Definition:
Complexity, Accuracy, and Fluency (CAF) is the standard framework for measuring the quality of L2 spoken and written production in research and assessment. Each dimension captures a different aspect of learner performance: complexity reflects grammatical range and sophistication; accuracy reflects error-free production; fluency reflects ease, speed, and automaticity.
In-Depth Explanation
Complexity:
Refers to the breadth and sophistication of the grammatical structures and vocabulary a learner uses. Measured by subordination ratios (subordinate clauses per clause), vocabulary richness scores (type-token ratio), and structural variety. High complexity indicates the learner is using more varied and elaborate language.
Accuracy:
Refers to freedom from errors — grammatical, lexical, and phonological. Measured by error-free clauses per total clauses, or error counts per hundred words. High accuracy indicates the learner has mastered the forms they use.
Refers to smooth, automatic production without excessive pause, reformulation, or slowdown. Measured by speech rate (words per minute), mean length of runs between pauses, and frequency of repair sequences. High fluency indicates automaticity in production.
The CAF trade-off hypothesis:
A central finding in CAF research (Skehan, 1998; Robinson, 2001) is that learners trade off dimensions under task conditions: when pushed to be more fluent, accuracy often drops; when pushed for complexity, both accuracy and fluency may suffer. This trade-off occurs because all three dimensions compete for limited working memory and attention in real-time production.
CAF and task design:
Task-based language teaching research uses CAF to assess the effects of task design variables (complexity, familiarity, planning time, structure) on learner production. Pre-task planning time, for instance, consistently improves both accuracy and fluency without necessarily increasing complexity.
History
- 1980s–1990s: SLA researchers adopt operational measures of complexity, accuracy, and fluency as standard production metrics, allowing comparison across studies.
- 1998: Skehan formalizes the CAF framework in A Cognitive Approach to Language Learning, arguing for a limited attention capacity model driving trade-offs.
- 2001: Robinson proposes an alternative Cognition Hypothesis, arguing that more complex tasks can increase all three dimensions simultaneously by forcing deeper processing.
- Present: CAF remains the standard output measurement framework, though debate continues about how to operationalize each dimension and how they interact.
Common Misconceptions
“Fluency means speaking quickly and without hesitation.” In the CAF framework, fluency refers to the smoothness and continuity of speech delivery — but this includes natural hesitations that allow the speaker to plan complex language. Fluency is not simply speaking fast; a learner who speaks rapidly with many fillers and false starts may score lower on fluency measures than a slower speaker who maintains forward movement without frequent breakdown.
“Higher complexity always indicates better learning.” Complexity in the CAF sense refers to the range and sophistication of grammatical structures used — not necessarily to comprehensibility or communicative effectiveness. A learner who uses very complex embedded clauses that obscure meaning is not necessarily a better communicator than one who uses simpler, clearer structures accurately.
Criticisms
The CAF trade-off hypothesis (Skehan, 1998) — which predicts that learners cannot focus equally on all three dimensions simultaneously and will sacrifice one to achieve another — has been empirically contested. Studies show that CAF dimensions interact in complex, task-specific ways that are difficult to predict from general principles. Different operationalizations of complexity, accuracy, and fluency across studies make meta-analytic comparison difficult. Critics also note that CAF measures are production-focused and do not directly reflect underlying linguistic competence development.
Social Media Sentiment
CAF as a framework appears in language teaching professional communities — teacher training discussions, academic course materials, and applied linguistics blogs. Among learner-facing communities, the underlying concepts appear in advice about balancing grammar accuracy with speaking fluency, particularly in discussions about when to prioritize speaking practice over error correction. The advice “don’t be afraid to make mistakes — focus on fluency first” implicitly draws on CAF thinking without using the technical framework.
Last updated: 2026-04
Practical Application
For language learners:
- When starting to speak or write in L2, prioritize fluency first (get comfortable with real-time production) rather than stopping constantly for accuracy
- Once fluency is comfortable, shift attention toward accuracy — self-correct more carefully, use focused output practice
- Complexity develops naturally with vocabulary and grammar growth from extensive input and SRS review
- Sakubo
Related Terms
See Also
Research
- Skehan, P. (1998). A Cognitive Approach to Language Learning. Oxford University Press. [Summary: Defines CAF as a three-dimensional production framework and proposes the Trade-Off Hypothesis — that attention trade-offs between dimensions occur under task conditions due to limited working memory capacity.]
- Robinson, P. (2001). Task complexity, task difficulty, and task production: Exploring interactions in a componential framework. Applied Linguistics, 22(1), 27–57. [Summary: Challenges the trade-off model, arguing that increased task complexity can enhance all three CAF dimensions by deepening cognitive processing and engagement.]
- Housen, A., & Kuiken, F. (2009). Complexity, accuracy, and fluency in second language acquisition. Applied Linguistics, 30(4), 461–473. [Summary: Overview of CAF as a research framework, discussing operationalization challenges, cross-study comparisons, and the relationship between CAF dimensions and SLA theory.]