Grammaticality Judgement Test
A grammaticality judgement test (GJT) presents participants with sentences and asks them to judge whether each is grammatically acceptable or unacceptable. It is one of the most widely used data collection instruments in SLA research, providing a window into learners' underlying linguistic knowledge without requiring production.
The Core Question
GJTs aim to access what learners know about the target language grammar, as distinct from what they can produce. A learner might fail to produce a correct relative clause in spontaneous speech but correctly judge one as grammatical on a GJT — revealing knowledge that is not yet available for production.
Timed vs Untimed GJTs
The critical design distinction is time pressure, which determines what type of knowledge is measured:
| Feature | Timed GJT | Untimed GJT |
|---|---|---|
| Time allowed | 3-5 seconds per item | Unlimited |
| Knowledge tapped | Implicit (automatised) | Explicit (analysed) |
| Processing | Fast, intuitive | Slow, reflective |
| Monitoring | Minimal — no time to apply rules | Maximum — learners can reason about rules |
R. Ellis (2005) demonstrated through factor analysis that timed GJTs loaded with other implicit knowledge measures (Elicited Imitation, oral narration), while untimed GJTs loaded with explicit knowledge measures (metalinguistic knowledge test). This validated the timed/untimed distinction as a methodological tool for separating the two knowledge types.
Item Types
- Grammatical sentences — correct items testing whether learners can recognise well-formed structures
- Ungrammatical sentences — items containing a specific violation (e.g., word order, morphology, agreement)
- Response format — binary (grammatical/ungrammatical), Likert scale, or binary with error identification and correction
Including a correction component strengthens the instrument: a learner who judges a sentence as ungrammatical and correctly identifies the error demonstrates more robust knowledge than one who merely rejects the sentence.
Concerns
- What is being judged? — participants may judge acceptability (pragmatic, stylistic) rather than grammaticality per se
- Guessing — binary forced-choice yields 50% accuracy by chance
- Construct validity — untimed GJTs may measure metalinguistic knowledge (knowledge about grammar) rather than functional linguistic knowledge
- Sentence decontextualisation — isolated sentences lack the discourse context of natural language use (see Ecological Validity)
- L1 influence — learners may judge sentences by comparison with L1 grammar rather than L2 rules
In SLA Research
GJTs have been central to debates about:
- Interlanguage grammar — what learners know at different developmental stages
- Universal Grammar access — whether L2 learners show sensitivity to UG constraints
- The interface debate — whether explicit knowledge can become implicit
- Corrective Feedback effects — whether treatment changes underlying representations or only production
Key References
- Ellis (2005) — GJTs in the implicit/explicit knowledge test battery
- Gass (1994) — methodological issues with GJTs
- Tremblay (2005) — timed and untimed GJTs as distinct measures
- Vafaee, Suzuki & Kachinske (2017) — construct validity of GJTs