Original data analysis
What 24 recent CELPE-BRAS tasks reveal
We analyzed the 24 written tasks in our curated practice corpus, covering both editions from 2023, 2024, and 2025. Every record links back to the official UFRGS archive so readers can reproduce the counts.
Trigger-type distribution
Half of the selected tasks begin with written material; audio and video together make up the other half.
| Type | Tasks | Share |
|---|---|---|
| text | 12 | 50% |
| video | 6 | 25% |
| audio | 6 | 25% |
Required writing genres
Genre labels follow the normalized labels used in this platform's task catalog.
| Type | Tasks | Share |
|---|---|---|
| texto divulgacao | 7 | 29.2% |
| 6 | 25% | |
| carta | 6 | 25% |
| artigo | 3 | 12.5% |
| cronica | 1 | 4.2% |
| carta aberta | 1 | 4.2% |
Methodology and limitations
The unit of analysis is one written task. We counted the year, edition, task number, trigger type, and normalized target genre stored in the platform catalog. The sample contains four tasks from each of six editions. It excludes oral-interaction prompts and all editions outside 2023–2025. Labels were reviewed on 21 August 2026 against the linked archive pages. Because the corpus is intentionally bounded, the findings must not be generalized to the full exam history.
Download the record-level JSON dataset
Published and last reviewed: 21 August 2026