Study Notes – Cloze Test
Subject: English Language (Nagaland TET)
Level: Advanced / MIT‑style
Format: Markdown
1. Introduction
| Item | Description |
|---|---|
| What is a cloze test? | A reading‑comprehension instrument in which words are omitted from a text and the test‑taker must supply the missing words. |
| Why it matters for TET? | Gauges inferential and contextual‑reading skills, lexical knowledge, syntactic awareness, and ability to use discourse cues. |
| Typical structure (Nagaland TET) | 1–2 short passages (≈120–150 words). 5–10 blanks. 4‑option multiple choice per blank. |
2. Theoretical Foundations
Cloze tests sit at the intersection of psycholinguistics, computational linguistics, and psychometrics.
They exploit the human ability to fill gaps by leveraging contextual constraints.
2.1 Linguistic Basis
| Constraint | What it provides | Example |
|---|---|---|
| Lexical | Word frequency, distributional semantics | “The cat sat on the mat.” → cat is high‑frequency in the context. |
| Syntactic | Part‑of‑speech (POS) compatibility, grammatical roles | “The cat sat on the mat.” → sat must be a verb. |
| Semantic | Coherence, thematic progression | “The cat sat on the mat, and it began to sleep.” |
| Pragmatic | World knowledge, discourse expectations | “The cat sat on the mat, so it was comfortable.” |
2.2 Psychometric Basis
- Item Response Theory (IRT) models the probability of a correct response as a function of latent ability (θ).
- For a 2‑parameter logistic (2PL) model:
[ P_{i}(\theta) = \frac{1}{1 + e^{-a_i(\theta - b_i)}} ]
- (a_i) = discrimination (slope)
- (b_i) = difficulty (point where (P=0.5))
- Graded Response Model (GRM) can handle multiple‑choice items with ordered options.
2.3 Cognitive Load
- Intrinsic load – complexity of the removed word’s semantic/syntactic role.
- Extraneous load – passage length, word density, cross‑sentence dependencies.
- Germane load – benefits from productive inference.
3. Design Principles – From First Principles
3.1 Passage Construction
| Rule | Rationale | Example |
|---|---|---|
| Select a coherent, domain‑neutral text | Reduces cultural bias. | A short paragraph about a simple science experiment. |
| Maintain grammaticality after removal | Avoid ambiguous or ungrammatical sentences. | “The teacher explained the concept.” → Remove teacher but keep “The ___ explained the concept.” |
| Ensure each blank has a single plausible answer | Prevents guessing. | Use a unique noun or verb that fits the syntax. |
| Balance difficulty | Avoid clustering easy or hard items. | Distribute high‑frequency and low‑frequency words. |
| Control for semantic drift | Preserve overall meaning. | Do not remove core content words that change the passage’s message. |
3.2 Word Selection Strategy
- POS Tagging – Identify nouns, verbs, adjectives that anchor meaning.
- Frequency Analysis – Use corpora (e.g., British National Corpus) to rank candidate words.
- Semantic Role Labeling – Ensure the blank’s role (agent, patient) is clear.
- Local vs. Global Context – Prefer words that can be inferred from the immediate sentence (local) for easier items; global context for harder items.
3.3 Difficulty Calibration
| Stage | Activity | Tool |
|---|---|---|
| Pilot testing | Administer to a sample of 200 |