Reading Comprehension – Advanced Study Notes
(MIT‑Level, First‑Principles Treatment)
Audience: Graduate students in English Linguistics / Cognitive Science
Purpose: A rigorous, theory‑driven foundation for reading‑comprehension research and instruction.
Format: Markdown, self‑contained, with equations, diagrams (ASCII), and a challenging problem set.
1. Introduction
| Question | Why is Reading Comprehension Fundamental? |
|---|---|
| What is reading? | The orthographic → phonological → semantic mapping of a text onto a mental representation. |
| Why study it? | It is the gatekeeper of knowledge acquisition, critical thinking, and communication. |
| How complex is it? | It integrates perceptual, cognitive, linguistic, and cultural processes in a single, dynamic system. |
2. First‑Principles Framework
2.1. The Core Mapping
[ \text{Input (orthography)} \xrightarrow{\text{Decoding}} \text{Phonology} \xrightarrow{\text{Lexical Access}} \text{Syntax} \xrightarrow{\text{Semantics}} \text{Pragmatics} \xrightarrow{\text{Metacognition}} \text{Comprehension} ]
- Decoding: Grapheme‑to‑phoneme conversion, visual‑motor integration.
- Lexical Access: Retrieval of word form, definition, usage.
- Syntax: Parsing of sentence structure.
- Semantics: Meaning construction (sense, reference).
- Pragmatics: Contextual inference, speaker intent.
- Metacognition: Self‑monitoring, strategy selection.
2.2. Information‑Theoretic View
Reading can be modelled as a noisy communication channel:
[ \text{Text} \xrightarrow{C} \text{Reader’s Internal Representation} ]
- Shannon’s Entropy (H = -\sum p_i \log_2 p_i) measures uncertainty of the next word or syntactic structure.
- Mutual Information (I(X;Y)) quantifies how much the reader’s internal state reduces uncertainty about the text.
- Error Correction: Working memory, context, prior knowledge act as error‑correcting codes.
3. Theoretical Models of Reading Comprehension
| Model | Core Idea | Key Equations / Mechanisms | Strengths | Limitations |
|---|---|---|---|---|
| Simple View of Reading (SVR) | Comprehension = Decoding × Linguistic Comprehension | (C = D \times L) | Clear diagnostic framework | Oversimplifies interaction effects |
| Dual‑Route Cascaded (DRC) | Lexical and non‑lexical routes in parallel | (L_{lex} = f(GR)), (L_{nonlex} = f(P)) | Explains irregularities | Requires many parameters |
| Connectionist Models (e.g., ACT‑R) | Parallel distributed processing | (\Delta w_{ij} = \eta (x_i y_j - w_{ij})) | Handles learning & memory decay | Hard to interpret |
| Interactive Activation Model | Bottom‑up + top‑down interactions | (\dot{R}i = \sum_j w{ij} R_j - \alpha R_i) | Captures context effects | Parameter tuning needed |
| Predictive Coding (Bayesian) | Brain predicts next input | (P(\theta | d) \propto P(d | \theta)P(\theta)) |
4. Cognitive Processes Involved
-
Attention & Working Memory
- Load capacity ( \approx 7 \pm 2 ) items (Miller, 1956).
- Executive control: selective attention, inhibition, switching (Baddeley, 2003).
-
Lexical Retrieval
- Lexicon as a network: Node activation spreads.
- Frequency effect: ( t \propto \frac{1}{\log f} ) – faster access for frequent words.
-
Sentence Parsing
- Incremental parsing: Garden‑path resolution.
- Constraint‑based vs rule‑based parsing.
-
Semantic Integration
- Semantic memory: Retrieval of conceptual knowledge.
- Integrative combinatory semantics: Lambda calculus representation.
-
Pragmatic Inference
- Relevance Theory (Sperber & Wilson, 1989).
- Theory of Mind for speaker intent.
6