Reproducibility, Recognition, and InterAb 2026

This edition marks a meaningful moment for Cadernos de Linguística. The first Registered Report protocols from our reproducibility theme issue are now online, alongside a research report that brings Portuguese into the heart of the debate on large language models. We also celebrate the Abralin Award for Best Article, won by a paper published in our journal, and share highlights from CadLin’s participation at InterAb 2026. Open calls for theme issues are listed at the end.

The theme issue Reproducibility in Linguistic Research is taking shape with the publication of its first three papers. Two are Registered Report Protocols — a format in which the research project is peer-reviewed and approved before data collection begins, with hypotheses, methods, and the analysis plan on public record from the start.

Why does this matter for open science? Registered Reports separate the research plan from its results, which reduces the incentive for post-hoc adjustments and makes the scientific process easier to trace and evaluate. Pre-registration is not a formality: it is a public commitment to answering specific questions in a specific way, independent of what the data turn out to show. The final reports — with collected and analysed data — will follow. Stay tuned.

Replicating the Effects of Iconicity in Lexical Decision Task: A Study in Brazilian Portuguese, by Mahayana C. Godoy, Thayná C. Ananias, and David M. Sidhu, sets out to replicate a classic experiment on lexical iconicity with Brazilian Portuguese speakers. Iconic words — such as sussurro (”whisper”), chiado (”hiss”), or zanzar (”to wander”) — bear a resemblance between sound and meaning that, in English, has been shown to speed up visual word recognition. Whether this holds for Portuguese remains an open question. The protocol lays out the stimulus selection criteria, experimental procedure, and statistical analysis plan that will guide the study.

Alves et al., from the University of São Paulo, present the protocol Testing the Efficiency of a Machine Learning-Based Automatic Prosodic Segmentation Method for Brazilian Portuguese. The study replicates, on a new corpus of spontaneous speech, a Random Forest classifier trained to identify terminal prosodic boundaries in Brazilian Portuguese. Testing the model on data outside its original training set — the NURC Minimal Corpus, with over 17 hours of recordings from the 1970s — allows the authors to assess how well the method holds up across different speakers and audio conditions.

Word Predictability in Portuguese: Cloze Norming Study vs. LLMs, by Jane Aristia (Université Paris Nanterre), compares a traditional cloze study — in which participants complete sentences — with the predictions of two Portuguese language models, Grevásio and Tucano. The correlations between human and model probabilities are weak to moderate, suggesting that, at least for Portuguese, LLMs are not yet a substitute for human participants in this kind of task. They may, however, be useful as complementary data.

Cadernos de Linguística is proud to share that the Abralin Award for Best Article has been awarded to Exploring variation and change in the distant past: /s/-variation and its implications for the reconstruction of Ancient Babylonian, by Rodrigo Hernáiz. The announcement was made by ABRALIN at InterAb 14, in the award’s second edition.

The paper examines variation in the written representation of /s/ in Ancient Babylonian, demonstrating that the variation is not random: it correlates statistically with geographical and temporal factors, points to a historical process of diffusion, and sheds light on the reconstruction of one of the oldest writing traditions in the world. ABRALIN singled out the work for its originality, analytical consistency, and the way it brings together variation and change, historical reconstruction, and computational analysis — connecting longstanding questions in Historical Linguistics and Sociolinguistics with contemporary quantitative methods.

Read the full announcement on the Cadernos de Linguística website.

CadLin participated in two roundtables at InterAb 2026 that addressed questions central to linguistic research and academic publishing in Brazil.

The roundtable Reproducibility and quantitative methods in linguistics gave us the opportunity to discuss CadLin’s experience with the Registered Reports format and the practical challenges of building open science practices in a field still working through what those practices mean.

The roundtable Evaluation policies, knowledge circulation, and challenges for Language Studies brought together a broader set of voices, including Dr. Antônio Gomes de Souza Filho, Director of Postgraduate Evaluation at CAPES. The conversation ranged across editorial practices, assessment criteria, and the place of the open science model in the Brazilian academic context — and kept returning to the tensions between established evaluation systems and the demands for greater transparency, data sharing, and reproducibility.

CadLin’s open calls are standing invitations for the academic community to shape the field’s conversations. The following are currently open:

Diachronic Construction Grammar — language change, cognition, and usage in a constructional perspective. Learn more.

Cognitive Linguistics in the Real World — research on language as it is actually used. Learn more.

Roots to New Directions: The Trajectory of Brazilian Dialectology — contributions engaging with the different approaches to Dialectology in Brazil. Learn more.

Media, Technologies, and Artificial Intelligence — research on media, technologies, and AI in political-discursive processes. Learn more.

Contemporary Historiographies of Linguistic Studies and Language Teaching — trajectories of knowledge and practice in language and teaching. Learn more.

Browse all open calls at cadernos.abralin.org.

CadLin’s work goes beyond publishing papers. We have a genuine commitment to helping the community learn what open and reproducible science looks like in practice — because we know that for many researchers, these models and formats are still new territory.

When a submission does not yet meet our requirements, the editorial team does not simply turn it away. We work through the issues with authors, showing them concretely what needs to change and why. The goal is to bring good research into a framework that makes it more transparent, more reusable, and easier for others to build on.

A reminder about what we publish: CadLin exclusively accepts Registered Reports (Protocols and full Registered Reports), reproducibility reports, critical perspectives, and papers submitted to theme issues. Please read our submission guidelines before submitting.

CadLin supports format-free submission — no house style required at the submission stage. Find out how to submit your manuscript here.