Study Notes – Idioms & Phrases
Program: Jail Warder
Subject: English Language
Level: MIT‑Level (graduate/advanced undergraduate)
These notes aim to provide a rigorous, first‑principles treatment of idioms and phrases, blending linguistic theory, cognitive science, and computational linguistics. The material is intended for students who already have a solid grounding in semantics, syntax, and psycholinguistics.
1. Introduction
Idioms are fixed expressions whose meanings are not deducible from the literal meanings of their constituents. They are a central phenomenon in natural language, illustrating how language can be both compositional and non‑compositional. A phrase is a syntactically coherent unit that can be a constituent (e.g., a noun phrase) or a multi‑word expression that behaves as a single semantic unit (e.g., kick the bucket).
Why study idioms?
- Cognitive insight – idioms reveal how the brain stores and retrieves knowledge.
- Linguistic theory – they challenge compositionality principles and drive developments in lexical semantics.
- Applied AI – idiom translation, speech recognition, and natural‑language generation all struggle with idiomaticity.
2. Definition & Scope
| Concept | Definition | Example |
|---|---|---|
| Idiom | A multi‑word expression whose overall meaning cannot be deduced from its parts. | Break the ice → “Initiate social interaction.” |
| Fixed Expression | A phrase that is highly resistant to word order change. | In the blink of an eye |
| Semi‑Fixed / Collocation | An expression that shows high lexical association but can still be partially recomposed. | Strong coffee |
| Literal Phrase | A multi‑word unit whose meaning is compositional. | The cat sat on the mat |
Edge‑case note: Some idioms have become so conventionalized that they are treated as lexical items (e.g., to be on cloud nine).
3. Historical Development
- Early Lexicography – 19th‑century dictionaries list idioms as “phrases of commonplace usage.”
- Structural Linguistics – 1950s focus on constituent structure; idioms were treated as “lexicalized" instances of syntactic templates.
- Cognitive Semantics – 1980s onward, the Conceptual Metaphor Theory (Lakoff & Johnson, 1980) explains idioms as metaphorical transfers.
- Formal Semantics – 1990s: Montague Grammar and Lambda Calculus attempted to capture idiomatic meaning via non‑compositional rules.
- Computational Linguistics – 2000s: distributional semantics, machine learning, and neural networks tackle idiom detection and generation.
4. Classification of Idioms
| Class | Properties | Example | Theoretical Lens |
|---|---|---|---|
| Literal Idioms | Occasionally used literally; high semantic ambiguity. | Run over (to exceed a limit vs. to drive over). | Pragmatics |
| Non‑Literal Idioms | Strictly idiomatic; no literal sense. | Spill the beans (reveal secret). | Conventionalization |
| Process Idioms | Describe actions. | Hit the road. | Frame semantics |
| Result Idioms | Denote outcomes. | Bite the bullet. | Reification |
| Lexical Idioms | Single lexical unit. | Heart‑broken. | Lexicon |
| Phrasal Idioms | Multi‑word. | Kick the bucket. | Syntax + semantics |
5. Semantic Theories
5.1 Conventionalization & Prototype Theory
Idioms are viewed as prototype-based items: the canonical form is stored as a unit in memory.
- Cognitive Basis: Cognitive Load Theory suggests that highly conventionalized expressions reduce processing load.
- Empirical Evidence: Faster reaction times in lexical decision tasks for idioms (e.g., spill the beans vs. spill the water).
5.2 Conceptual Metaphor Theory (CMT)
Idioms often embody metaphorical mappings.
- “Break the ice” → Social Interaction → Ice as barrier.
- “Kick the bucket” → Death → Bucket as life support.
5.3 Frame Semantics
Idioms evoke a frame (S. Fillmore).
- “Get the ball rolling” evokes the Initiation frame: ball → activity.
5.4 Distributional Semantics
Modern embeddings encode idiom usage patterns.
- Challenge: Idioms often have low frequency → sparse data.
- Solution: Idiomaticity detection via contextualized embeddings (BERT, GPT) captures subtle co‑occurrence patterns.