Writing quality is often reduced to a simple idea: students with better English use more sophisticated language. But linguistic complexity is a surprisingly slippery concept. It isn’t just about using longer words or longer sentences. It spans vocabulary, syntax, cohesion, coherence & the relationships between them.
A recent study by Jinhua Zhang, published in PLOS One, explored this question in depth by examining how linguistic complexity differs across EFL learners at different proficiency levels. Rather than focusing on a single feature, the study looked simultaneously at lexical, syntactic & textual complexity, offering a more complete picture of what distinguishes stronger writers from weaker ones.
The Study
Zhang analysed 66 essays written by Chinese university students studying at an application-oriented university. The essays were written under timed conditions (60 minutes) on the topic of cloud tourism (sometimes called virtual tourism or online tourism). After being rated by two experienced assessors, the essays were divided into three equally sized groups: high-scoring, middle-scoring & low-scoring writers.
The researcher then built a corpus of over 22,000 words & analysed the texts across three dimensions:
- Lexical complexity (vocabulary diversity, sophistication, density, frequency profiles)
- Syntactic complexity (sentence length, clause length, T-units, clause density, syntactic variation)
- Textual complexity (cohesion, coherence, semantic connectedness)
A range of corpus-analysis tools including Range, Coh-Metrix & the L2 Syntactic Complexity Analyzer were used to identify patterns across proficiency levels.
The Findings
1. Not all complexity measures matter equally
One of the clearest findings was that some measures were much better at distinguishing stronger writers than others.
For vocabulary, lexical sophistication mattered more than vocabulary diversity. High-scoring writers used substantially more low-frequency vocabulary. Around 21.3% of their word types came from academic or less-frequent vocabulary, compared with 16.3% for middle-level writers & 14.6% for lower-level writers.
By contrast, vocabulary diversity itself showed no significant differences between groups. In other words, simply using a wider range of words wasn’t enough.
Imagine two learners writing about remote working:
- “People can work from home because technology is good.”
vs
- “Digital platforms enable employees to collaborate remotely with increased flexibility.”
The second example isn’t necessarily more diverse lexically, but its vocabulary is noticeably more sophisticated.
2. Development isn’t linear
Perhaps the most interesting finding was that both vocabulary & syntax appeared to develop in non-linear ways.
The middle-proficiency group often looked surprisingly similar to the low-proficiency group. Large jumps tended to appear only among the highest-performing writers.
This suggests a possible “threshold” effect. Progress may accumulate gradually below the surface before becoming visible in writing performance.
The finding echoes ideas from Complexity Theory in SLA, where development is often viewed as uneven, dynamic & characterised by sudden leaps rather than smooth progression.
3. Longer & more varied sentences helped
The strongest writers consistently produced:
• longer sentences
• longer clauses
• longer T-units (a main clause plus any dependent clauses attached to it).
• greater syntactic variety
High-scoring writers differed significantly from low-scoring writers across most syntactic measures. Mean sentence length & clause length were among the strongest predictors of writing quality.
However, one finding stood out.
Using more subordinate clauses alone did not automatically improve writing quality. The dependent-clause ratio showed no significant differences between groups. In practical terms, this means that writing quality isn’t improved by mechanically adding relative clauses or subordinators. Quality depends on how effectively those structures carry meaning.
4. Coherence beat cohesion
For many teachers, the most thought-provoking result may be the distinction between cohesion & coherence.
Lower-scoring writers used far more visible cohesive devices, especially pronouns & conjunctions. They often relied on “because”, “and”, “so”, “however” & similar connectors to create links.
Yet stronger writers displayed higher levels of overall semantic coherence rather than simply more connecting words.
Think of the difference between: “Firstly… Secondly… Thirdly… Therefore…” & a text whose ideas naturally build on one another even without constant signalling.
The study suggests weaker writers often focus on surface connections, while stronger writers focus on deeper conceptual organisation.
This aligns closely with research by Halliday & Hasan on cohesion, McNamara et al. on writing quality & Crossley et al. on discourse features, all of whom have highlighted the limits of judging writing purely through visible cohesive markers.
Teacher Takeaways?
- Don’t equate “better vocabulary” with simply “more vocabulary”. Help learners notice how low-frequency, precise vocabulary contributes to writing quality.
- Treat cohesion & coherence as different beasts. A paragraph full of linking words is not necessarily a coherent paragraph.
- Expect uneven development. Progress in writing may appear stalled for long periods before important breakthroughs emerge.
One broader implication stood out to me: many of our assessment practices still reward visible linguistic features more easily than deeper organisational qualities. Yet this study suggests that what separates stronger writers may be less about how many language features they use & more about how effectively those features work together.
What if the biggest difference between weaker & stronger writers isn’t linguistic complexity itself, but their growing ability to organise meaning?
Leave a Reply