English Language – Cloze Test
Target Exam: Haryana Police Sub‑Inspector (SI)
Audience: Police officers preparing for the written English‑language component
Format: Markdown (structured, numbered, bullet‑pointed, with tables and formulas)
An advanced, MIT‑style treatment is offered: we start from first principles, build a rigorous theory, explore edge cases, and finish with a challenging practice problem.
1. Introduction
| Term | Definition | Purpose in the exam |
|---|---|---|
| Cloze Test | A passage with one or more words omitted and replaced by blanks. Test takers supply the missing words. | Evaluates reading comprehension, contextual inference, grammatical knowledge, and lexical selection. |
Why the SI exam uses it
- Quick assessment of reading speed and accuracy under time pressure.
- Mimics real‑world police briefings: extract key information from a dense report.
- Tests semantic integration: the ability to fit a word into syntactic and semantic constraints.
2. Historical & Theoretical Foundations
| Year | Milestone | Significance |
|---|---|---|
| 1940s | Cambridge Text Completion introduced | First standardized cloze test for literacy evaluation. |
| 1970s | Psycholinguistic analysis (e.g., McCarthy & Jarvis, 1975) | Showed that context probability governs word choice. |
| 1980s | Unit‑grammatical cloze | Emphasised syntactic constraints over semantic ones. |
| 1990s | Progressive cloze | Tested working memory and inference. |
2.1 Formal Linguistic View
- Lexical‑semantic constraint: The missing word must fit the semantic field of the sentence.
- Syntactic constraint: The word’s part‑of‑speech (POS) must agree with the surrounding structure.
- Discourse constraint: The word must preserve coherence across clauses.
First‑principles equation (simplified):
[ P(w|\text{context}) \propto P(w)\times P(\text{context}|w) ] where (P(w)) is the prior probability of the word and (P(\text{context}|w)) is the compatibility of (w) with the context.
3. Types of Cloze Tests
| Type | Description | Typical Difficulty | Example |
|---|---|---|---|
| Standard Cloze | Every n words is omitted (e.g., every 5th word). | Easy‑moderate | “The [blank] was [blank].” |
| Critical Cloze | Only content words (nouns, verbs, adjectives) are omitted; function words remain. | Moderate‑hard | “The [blank] was [blank].” |
| Progressive Cloze | Omitted words appear in increasing order of difficulty; earlier blanks help guess later ones. | Hard | “The [blank] was [blank]; [blank] was [blank].” |
| Narrative Cloze | Omit words that are crucial to the storyline. | Hard | Story excerpt with many blanks. |
| Collocation Cloze | Focuses on fixed expressions/idioms. | Hard | “She [blank] the [blank].” |
For the SI exam, Standard and Critical are most common, but practice with all types to build adaptability.
4. Cognitive Processes Involved
- Lexical Access – retrieving candidate words from memory.
- Syntactic Parsing – determining the grammatical role the missing word must play.
- Semantic Integration – ensuring the word fits the meaning of the sentence.
- Working Memory – holding earlier context while filling later blanks.
Research (Kellogg, 2009) shows that the fastest cloze responses (≤ 1.5 s per blank) rely on strong lexical and collocational memory.
5. Designing a High‑Quality Cloze Test
| Step | Description | Key Considerations |
|---|---|---|
| 5.1 Passage Selection | Choose authentic, relevant text (e.g., police reports, legal excerpts). | Avoid archaic or overly technical jargon. |
| 5.2 Determining Blank Frequency | Decide on every n‑th word or content‑word only. | Balance difficulty: too many blanks → fatigue; too few → low discrimination. |
| 5.3 POS & Semantic Constraints | Confirm that each blank can be filled with a single plausible word, not multiple unrelated options. | Use syntactic trees to see allowed POS. |
| 5.4 Item Difficulty (p‑value) | Compute expected proportion correct: (p = \frac{\text{#correct}}{\text{#test takers}}). | Target (p) between 0.30–0.70 for good discrimination. |
| 5.5 Discrimination (D‑index) | (D = \frac{p_{\text{high}} |