. 10-15%
C. 40-50%
D. 80-90%
Answer: B
17. Parafoveal preview primarily provides
information about:A. Precise letter identities of all upcoming words.
B. Word shape and initial letters of an upcoming word.
C. The syntactic category of the next word.
D. The meaning of the next sentence.
Answer: B
18. Which factor typically leads to slower reading times?
A. High word frequency
B. Short word length
C. A surprising or low-frequency word
D. High concreteness
Answer: C
19. The ‘MegaStudy’ approach (e.g., Balota et al., 2004) is used to:
A. Study a single participant reading for hundreds of hours.
B. Collect very large datasets across many words to distinguish correlated factors.
C. Measure brain activity only.
D. Test strictly serial processing models.
Answer: B
20. Age of acquisition (AoA) effects show that:
A. Words learned at age 50 are easier.
B. Early-acquired words can be processed faster even if they are currently lower in frequency.
C. Age of acquisition never matters once frequency is controlled.
D. Infants cannot acquire any words.
Answer: B
21. Forster’s (1976) serial search model proposed that word recognition is like:
A. A massive parallel search.
B. Looking up a word in a dictionary, checking entries one at a time.
C. Recognizing a face holistically.
D. Understanding a sentence.
Answer: B
22. Morton’s Logogen Model (1969) is an early example of what type of processing?
A. Serial processing
B. Parallel processing with independent “word detector” units
C. Strictly bottom-up processing with no top-down influence
D. A purely grammatical model
Answer: B
23. Connectionist models are also known as:
A. Serial search models
B. Box-and-arrow diagrams only
C. Parallel Distributed Processing (PDP) models or artificial neural networks
D. Purely behavioural models with no computational implementation
Answer: C
24. In the Interactive Activation and Competition (IAC) model, what are the three levels of representation (bottom to top)?
A. Words, letters, visual features
B. Visual features, letters, words
C. Phonemes, letters, meanings
D. Letters, words, sentences
Answer: B
25. In the IAC model, connections can be:
A. Only excitatory (positive).
B. Only inhibitory (negative).
C. Both excitatory and inhibitory.
D. Neither excitatory nor inhibitory; they are neutral.
Answer: C
26. The “Word Superiority Effect” refers to the finding that:
A. Words are remembered better than pictures.
B. A letter is identified more accurately when it appears in a real word than in a nonword.
C. High-frequency words are always superior to low-frequency words.
D. Spoken words have superior processing to printed words.
Answer: B
27. The IAC model explains the Word Superiority Effect via:
A. Slower serial checking.
B. Top-down feedback from word units reinforcing letter units.
C. A purely bottom-up feature detection mechanism.
D. Disconnection between the word and letter layers.
Answer: B
28. What is the “Ganong effect” (recapped from week 1) analogous to in reading?
A. The neighbourhood effect
B. The Word Superiority Effect and interactive feedback
C. The frequency effect
D. Eye movement regressions
Answer: B
29. “Slot-based coding” in the IAC model means:
A. Words are recognized as whole shapes.
B. Letter detectors are specific to a position in the word (e.g., C in position 1, C in position 2).
C. The model can read any font.
D. The model processes all letters simultaneously without position tracking.
Answer: B
30. A major limitation of the IAC model’s slot-based coding is that it:
A. Overestimates the tolerance for letter transpositions.
B. Cannot explain the Word Superiority Effect.
C. Is completely intolerant to letter switches; real readers can cope with transpositions.
D. Predicts that letter order is irrelevant.
Answer: C
31. The famous ‘Cambridge University’ scrambled-letter passage demonstrates that:
A. People cannot read words with transposed interior letters.
B. People can read text with transposed interior letters as long as first and last letters are in correct place.
C. Reading requires every letter to be in its exact position.
D. The IAC model perfectly predicts human reading behaviour.
Answer: B
32. According to the lecture, does the IAC model include phonological (sound) representations?
A. Yes, it was the first model to integrate phonology.
B. No, it is a limitation of the model’s scope; it only deals with orthographic form.
C. Yes, but only for regular words.
D. No, it was a deliberate theoretical claim that phonology plays no role in reading.
Answer: B
33. The statistician Box is quoted as saying, “All models are wrong, but some are interesting.” The lecture used this to mean:
A. The IAC model is useless.
B. Models, despite inaccuracies, can be illuminating and useful for understanding cognitive mechanisms.
C. We should only build models with no errors.
D. Only statistical models are valid.
Answer: B
34. For an English-speaking child learning to read, the mapping from print to meaning is considered harder initially than print-to-sound because:
A. Print-to-sound is fully consistent in English.
B. The print-to-meaning mapping is relatively arbitrary and the child already knows sound-to-meaning.
C. Children are not interested in meaning.
D. Sound-to-meaning mapping doesn’t exist yet.
Answer: B
35. Phonics instruction primarily targets which mapping?
A. Print to meaning directly
B. Sound to meaning
C. Print to sound
D. Meaning to syntax
Answer: C
36. The lecture described phonics knowledge as:
A. All you need for skilled reading comprehension.
B. Necessary but not sufficient for skilled reading.
C. Harmful to reading development.
D. Irrelevant to reading success.
Answer: B
37. A skilled adult reader encountering the novel word “wug” for the first time would rely most heavily on:
A. The direct print-to-meaning route.
B. The indirect route, sounding out the word.
C. Guessing solely from context.
D. Whole-word shape recognition.
Answer: B
38. Words like “yacht” that cannot be easily sounded out are argued to rely on which route in skilled readers?
A. Indirect phonological route
B. Direct print-to-meaning route
C. A tactile route
D. Only the visual feature level
Answer: B
39. In the lecture, “homonyms” are described as:
A. Words with highly related senses.
B. Words with identical spellings and sounds but distinct, unrelated meanings (e.g., trunk, bark).
C. Words that sound different but mean the same.
D. Newly invented words.
Answer: B
40. Polysemy is more common than homonymy and refers to:
A. A word having only one precise meaning.
B. Words with multiple, related senses (e.g., run).
C. The complete absence of ambiguity.
D. The sound structure of a word.
Answer: B
41. The 2016 SATs reading test sentence “Dawn was casting spun gold threads…” illustrates the challenge of:
A. Decoding irregular words.
B. Resolving pervasive ambiguity and selecting contextually appropriate meanings.
C. Phonics application to complex graphemes.
D. Understanding narrative plot.
Answer: B
42. According to consensus described in the lecture, upon encountering an ambiguous word, we:
A. Retrieve only the most frequent meaning.
B. Activate multiple meanings in parallel and then rapidly select the appropriate one.
C. Process meaning only after finishing the sentence.
D. Always consciously deliberate before selecting a meaning.
Answer: B
43. In Duffy et al. (1988), what words were “biased” ambiguous words?
A. Words with two equally common meanings, like “bark”.
B. Words with one very dominant meaning and one subordinate meaning, like “port”.
C. Words that had never been seen before.
D. Words that were always disambiguated early.
Answer: B
44. In Duffy et al. (1988), for early-disambiguation sentences, reading times on the ambiguous word were slower for biased words because:
A. The context was completely unhelpful.
B. The high-frequency dominant meaning interfered with selection of the context-supported subordinate meaning.
C. Participants couldn’t read the font.
D. Balanced words had no activated meanings.
Answer: B
45. The Reordered Access Model (Duffy et al., 1988) states that meaning access is modulated by:
A. Only sentence context.
B. Only dominance (meaning frequency).
C. Both dominance and sentence context, which interact.
D. The reader’s mood.
Answer: C
46. Rodd et al. (2016) studied changes in meaning preferences using:
A. Aphasic patients.
B. Recreational rowers who had acquired niche vocabulary.
C. Pre-lingual infants.
D. Artificial grammar learning.
Answer: B
47. The rowing study found that accessibility of rowing-specific meanings depended on:
A. Only recent same-day experience.
B. Only long-term years of experience.
C. Both recent and long-term experience.
D. The weather on the day of testing.
Answer: C
48. In the word-meaning priming experiment (Rodd et al., 2013), the effect of a single prior exposure on meaning selection:
A. Was permanent and never decayed.
B. Was a significant temporary boost compared to a no-prime control.
C. Actually decreased responding with that meaning.
D. Only worked in visual, not auditory, modality.
Answer: B
49. The term “lexical upkeep” metaphorically suggests that:
A. The lexicon is a static dictionary fixed by adulthood.
B. We must continuously update and maintain our mental lexicon to preserve comprehension.
C. Lexicons need to be physically cleaned.
D. Only children need to learn new words.
Answer: B
50. The lecture suggested that similar flexible learning mechanisms for ambiguity resolution may also apply to:
A. Visual object recognition only.
B. Grammatical (syntactic) ambiguity.
C. Only spoken language, not reading.
D. None of the above.
Answer: B
51. Why is eye-tracking a particularly useful method for studying meaning access during reading?
A. It measures brain activity directly.
B. It provides an online index of processing difficulty as participants control their reading pace.
C. It is cheaper than all other methods.
D. It relies on subjective reports.
Answer: B
52. The lecture argues that the introduction of phonics in UK schools was underpinned by:
A. Political ideology alone.
B. Basic cognitive science on the mechanisms of reading.
C. Parental pressure groups.
D. Financial incentives from publishers.
Answer: B
53. Which of the following is NOT a factor demonstrated to influence visual word recognition speed?
A. Word length
B. Concreteness
C. The font colour (provided contrast is sufficient)
D. Orthographic neighbourhood size
Answer: C
54. In the IAC model, activation spreads through the network dependent on:
A. The weather.
B. The connection strengths between units.
C. The number of letters in the word only.
D. The participant’s IQ.
Answer: B
55. The absence of gaps between words in spoken language is called:
A. The invariance problem
B. The segmentation problem
C. The Ganong problem
D. The phonics problem
Answer: B
56. According to the lecture, the fact that we perceive a familiar language as a series of discrete words (instead of continuous babble) arises because:
A. Speakers do insert tiny silent pauses we cannot consciously perceive.
B. Our brains actively impose structure by chunking the input using multiple cues.
C. All words are acoustically unique.
D. Written language has taught us the boundaries.
Answer: B
57. Which cue to word meaning selection involves knowledge about who is speaking?
A. Dominance
B. Recency
C. Knowledge about the speaker/writer (e.g., their accent or age)
D. Sentence context
Answer: C
58. What does the “phonics revolution” refer to in the context of the lecture?
A. A shift from teaching reading to teaching only listening skills.
B. A widespread change in reading instruction emphasizing systematic letter-sound correspondences.
C. A movement to stop teaching reading altogether.
D. Using technology to replace teachers.
Answer: B
59. The IAC model’s top-down feedback from word to letter units is functionally similar to which speech perception effect?
A. The McGurk effect
B. The Ganong effect
C. The phoneme restoration effect
D. Categorical perception
Answer: B
60. In computational modelling, what distinguishes a computational model from a purely verbal theory?
A. It is published in a journal.
B. It is implemented as a running computer program.
C. It always uses very few parameters.
D. It only uses brain imaging data.
Answer: B
61. The scrambled letter effect (reading transposed words) is a problem for the IAC model because the model:
A. Predicts that readers will be perfectly tolerant of all switches.
B. Predicts that all transpositions will be equally damaging, with no benefit for “right letter, wrong place”.
C. Cannot read any text with errors.
D. Assumes letters are processed as whole words.
Answer: B
62. Which dual-route is said to be vital when learning to read?
A. The direct route from print to meaning
B. The indirect route from print to sound (phonology)
C. The semantic route
D. The eye-movement route
Answer: B
63. “Phonology” in the context of reading refers to:
A. The visual shape of words.
B. The sound structure of words.
C. The meaning of words.
D. The grammatical rules.
Answer: B
64. A person who says “sabotage” as “satur-bugging” has likely:
A. Only ever heard the word spoken, never read it.
B. Only ever read the word, never heard it spoken, and misapplied spelling-to-sound rules.
C. A speech impediment.
D. Read the word in a font with poor letter clarity.
Answer: B
65. The concept of “parallel processing” in models like IAC means that:
A. Processing occurs one step at a time.
B. Many different units can be active and interact simultaneously.
C. Processing is always serial.
D. Only one word can be recognized at a time.
Answer: B
66. In the lecture, what is the “fovea”?
A. The part of the ear that processes sound.
B. The central portion of the retina providing detailed visual input.
C. A type of eye movement.
D. A brain area for meaning.
Answer: B
67. The fact that we can read about 4 words per second in fluent reading, often faster than listening to an audiobook, highlights that:
A. Reading is always a slower process.
B. Reading rate is under the reader’s control and can be very efficient.
C. Audiobooks are always poorly narrated.
D. Visual processing is slower than auditory.
Answer: B
68. The IAC model was developed in the:
A. 1960s
B. 1970s
C. 1980s
D. 2000s
Answer: C
69. The “connectionist bible” mentioned refers to works by:
A. Forster.
B. Morton.
C. McClelland & Rumelhart.
D. Duffy.
Answer: C
70. Meaning dominance, as a cue for disambiguation, relies on:
A. The visual salience of the word.
B. The relative frequency of the different meanings from long-term experience.
C. The number of letters in the word.
D. The emotional content of the word.
Answer: B
---
End of Questions.